MIT News - CSAIL - Robotics - Computer Science and Artificial Intelligence Laboratory (CSAIL) - Robots - Artificial intelligence MIT News is dedicated to communicating to the media and the public the news and achievements of the students, faculty, staff and the greater MIT community. en Mon, 09 Mar 2020 00:00:00 -0400 The elephant in the server room Catherine D’Ignazio’s new book, “Data Feminism,” examines problems of bias and power that beset modern information. Mon, 09 Mar 2020 00:00:00 -0400 Peter Dizikes | MIT News Office <p>Suppose you would like to know mortality rates for women during childbirth, by country, around the world. Where would you look? One option is the <a href="" target="_blank">WomanStats</a> Project, the website of an academic research effort investigating the links between the security and activities of nation-states, and the security of the women who live in them.</p> <p>The project, founded in 2001, meets a need by patching together data from around the world. Many countries are indifferent to collecting statistics about women’s lives. But even where countries try harder to gather data, there are clear challenges to arriving at useful numbers — whether it comes to women’s physical security, property rights, and government participation, among many other issues. &nbsp;</p> <p>For instance: In some countries, violations of women’s rights may be reported more regularly than in other places. That means a more responsive legal system may create the appearance of greater problems, when it provides relatively more support for women. The WomanStats Project notes many such complications.</p> <p>Thus the WomanStats Project offers some answers — for example, Australia, Canada, and much of Western Europe have low childbirth mortality rates — while also showing what the challenges are to taking numbers at face value. This, according to MIT professor Catherine D’Ignazio, makes the site unusual, and valuable.</p> <p>“The data never speak for themselves,” says D’Ignazio, referring to the general problem of finding reliable numbers about women’s lives. “There are always humans and institutions speaking for the data, and different people have their own agendas. The data are never innocent.”</p> <p>Now D’Ignazio, an assistant professor in MIT’s Department of Urban Studies and Planning, has taken a deeper look at this issue in a new book, co-authored with Lauren Klein, an associate professor of English and quantitative theory and methods at Emory University. In the book, “<a href="" target="_blank">Data Feminism</a>,” published this month by the MIT Press, the authors use the lens of intersectional feminism to scrutinize how data science reflects the social structures it emerges from.</p> <p>“Intersectional feminism examines unequal power,” write D’Ignazio and Klein, in the book’s introduction. “And in our contemporary world, data is power too. Because the power of data is wielded unjustly, it must be challenged and changed.”</p> <p><strong>The 4 percent problem</strong></p> <p>To see a clear case of power relations generating biased data, D’Ignazio and Klein note, consider research led by MIT’s own Joy Buolamwini, who as a graduate student in a class studying facial-recognition programs, observed that the software in question could not “see” her face. Buolamwini found that for the facial-recognition system in question, the software was based on a set of faces which were 78 percent male and 84 percent white; only 4 percent were female and dark-skinned, like herself.&nbsp;</p> <p>Subsequent media coverage of Buolamwini’s work, D’Ignazio and Klein write, contained “a hint of shock.” But the results were probably less surprising to those who are not white males, they think.&nbsp;&nbsp;</p> <p>“If the past is racist, oppressive, sexist, and biased, and that’s your training data, that is what you are tuning for,” D’Ignazio says.</p> <p>Or consider another example, from tech giant Amazon, which tested an automated system that used AI to sort through promising CVs sent in by job applicants. One problem: Because a high percentage of company employees were men, the algorithm favored men’s names, other things being equal.&nbsp;</p> <p>“They thought this would help [the] process, but of course what it does is train the AI [system] to be biased toward women, because they themselves have not hired that many women,” D’Ignazio observes.</p> <p>To Amazon’s credit, it did recognize the problem. Moreover, D’Ignazio notes, this kind of issue is a problem that can be addressed. “Some of the technologies can be reformed with a more participatory process, or better training data. … If we agree that’s a good goal, one path forward is to adjust your training set and include more people of color, more women.”</p> <p><strong>“Who’s on the team? Who had the idea? Who’s benefiting?” </strong></p> <p>Still, the question of who participates in data science is, as the authors write, “the elephant in the server room.” As of 2011, only 26 percent of all undergraduates receiving computer science degrees in the U.S. were women. That is not only a low figure, but actually a decline from past levels: In 1985, 37 percent of computer science graduates were women, the highest mark on record.</p> <p>As a result of the lack of diversity in the field, D’Ignazio and Klein believe, many data projects are radically limited in their ability to see all facets of the complex social situations they purport to measure.&nbsp;</p> <p>“We want to try to tune people in to these kinds of power relationships and why they matter deeply,” D’Ignazio says. “Who’s on the team? Who had the idea? Who’s benefiting from the project? Who’s potentially harmed by the project?”</p> <p>In all, D’Ignazio and Klein outline seven principles of data feminism, from examining and challenging power, to rethinking binary systems and hierarchies, and embracing pluralism. (Those statistics about gender and computer science graduates are limited, they note, by only using the “male” and “female” categories, thus excluding people who identify in different terms.)</p> <p>People interested in data feminism, the authors state, should also “value multiple forms of knowledge,” including firsthand knowledge that may lead us to question seemingly official data. Also, they should always consider the context in which data are generated, and “make labor visible” when it comes to data science. This last principle, the researchers note, speaks to the problem that even when women and other excluded people contribute to data projects, they often receive less credit for their work.</p> <p>For all the book’s critique of existing systems, programs, and practices, D’Ignazio and Klein are also careful to include examples of positive, successful efforts, such as the WomanStats project, which has grown and thrived over two decades.</p> <p>“For people who are data people but are new to feminism, we want to provide them with a very accessible introduction, and give them concepts and tools they can use in their practice,” D’Ignazio says. “We’re not imagining that people already have feminism in their toolkit. On the other hand, we are trying to speak to folks who are very tuned in to feminism or social justice principles, and highlight for them the ways data science is both problematic, but can be marshalled in the service of justice.”</p> Catherine D’Ignazio is the co-author of a new book, “Data Feminism,” published by MIT Press in March 2020. Image: Diana Levine and MIT PressData, Women, Faculty, Research, Books and authors, MIT Press, Diversity and inclusion, Ethics, Technology and society, Artificial intelligence, Machine learning, Computer science and technology, Urban studies and planning, School of Architecture and Planning “Doing machine learning the right way” Professor Aleksander Madry strives to build machine-learning models that are more reliable, understandable, and robust. Sat, 07 Mar 2020 23:59:59 -0500 Rob Matheson | MIT News Office <p>The work of MIT computer scientist Aleksander Madry is fueled by one core mission: “doing machine learning the right way.”</p> <p>Madry’s research centers largely on making machine learning — a type of artificial intelligence — more accurate, efficient, and robust against errors. In his classroom and beyond, he also worries about questions of ethical computing, as we approach an age where artificial intelligence will have great impact on many sectors of society.</p> <p>“I want society to truly embrace machine learning,” says Madry, a recently tenured professor in the Department of Electrical Engineering and Computer Science. “To do that, we need to figure out how to train models that people can use safely, reliably, and in a way that they understand.”</p> <p>Interestingly, his work with machine learning dates back only a couple of years, to shortly after he joined MIT in 2015. In that time, his research group has published several critical papers demonstrating that certain models can be easily tricked to produce inaccurate results — and showing how to make them more robust.</p> <p>In the end, he aims to make each model’s decisions more interpretable by humans, so researchers can peer inside to see where things went awry. At the same time, he wants to enable nonexperts to deploy the improved models in the real world for, say, helping diagnose disease or control driverless cars.</p> <p>“It’s not just about trying to crack open the machine-learning black box. I want to open it up, see how it works, and pack it back up, so people can use it without needing to understand what’s going on inside,” he says.</p> <p><strong>For the love of algorithms</strong></p> <p>Madry was born in Wroclaw, Poland, where he attended the University of Wroclaw as an undergraduate in the mid-2000s. While he harbored interest in computer science and physics, “I actually never thought I’d become a scientist,” he says.</p> <p>An avid video gamer, Madry initially enrolled in the computer science program with intentions of programming his own games. But in joining friends in a few classes in theoretical computer science and, in particular, theory of algorithms, he fell in love with the material. Algorithm theory aims to find efficient optimization procedures for solving computational problems, which requires tackling difficult mathematical questions. “I realized I enjoy thinking deeply about something and trying to figure it out,” says Madry, who wound up double-majoring in physics and computer science.</p> <p>When it came to delving deeper into algorithms in graduate school, he went to his first choice: MIT. Here, he worked under both Michel X. Goemans, who was a major figure in applied math and algorithm optimization, and Jonathan A. Kelner, who had just arrived to MIT as a junior faculty working in that field. For his PhD dissertation, Madry developed algorithms that solved a number of longstanding problems in graph algorithms, earning the 2011 George M. Sprowls Doctoral Dissertation Award for the best MIT doctoral thesis in computer science.</p> <p>After his PhD, Madry spent a year as a postdoc at Microsoft Research New England, before teaching for three years at the Swiss Federal Institute of Technology Lausanne — which Madry calls “the Swiss version of MIT.” But his alma mater kept calling him back: “MIT has the thrilling energy I was missing. It’s in my DNA.”</p> <p><strong>Getting adversarial</strong></p> <p>Shortly after joining MIT, Madry found himself swept up in a novel science: machine learning. In particular, he focused on understanding the re-emerging paradigm of deep learning. That’s an artificial-intelligence application that uses multiple computing layers to extract high-level features from raw input — such as using pixel-level data to classify images. MIT’s campus was, at the time, buzzing with new innovations in the domain.</p> <p>But that begged the question: Was machine learning all hype or solid science? “It seemed to work, but no one actually understood how and why,” Madry says.</p> <p>Answering that question set his group on a long journey, running experiment after experiment on deep-learning models to understand the underlying principles. A major milestone in this journey was an influential paper they published in 2018, developing a methodology for making machine-learning models more resistant to “adversarial examples.” Adversarial examples are slight perturbations to input data that are imperceptible to humans — such as changing the color of one pixel in an image — but cause a model to make inaccurate predictions. They illuminate a major shortcoming of existing machine-learning tools.</p> <p>Continuing this line of work, Madry’s group showed that the existence of these mysterious adversarial examples may contribute to how machine-learning models make decisions. In particular, models designed to differentiate images of, say, cats and dogs, make decisions based on features that do not align with how humans make classifications. Simply changing these features can make the model consistently misclassify cats as dogs, without changing anything in the image that’s really meaningful to humans.</p> <p>Results indicated some models — which may be used to, say, identify abnormalities in medical images or help autonomous cars identify objects in the road —&nbsp;aren’t exactly up to snuff. “People often think these models are superhuman, but they didn’t actually solve the classification problem we intend them to solve,” Madry says. “And their complete vulnerability to adversarial examples was a manifestation of that fact. That was an eye-opening finding.”</p> <p>That’s why Madry seeks to make machine-learning models more interpretable to humans. New models he’s developed show how much certain pixels in images the system is trained on can influence the system’s predictions. Researchers can then tweak the models to focus on pixels clusters more closely correlated with identifiable features — such as detecting an animal’s snout, ears, and tail. In the end, that will help make the models more humanlike —&nbsp;or “superhumanlike” —&nbsp;in their decisions. To further this work, Madry and his colleagues recently founded the <a href="">MIT Center for Deployable Machine Learning</a>, a collaborative research effort within the <a href="" target="_blank">MIT Quest for Intelligence</a> that is working toward building machine-learning tools ready for real-world deployment.&nbsp;</p> <p>“We want machine learning not just as a toy, but as something you can use in, say, an autonomous car, or health care. Right now, we don’t understand enough to have sufficient confidence in it for those critical applications,” Madry says.</p> <p><strong>Shaping education and policy</strong></p> <p>Madry views artificial intelligence and decision making (“AI+D” is one of the three <a href="">new academic units</a> in the Department of Electrical Engineering and Computer Science) as “the interface of computing that’s going to have the biggest impact on society.”</p> <p>In that regard, he makes sure to expose his students to the human aspect of computing. In part, that means considering consequences of what they’re building. Often, he says, students will be overly ambitious in creating new technologies, but they haven’t thought through potential ramifications on individuals and society. “Building something cool isn’t a good enough reason to build something,” Madry says. “It’s about thinking about not if we can build something, but if we should build something.”</p> <p>Madry has also been engaging in conversations about laws and policies to help regulate machine learning. A point of these discussions, he says, is to better understand the costs and benefits of unleashing machine-learning technologies on society.</p> <p>“Sometimes we overestimate the power of machine learning, thinking it will be our salvation. Sometimes we underestimate the cost it may have on society,” Madry says. “To do machine learning right, there’s still a lot still left to figure out.”</p> Alexander MadryImage: Ian MacLellanComputer science and technology, Algorithms, Artificial intelligence, Machine learning, Computer vision, Technology and society, Faculty, Profile, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering, MIT Schwarzman College of Computing, Quest for Intelligence Showing robots how to do your chores By observing humans, robots learn to perform complex tasks, such as setting a table. Thu, 05 Mar 2020 23:59:59 -0500 Rob Matheson | MIT News Office <p>Training interactive robots may one day be an easy job for everyone, even those without programming expertise. Roboticists are developing automated robots that can learn new tasks solely by observing humans. At home, you might someday show a domestic robot how to do routine chores. In the workplace, you could train robots like new employees, showing them how to perform many duties.</p> <p>Making progress on that vision, MIT researchers have designed a system that lets these types of robots learn complicated tasks that would otherwise stymie them with too many confusing rules. One such task is setting a dinner table under certain conditions. &nbsp;</p> <p>At its core, the researchers’ “Planning with Uncertain Specifications” (PUnS) system gives robots the humanlike planning ability to simultaneously weigh many ambiguous —&nbsp;and potentially contradictory —&nbsp;requirements to reach an end goal. In doing so, the system always chooses the most likely action to take, based on a “belief” about some probable specifications for the task it is supposed to perform.</p> <p>In their work, the researchers compiled a dataset with information about how eight objects — a mug, glass, spoon, fork, knife, dinner plate, small plate, and bowl — could be placed on a table in various configurations. A robotic arm first observed randomly selected human demonstrations of setting the table with the objects. Then, the researchers tasked the arm with automatically setting a table in a specific configuration, in real-world experiments and in simulation, based on what it had seen.</p> <p>To succeed, the robot had to weigh many possible placement orderings, even when items were purposely removed, stacked, or hidden. Normally, all of that would confuse robots too much. But the researchers’ robot made no mistakes over several real-world experiments, and only a handful of mistakes over tens of thousands of simulated test runs. &nbsp;</p> <p>“The vision is to put programming in the hands of domain experts, who can program robots through intuitive ways, rather than describing orders to an engineer to add to their code,” says first author Ankit Shah, a graduate student in the Department of Aeronautics and Astronautics (AeroAstro) and the Interactive Robotics Group, who emphasizes that their work is just one step in fulfilling that vision. “That way, robots won’t have to perform preprogrammed tasks anymore. Factory workers can teach a robot to do multiple complex assembly tasks. Domestic robots can learn how to stack cabinets, load the dishwasher, or set the table from people at home.”</p> <p>Joining Shah on the paper are AeroAstro and Interactive Robotics Group graduate student Shen Li and Interactive Robotics Group leader Julie Shah, an associate professor in AeroAstro and the Computer Science and Artificial Intelligence Laboratory.</p> <div class="cms-placeholder-content-video"></div> <p><strong>Bots hedging bets</strong></p> <p>Robots are fine planners in tasks with clear “specifications,” which help describe the task the robot needs to fulfill, considering its actions, environment, and end goal. Learning to set a table by observing demonstrations, is full of uncertain specifications. Items must be placed in certain spots, depending on the menu and where guests are seated, and in certain orders, depending on an item’s immediate availability or social conventions. Present approaches to planning are not capable of dealing with such uncertain specifications.</p> <p>A popular approach to planning is “reinforcement learning,” a trial-and-error machine-learning technique that rewards and penalizes them for actions as they work to complete a task. But for tasks with uncertain specifications, it’s difficult to define clear rewards and penalties. In short, robots never fully learn right from wrong.</p> <p>The researchers’ system, called PUnS (for Planning with Uncertain Specifications), enables a robot to hold a “belief” over a range of possible specifications. The belief itself can then be used to dish out rewards and penalties. “The robot is essentially hedging its bets in terms of what’s intended in a task, and takes actions that satisfy its belief, instead of us giving it a clear specification,” Ankit Shah says.</p> <p>The system is built on “linear temporal logic” (LTL), an expressive language that enables robotic reasoning about current and future outcomes. The researchers defined templates in LTL that model various time-based conditions, such as what must happen now, must eventually happen, and must happen until something else occurs. The robot’s observations of 30 human demonstrations for setting the table yielded a probability distribution over 25 different LTL formulas. Each formula encoded a slightly different preference — or specification — for setting the table. That probability distribution becomes its belief.</p> <p>“Each formula encodes something different, but when the robot considers various combinations of all the templates, and tries to satisfy everything together, it ends up doing the right thing eventually,” Ankit Shah says.</p> <p><strong>Following criteria</strong></p> <p>The researchers also developed several criteria that guide the robot toward satisfying the entire belief over those candidate formulas. One, for instance, satisfies the most likely formula, which discards everything else apart from the template with the highest probability. Others satisfy the largest number of unique formulas, without considering their overall probability, or they satisfy several formulas that represent highest total probability. Another simply minimizes error, so the system ignores formulas with high probability of failure.</p> <p>Designers can choose any one of the four criteria to preset before training and testing. Each has its own tradeoff between flexibility and risk aversion. The choice of criteria depends entirely on the task. In safety critical situations, for instance, a designer may choose to limit possibility of failure. But where consequences of failure are not as severe, designers can choose to give robots greater flexibility to try different approaches.</p> <p>With the criteria in place, the researchers developed an algorithm to convert the robot’s belief — the probability distribution pointing to the desired formula — into an equivalent reinforcement learning problem. This model will ping the robot with a reward or penalty for an action it takes, based on the specification it’s decided to follow.</p> <p>In simulations asking the robot to set the table in different configurations, it only made six mistakes out of 20,000 tries. In real-world demonstrations, it showed behavior similar to how a human would perform the task. If an item wasn’t initially visible, for instance, the robot would finish setting the rest of the table without the item. Then, when the fork was revealed, it would set the fork in the proper place. “That’s where flexibility is very important,” Ankit Shah says. “Otherwise it would get stuck when it expects to place a fork and not finish the rest of table setup.”</p> <p>Next, the researchers hope to modify the system to help robots change their behavior based on verbal instructions, corrections, or a user’s assessment of the robot’s performance. “Say a person demonstrates to a robot how to set a table at only one spot. The person may say, ‘do the same thing for all other spots,’ or, ‘place the knife before the fork here instead,’” Ankit Shah says. “We want to develop methods for the system to naturally adapt to handle those verbal commands, without needing additional demonstrations.”&nbsp;&nbsp;</p> Roboticists are developing automated robots that can learn new tasks solely by observing humans. At home, you might someday show a domestic robot how to do routine chores.Image: Christine Daniloff, MITResearch, Computer science and technology, Algorithms, Artificial intelligence, Machine learning, Robots, Robotics, Assistive technology, Aeronautical and astronautical engineering, Computer Science and Artificial Intelligence Laboratory (CSAIL), School of Engineering A new model of vision Computer model of face processing could reveal how the brain produces richly detailed visual representations so quickly. Wed, 04 Mar 2020 14:00:00 -0500 Anne Trafton | MIT News Office <p>When we open our eyes, we immediately see our surroundings in great detail. How the brain is able to form these richly detailed representations of the world so quickly is one of the biggest unsolved puzzles in the study of vision.</p> <p>Scientists who study the brain have tried to replicate this phenomenon using computer models of vision, but so far, leading models only perform much simpler tasks such as picking out an object or a face against a cluttered background. Now, a team led by MIT cognitive scientists has produced a computer model that captures the human visual system’s ability to quickly generate a detailed scene description from an image, and offers some insight into how the brain achieves this.</p> <p>“What we were trying to do in this work is to explain how perception can be so much richer than just attaching semantic labels on parts of an image, and to explore the question of how do we see all of the physical world,” says Josh Tenenbaum, a professor of computational cognitive science and a member of MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Center for Brains, Minds, and Machines (CBMM).</p> <p>The new model posits that when the brain receives visual input, it quickly performs a series of computations that reverse the steps that a computer graphics program would use to generate a 2D representation of a face or other object. This type of model, known as efficient inverse graphics (EIG), also correlates well with electrical recordings from face-selective regions in the brains of nonhuman primates, suggesting that the primate visual system may be organized in much the same way as the computer model, the researchers say.</p> <p>Ilker Yildirim, a former MIT postdoc who is now an assistant professor of psychology at Yale University, is the lead author of the paper, which appears today in <em>Science Advances</em>. Tenenbaum and Winrich Freiwald, a professor of neurosciences and behavior at Rockefeller University, are the senior authors of the study. Mario Belledonne, a graduate student at Yale, is also an author.</p> <p><strong>Inverse graphics</strong></p> <p>Decades of research on the brain’s visual system has studied, in great detail, how light input onto the retina is transformed into cohesive scenes. This understanding has helped artificial intelligence researchers develop computer models that can replicate aspects of this system, such as recognizing faces or other objects.</p> <p>“Vision is the functional aspect of the brain that we understand the best, in humans and other animals,” Tenenbaum says. “And computer vision is one of the most successful areas of AI at this point. We take for granted that machines can now look at pictures and recognize faces very well, and detect other kinds of objects.”</p> <p>However, even these sophisticated artificial intelligence systems don’t come close to what the human visual system can do, Yildirim says.</p> <p>“Our brains don’t just detect that there’s an object over there, or recognize and put a label on something,” he says. “We see all of the shapes, the geometry, the surfaces, the textures. We see a very rich world.”</p> <p>More than a century ago, the physician, physicist, and philosopher Hermann von Helmholtz theorized that the brain creates these rich representations by reversing the process of image formation. He hypothesized that the visual system includes an image generator that would be used, for example, to produce the faces that we see during dreams. Running this generator in reverse would allow the brain to work backward from the image and infer what kind of face or other object would produce that image, the researchers say.</p> <p>However, the question remained: How could the brain perform this process, known as inverse graphics, so quickly? Computer scientists have tried to create algorithms that could perform this feat, but the best previous systems require many cycles of iterative processing, taking much longer than the 100 to 200 milliseconds the brain requires to create a detailed visual representation of what you’re seeing. Neuroscientists believe perception in the brain can proceed so quickly because it is implemented in a mostly feedforward pass through several hierarchically organized layers of neural processing.</p> <p>The MIT-led team set out to build a special kind of deep neural network model to show how a neural hierarchy can quickly infer the underlying features of a scene — in this case, a specific face. In contrast to the standard deep neural networks used in computer vision, which are trained from labeled data indicating the class of an object in the image, the researchers’ network is trained from a model that reflects the brain’s internal representations of what scenes with faces can look like.</p> <p>Their model thus learns to reverse the steps performed by a computer graphics program for generating faces. These graphics programs begin with a three-dimensional representation of an individual face and then convert it into a two-dimensional image, as seen from a particular viewpoint. These images can be placed on an arbitrary background image. The researchers theorize that the brain’s visual system may do something similar when you dream or conjure a mental image of someone’s face.</p> <p>The researchers trained their deep neural network to perform these steps in reverse — that is, it begins with the 2D image and then adds features such as texture, curvature, and lighting, to create what the researchers call a “2.5D” representation. These 2.5D images specify the shape and color of the face from a particular viewpoint. Those are then converted into 3D representations, which don’t depend on the viewpoint.</p> <p>“The model gives a systems-level account of the processing of faces in the brain, allowing it to see an image and ultimately arrive at a 3D object, which includes representations of shape and texture, through this important intermediate stage of a 2.5D image,” Yildirim says.</p> <p><strong>Model performance</strong></p> <p>The researchers found that their model is consistent with data obtained by studying certain regions in the brains of macaque monkeys. In a study published in 2010, Freiwald and Doris Tsao of Caltech recorded the activity of neurons in those regions and analyzed how they responded to 25 different faces, seen from seven different viewpoints. That study revealed three stages of higher-level face processing, which the MIT team now hypothesizes correspond to three stages of their inverse graphics model: roughly, a 2.5D viewpoint-dependent stage; a stage that bridges from 2.5 to 3D; and a 3D, viewpoint-invariant stage of face representation.</p> <p>“What we show is that both the quantitative and qualitative response properties of those three levels of the brain seem to fit remarkably well with the top three levels of the network that we’ve built,” Tenenbaum says.</p> <p>The researchers also compared the model’s performance to that of humans in a task that involves recognizing faces from different viewpoints. This task becomes harder when researchers alter the faces by removing the face’s texture while preserving its shape, or distorting the shape while preserving relative texture. The new model’s performance was much more similar to that of humans than computer models used in state-of-the-art face-recognition software, additional evidence that this model may be closer to mimicking what happens in the human visual system.</p> <p>“This work is exciting because it introduces interpretable stages of intermediate representation into a feedforward neural network model of face recognition,” says Nikolaus Kriegeskorte, a professor of psychology and neuroscience at Columbia University, who was not involved in the research. “Their approach merges the classical idea that vision inverts a model of how the image was generated, with modern deep feedforward networks. It’s very interesting that this model better explains neural representations and behavioral responses.”</p> <p>The researchers now plan to continue testing the modeling approach on additional images, including objects that aren’t faces, to investigate whether inverse graphics might also explain how the brain perceives other kinds of scenes. In addition, they believe that adapting this approach to computer vision could lead to better-performing AI systems.</p> <p>“If we can show evidence that these models might correspond to how the brain works, this work could lead computer vision researchers to take more seriously and invest more engineering resources in this inverse graphics approach to perception,” Tenenbaum says. “The brain is still the gold standard for any kind of machine that sees the world richly and quickly.”</p> <p>The research was funded by the Center for Brains, Minds, and Machines at MIT, the National Science Foundation, the National Eye Institute, the Office of Naval Research, the New York Stem Cell Foundation, the Toyota Research Institute, and Mitsubishi Electric.</p> MIT cognitive scientists have developed a computer model of face recognition that performs a series of computations that reverse the steps that a computer graphics program would use to generate a 2D representation of a face.Image: courtesy of the researchersResearch, Computer vision, Brain and cognitive sciences, Center for Brains Minds and Machines, Computer Science and Artificial Intelligence Laboratory (CSAIL), School of Science, School of Engineering, National Science Foundation (NSF), Artificial intelligence, Machine learning, Neuroscience Integrating electronics onto physical prototypes In place of flat “breadboards,” 3D-printed CurveBoards enable easier testing of circuit design on electronics products. Tue, 03 Mar 2020 23:59:59 -0500 Rob Matheson | MIT News Office <p>MIT researchers have invented a way to integrate “breadboards” — flat platforms widely used for electronics prototyping — directly onto physical products. The aim is to provide a faster, easier way to test circuit functions and user interactions with products such as smart devices and flexible electronics.</p> <p>Breadboards are rectangular boards with arrays of pinholes drilled into the surface. Many of the holes have metal connections and contact points between them. Engineers can plug components of electronic systems — from basic circuits to full computer processors — into the pinholes where they want them to connect. Then, they can rapidly test, rearrange, and retest the components as needed.</p> <p>But breadboards have remained that same shape for decades. For that reason, it’s difficult to test how the electronics will look and feel on, say, wearables and various smart devices. Generally, people will first test circuits on traditional breadboards, then slap them onto a product prototype. If the circuit needs to be modified, it’s back to the breadboard for testing, and so on.</p> <p>In a paper being presented at CHI (Conference on Human Factors in Computing Systems), the researchers describe “CurveBoards,” 3D-printed objects with the structure and function of a breadboard integrated onto their surfaces. Custom software automatically designs the objects, complete with distributed pinholes that can be filled with conductive silicone to test electronics. The end products are accurate representations of the real thing, but with breadboard surfaces.</p> <p>CurveBoards “preserve an object’s look and feel,” the researchers write in their paper, while enabling designers to try out component configurations and test interactive scenarios during prototyping iterations. In their work, the researchers printed CurveBoards for smart bracelets and watches, Frisbees, helmets, headphones, a teapot, and a flexible, wearable e-reader.</p> <p>“On breadboards, you prototype the function of a circuit. But you don’t have context of its form — how the electronics will be used in a real-world prototype environment,” says first author Junyi Zhu, a graduate student in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “Our idea is to fill this gap, and merge form and function testing in very early stage of prototyping an object. … &nbsp;CurveBoards essentially add an additional axis to the existing [three-dimensional] XYZ axes of the object — the ‘function’ axis.”</p> <p>Joining Zhu on the paper are CSAIL graduate students Lotta-Gili Blumberg, Martin Nisser, and Ethan Levi Carlson; Department of Electrical Engineering and Computer Science (EECS) undergraduate students Jessica Ayeley Quaye and Xin Wen; former EECS undergraduate students Yunyi Zhu and Kevin Shum; and Stefanie Mueller, the X-Window Consortium Career Development Assistant Professor in EECS.</p> <div class="cms-placeholder-content-video"></div> <p><strong>Custom software and hardware</strong></p> <p>A core component of the CurveBoard is custom design-editing software. Users import a 3D model of an object. Then, they select the command “generate pinholes,” and the software automatically maps all pinholes uniformly across the object. Users then choose automatic or manual layouts for connectivity channels. The automatic option lets users explore a different layout of connections across all pinholes with the click of a button. For manual layouts, interactive tools can be used to select groups of pinholes and indicate the type of connection between them. The final design is exported to a file for 3D printing.</p> <p>When a 3D object is uploaded, the software essentially forces its shape into a “quadmesh” — where the object is represented as a bunch of small squares, each with individual parameters. In doing so, it creates a fixed spacing between the squares. Pinholes — which are cones, with the wide end on the surface and tapering down —&nbsp;will be placed at each point where the corners of the squares touch. For channel layouts, some geometric techniques ensure the chosen channels will connect the desired electrical components without crossing over one another.</p> <p>In their work, the researchers 3D printed objects using a flexible, durable, nonconductive silicone. To provide connectivity channels, they created a custom conductive silicone that can be syringed into the pinholes and then flows through the channels after printing. The silicone is a mixture of a silicone materials designed to have minimal electricity resistance, allowing various types electronics to function.</p> <p>To validate the CurveBoards, the researchers printed a variety of smart products. Headphones, for instance, came equipped with menu controls for speakers and music-streaming capabilities. An interactive bracelet included a digital display, LED, and photoresistor for heart-rate monitoring, and a step-counting sensor. A teapot included a small camera to track the tea’s color, as well as colored lights on the handle to indicate hot and cold areas. They also printed a wearable e-book reader with a flexible display.</p> <p><strong>Better, faster prototyping</strong></p> <p>In a user study, the team investigated the benefits of CurveBoards prototyping. They split six participants with varying prototyping experience into two sections: One used traditional breadboards and a 3D-printed object, and the other used only a CurveBoard of the object. Both sections designed the same prototype but switched back and forth between sections after completing designated tasks. In the end, five of six of the participants preferred prototyping with the CurveBoard. Feedback indicated the CurveBoards were overall faster and easier to work with.</p> <p>But CurveBoards are not designed to replace breadboards, the researchers say. Instead, they’d work particularly well as a so-called “midfidelity” step in the prototyping timeline, meaning between initial breadboard testing and the final product. “People love breadboards, and there are cases where they’re fine to use,” Zhu says. “This is for when you have an idea of the final object and want to see, say, how people interact with the product. It’s easier to have a CurveBoard instead of circuits stacked on top of a physical object.”</p> <p>Next, the researchers hope to design general templates of common objects, such as hats and bracelets. Right now, a new CurveBoard must built for each new object. Ready-made templates, however, would let designers quickly experiment with basic circuits and user interaction, before designing their specific CurveBoard.</p> <p>Additionally, the researchers want to move some early-stage prototyping steps entirely to the software side. The idea is that people can design and test circuits — and possibly user interaction — entirely on the 3D model generated by the software. After many iterations, they can 3D print a more finalized CurveBoard. “That way you’ll know exactly how it’ll work in the real world, enabling fast prototyping,” Zhu says. “That would be a more ‘high-fidelity’ step for prototyping.”</p> CurveBoards are 3D breadboards — which are commonly used to prototype circuits — that can be designed by custom software, 3D printed, and directly integrated into the surface of physical objects, such as smart watches, bracelets, helmets, headphones, and even flexible electronics. CurveBoards can give designers an additional prototyping technique to better evaluate how circuits will look and feel on physical products that users interact with.Image: Dishita Turakhia and Junyi ZhuResearch, Computer science and technology, 3-D printing, Design, Manufacturing, electronics, Computer graphics, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering Demystifying the world of deep networks Researchers discover that no magic is required to explain why deep networks generalize despite going against statistical intuition. Fri, 28 Feb 2020 14:40:01 -0500 Kris Brewer | Center for Brains, Minds and Machines <p>Introductory statistics courses teach us that, when fitting a model to some data, we should have more data than free parameters to avoid the danger of overfitting — fitting noisy data too closely, and thereby failing to fit new data. It is surprising, then, that in modern deep learning the practice is to have orders of magnitude more parameters than data. Despite this, deep networks show good predictive performance, and in fact do better the more parameters they have. Why would that be?</p> <p>It has been known for some time that good performance in machine learning comes from controlling the complexity of networks, which is not just a simple function of the number of free parameters. The complexity of a classifier, such as a neural network, depends on measuring the “size” of the space of functions that this network represents, with multiple technical measures previously suggested: Vapnik–Chervonenkis dimension, covering numbers, or Rademacher complexity, to name a few. Complexity, as measured by these notions, can be controlled during the learning process by imposing a constraint on the norm of the parameters — in short, on how “big” they can get. The surprising fact is that no such explicit constraint seems to be needed in training deep networks. Does deep learning lie outside of the classical learning theory? Do we need to rethink the foundations?</p> <p>In a new <em>Nature Communications</em> paper, “Complexity Control by Gradient Descent in Deep Networks,” a team from the Center for Brains, Minds, and Machines led by Director Tomaso Poggio, the Eugene McDermott Professor in the MIT Department of Brain and Cognitive Sciences, has shed some light on this puzzle by addressing the most practical and successful applications of modern deep learning: classification problems.</p> <p>“For classification problems, we observe that in fact the parameters of the model do not seem to converge, but rather grow in size indefinitely during gradient descent. However, in classification problems only the normalized parameters matter — i.e., the direction they define, not their size,” says co-author and MIT PhD candidate Qianli Liao. “The not-so-obvious thing we showed is that the commonly used gradient descent on the unnormalized parameters induces the desired complexity control on the normalized ones.”</p> <p>“We have known for some time in the case of regression for shallow linear networks, such as kernel machines, that iterations of gradient descent provide an implicit, vanishing regularization effect,” Poggio says. “In fact, in this simple case we probably know that we get the best-behaving maximum-margin, minimum-norm solution. The question we asked ourselves, then, was: Can something similar happen for deep networks?”</p> <p>The researchers found that it does. As co-author and MIT postdoc Andrzej Banburski explains, “Understanding convergence in deep networks shows that there are clear directions for improving our algorithms. In fact, we have already seen hints that controlling the rate at which these unnormalized parameters diverge allows us to find better performing solutions and find them faster.”</p> <p>What does this mean for machine learning? There is no magic behind deep networks. The same theory behind all linear models is at play here as well. This work suggests ways to improve deep networks, making them more accurate and faster to train.</p> MIT researchers (left to right) Qianli Liao, Tomaso Poggio, and Andrzej Banburski stand with their equations. Image: Kris BrewerCenter for Brains Minds and Machines, Brain and cognitive sciences, Electrical engineering and computer science (EECS), Machine learning, Artificial intelligence, Research Machine learning picks out hidden vibrations from earthquake data Technique may help scientists more accurately map vast underground geologic structures. Fri, 28 Feb 2020 13:00:46 -0500 Jennifer Chu | MIT News Office <p>Over the last century, scientists have developed methods to map the structures within the Earth’s crust, in order to identify resources such as oil reserves, geothermal sources, and, more recently, reservoirs where excess carbon dioxide could potentially be sequestered. They do so by tracking seismic waves that are produced naturally by earthquakes or artificially via explosives or underwater air guns. The way these waves bounce and scatter through the Earth can give scientists an idea of the type of structures that lie beneath the surface.</p> <p>There is a narrow range of seismic waves — those that occur at low frequencies of around 1 hertz — that could give scientists the clearest picture of underground structures spanning wide distances. But these waves are often drowned out by Earth’s noisy seismic hum, and are therefore difficult to pick up with current detectors. Specifically generating low-frequency waves would require pumping in enormous amounts of energy. For these reasons, low-frequency seismic waves have largely gone missing in human-generated seismic data.</p> <p>Now MIT researchers have come up with a machine learning workaround to fill in this gap.</p> <p>In a paper appearing in the journal <em>Geophysics</em>, they describe a method in which they trained a neural network on hundreds of different simulated earthquakes. When the researchers presented the trained network with only the high-frequency seismic waves produced from a new simulated earthquake, the neural network was able to imitate the physics of wave propagation and accurately estimate the quake’s missing low-frequency waves.</p> <p>The new method could allow researchers to artificially synthesize the low-frequency waves that are hidden in seismic data, which can then be used to more accurately map the Earth’s internal structures.</p> <p>“The ultimate dream is to be able to map the whole subsurface, and be able to say, for instance, ‘this is exactly what it looks like underneath Iceland, so now you know where to explore for geothermal sources,’” says co-author Laurent Demanet, professor of applied mathematics at MIT. “Now we’ve shown that deep learning offers a solution to be able to fill in these missing frequencies.”</p> <p>Demanet’s co-author is lead author Hongyu Sun, a graduate student in MIT’s Department of Earth, Atmospheric and Planetary Sciences.</p> <p><strong>Speaking another frequency</strong></p> <p>A neural network is a set of algorithms modeled loosely after the neural workings of the human brain. The algorithms are designed to recognize patterns in data that are fed into the network, and to cluster these data into categories, or labels. A common example of a neural network involves visual processing; the model is trained to classify an image as either a cat or a dog, based on the patterns it recognizes between thousands of images that are specifically labeled as cats, dogs, and other objects.</p> <p>Sun and Demanet adapted a neural network for signal processing, specifically, to recognize patterns in seismic data. They reasoned that if a neural network was fed enough examples of earthquakes, and the ways in which the resulting high- and low-frequency seismic waves travel through a particular composition of the Earth, the network should be able to, as they write in their paper, “mine the hidden correlations among different frequency components” and extrapolate any missing frequencies if the network were only given an earthquake’s partial seismic profile.</p> <p>The researchers looked to train a convolutional neural network, or CNN, a class of deep neural networks that is often used to analyze visual information. A CNN very generally consists of an input and output layer, and multiple hidden layers between, that process inputs to identify correlations between them.</p> <p>Among their many applications, CNNs have been used as a means of generating visual or auditory “deepfakes” — content that has been extrapolated or manipulated through deep-learning and neural networks, to make it seem, for example, as if a woman were talking with a man’s voice.</p> <p>“If a network has seen enough examples of how to take a male voice and transform it into a female voice or vice versa, you can create a sophisticated box to do that,” Demanet says. “Whereas here we make the Earth speak another frequency — one that didn’t originally go through it.”</p> <p><strong>Tracking waves</strong></p> <p>The researchers trained their neural network with inputs that they generated using the Marmousi model, a complex two-dimensional geophysical model that simulates the way seismic waves travel through geological structures of varying density and composition. &nbsp;</p> <p>In their study, the team used the model to simulate nine “virtual Earths,” each with a different subsurface composition. For each Earth model, they simulated 30 different earthquakes, all with the same strength, but different starting locations. In total, the researchers generated hundreds of different seismic scenarios. They fed the information from almost all of these simulations into their neural network and let the network find correlations between seismic signals.</p> <p>After the training session, the team introduced to the neural network a new earthquake that they simulated in the Earth model but did not include in the original training data. They only included the high-frequency part of the earthquake’s seismic activity, in hopes that the neural network learned enough from the training data to be able to infer the missing low-frequency signals from the new input.</p> <p>They found that the neural network produced the same low-frequency values that the Marmousi model originally simulated.</p> <p>“The results are fairly good,” Demanet says. “It’s impressive to see how far the network can extrapolate to the missing frequencies.”</p> <p>As with all neural networks, the method has its limitations. Specifically, the neural network is only as good as the data that are fed into it. If a new input is wildly different from the bulk of a network’s training data, there’s no guarantee that the output will be accurate. To contend with this limitation, the researchers say they plan to introduce a wider variety of data to the neural network, such as earthquakes of different strengths, as well as subsurfaces of more varied composition.</p> <p>As they improve the neural network’s predictions, the team hopes to be able to use the method to extrapolate low-frequency signals from actual seismic data, which can then be plugged into seismic models to more accurately map the geological structures below the Earth’s surface. The low frequencies, in particular, are a key ingredient for solving the big puzzle of finding the correct physical model.</p> <p>“Using this neural network will help us find the missing frequencies to ultimately improve the subsurface image and find the composition of the Earth,” Demanet says.</p> <p>This research was supported, in part, by Total SA and the U.S. Air Force Office of Scientific Research.</p> MIT researchers have used a neural network to identify low-frequency seismic waves hidden in earthquake data. The technique may help scientists more accurately map the Earth’s interior.Image: Christine Daniloff, MITEAPS, Earthquakes, Environment, Geology, Mathematics, Research, School of Science, Machine learning, Artificial intelligence, Earth and atmospheric sciences To self-drive in the snow, look under the road Weather’s a problem for autonomous cars. MIT’s new system shows promise by using “ground-penetrating radar” instead of cameras or lasers. Wed, 26 Feb 2020 14:50:00 -0500 Adam Conner-Simons | MIT CSAIL <p>Car companies have been feverishly working to improve the technologies behind self-driving cars. But so far even the most high-tech vehicles still fail when it comes to safely navigating in rain and snow.&nbsp;</p> <p>This is because these weather conditions wreak havoc on the most common approaches for sensing, which usually involve either lidar sensors or cameras. In the snow, for example, cameras can no longer recognize lane markings and traffic signs, while the lasers of lidar sensors malfunction when there’s, say, stuff flying down from the sky.</p> <p>MIT researchers have recently been wondering whether an entirely different approach might work. Specifically, what if we instead looked under the road?&nbsp;</p> <div class="cms-placeholder-content-video"></div> <p>A team from MIT’s <a href="">Computer Science and Artificial Intelligence Laboratory</a> (CSAIL) has developed a new system that uses an existing technology called ground-penetrating radar (GPR) to send electromagnetic pulses underground that measure the area’s specific combination of soil, rocks, and roots. Specifically, the CSAIL team used a particular form of GPR instrumentation developed at <a href="" target="_blank">MIT Lincoln Laboratory</a> called <a href="" target="_self">localizing ground-penetrating radar</a>, or LGPR. The mapping process creates a unique fingerprint of sorts that the car can later use to localize itself when it returns to that particular plot of land.</p> <p>“If you or I grabbed a shovel and dug it into the ground, all we’re going to see is a bunch of dirt,” says CSAIL PhD student Teddy Ort, lead author on a new paper about the project that will be published in the <em>IEEE Robotics and Automation Letters</em> journal later this month. “But LGPR can quantify the specific elements there and compare that to the map it’s already created, so that it knows exactly where it is, without needing cameras or lasers.”</p> <p>In tests, the team found that in snowy conditions the navigation system’s average margin of error was on the order of only about an inch compared to clear weather. The researchers were surprised to find that it had a bit more trouble with rainy conditions, but was still only off by an average of 5.5 inches. (This is because rain leads to more water soaking into the ground, leading to a larger disparity between the original mapped LGPR reading and the current condition of the soil.)</p> <p>The researchers said the system’s robustness was further validated by the fact that, over a period of six months of testing, they never had to unexpectedly step in to take the wheel.&nbsp;</p> <p>“Our work demonstrates that this approach is actually a practical way to help self-driving cars navigate poor weather without actually having to be able to ‘see’ in the traditional sense using laser scanners or cameras,” says MIT Professor Daniela Rus, director of CSAIL and senior author on the new paper, which will also be presented in May at the International Conference on Robotics and Automation in Paris.</p> <p>While the team has only tested the system at low speeds on a closed country road, Ort said that existing work from Lincoln Laboratory suggests that the system could easily be extended to highways and other high-speed areas.&nbsp;</p> <p>This is the first time that developers of self-driving systems have employed ground-penetrating radar, which has previously been used in fields like construction planning, landmine detection, and even <a href="" target="_blank">lunar exploration</a>. The approach wouldn’t be able to work completely on its own, since it can’t detect things above ground. But its ability to localize in bad weather means that it would couple nicely with lidar and vision approaches.</p> <p>“Before releasing autonomous vehicles on public streets, localization and navigation have to be totally reliable at all times,” says Roland Siegwart, a professor of autonomous systems at ETH Zurich who was not involved in the project. “The CSAIL team’s innovative and novel concept has the potential to push autonomous vehicles much closer to real-world deployment.”&nbsp;</p> <p>One major benefit of mapping out an area with LGPR is that underground maps tend to hold up better over time than maps created using vision or lidar, since features of an above-ground map are much more likely to change. LGPR maps also take up only about 80 percent of the space used by traditional 2D sensor maps that many companies use for their cars.&nbsp;</p> <p>While the system represents an important advance, Ort notes that it’s far from road-ready. Future work will need to focus on designing mapping techniques that allow LGPR datasets to be stitched together to be able to deal with multi-lane roads and intersections. In addition, the current hardware is bulky and 6 feet wide, so major design advances need to be made before it’s small and light enough to fit into commercial vehicles.</p> <p>Ort and Rus co-wrote the paper with CSAIL postdoc Igor Gilitschenski. The project was supported, in part, by MIT Lincoln Laboratory.</p> MIT's new system allows a self-driving car to situate itself in snowy conditions.Photo courtesy of the researchersComputer Science and Artificial Intelligence Laboratory (CSAIL), Lincoln Laboratory, Robotics, Electrical engineering and computer science (EECS), School of Engineering, Research, Distributed Robotics Laboratory, Artificial intelligence, Automobiles, Autonomous vehicles, Faculty, Computer vision, Machine learning, Transportation, MIT Schwarzman College of Computing Protecting sensitive metadata so it can’t be used for surveillance System ensures hackers eavesdropping on large networks can’t find out who’s communicating and when they’re doing so. Wed, 26 Feb 2020 00:00:00 -0500 Rob Matheson | MIT News Office <p>MIT researchers have designed a scalable system that secures the metadata — such as who’s corresponding and when — of millions of users in communications networks, to help protect the information against possible state-level surveillance.</p> <p>Data encryption schemes that protect the content of online communications are prevalent today. Apps like WhatsApp, for instance, use “end-to-end encryption” (E2EE), a scheme that ensures third-party eavesdroppers can’t read messages sent by end users.</p> <p>But most of those schemes overlook metadata, which contains information about who’s talking, when the messages are sent, the size of message, and other information. Many times, that’s all a government or other hacker needs to know to track an individual. This can be especially dangerous for, say, a government whistleblower or people living in oppressive regimes talking with journalists.</p> <p>Systems that fully protect user metadata with cryptographic privacy are complex, and they suffer scalability and speed issues that have so far limited their practicality. Some methods can operate quickly but provide much weaker security. In a paper being presented at the USENIX Symposium on Networked Systems Design and Implementation, the MIT researchers describe “XRD” (for Crossroads), a metadata-protection scheme that can handle cryptographic communications from millions of users in minutes, whereas traditional methods with the same level of security would take hours to send everyone’s messages.</p> <p>“There is a huge lack in protection for metadata, which is sometimes very sensitive. The fact that I’m sending someone a message at all is not protected by encryption,” says first author Albert Kwon PhD ’19, a recent graduate from the Computer Science and Artificial Intelligence Laboratory (CSAIL). “Encryption can protect content well. But how can we fully protect users from metadata leaks that a state-level adversary can leverage?”</p> <p>Joining Kwon on the paper are David Lu, an undergraduate in the Department of Electrical Engineering and Computer Science; and Srinivas Devadas, the Edwin Sibley Webster Professor of Electrical Engineering and Computer Science in CSAIL.</p> <p><strong>New spin on mix nets</strong></p> <p>Starting in 2013, disclosures of classified information by Edward Snowden revealed widespread global surveillance by the U.S. government. Although the mass collection of metadata by the National Security Agency was subsequently discontinued, in 2014 former director of the NSA and the Central Intelligence Agency Michael Hayden explained that the government can often rely solely on metadata to find the information it’s seeking. As it happens, this is right around the time Kwon started his PhD studies.</p> <p>“That was like a punch to the cryptography and security communities,” Kwon says. “That meant encryption wasn’t really doing anything to stop spying in that regard.”</p> <p>Kwon spent most of his PhD program focusing on metadata privacy. With XRD, Kwon says he “put a new spin” on a traditional E2EE metadata-protecting scheme, called “mix nets,” which was invented decades ago but suffers from scalability issues.</p> <p>Mix nets use chains of servers, known as mixes, and public-private key encryption. The first server receives encrypted messages from many users and decrypts a single layer of encryption from each message. Then, it shuffles the messages in random order and transmits them to the next server, which does the same thing, and so on down the chain. The last server decrypts the final encryption layer and sends the message to the target receiver.</p> <p>Servers only know the identities of the immediate source (the previous server) and immediate destination (the next server). Basically, the shuffling and limited identity information breaks the link between source and destination users, making it very difficult for eavesdroppers to get that information. As long as one server in the chain is “honest”— meaning it follows protocol — metadata is almost always safe.</p> <p>However, “active attacks” can occur, in which a malicious server in a mix net tampers with the messages to reveal user sources and destinations. In short, the malicious server can drop messages or modify sending times to create communications patterns that reveal direct links between users.</p> <p>Some methods add cryptographic proofs between servers to ensure there’s been no tampering. These rely on public key cryptography, which is secure, but it’s also slow and limits scaling. For XRD, the researchers invented a far more efficient version of the cryptographic proofs, called “aggregate hybrid shuffle,” that guarantees servers are receiving and shuffling message correctly, to detect any malicious server activity.</p> <p>Each server has a secret private key and two shared public keys. Each server must know all the keys to decrypt and shuffle messages. Users encrypt messages in layers, using each server’s secret private key in its respective layer. When a server receives messages, it decrypts and shuffles them using one of the public keys combined with its own private key. Then, it uses the second public key to generate a proof confirming that it had, indeed, shuffled every message without dropping or manipulating any. All other servers in the chain use their secret private keys and the other servers’ public keys in a way that verifies this proof. If, at any point in the chain, a server doesn’t produce the proof or provides an incorrect proof, it’s immediately identified as malicious.</p> <p>This relies on a clever combination of the popular public key scheme with one called “authenticated encryption,” which uses only private keys but is very quick at generating and verifying the proofs. In this way, XRD achieves tight security from public key encryption while running quickly and efficiently.&nbsp;&nbsp;&nbsp;</p> <p>To further boost efficiency, they split the servers into multiple chains and divide their use among users. (This is another traditional technique they improved upon.) Using some statistical techniques, they estimate how many servers in each chain could be malicious, based on IP addresses and other information. From that, they calculate how many servers need to be in each chain to guarantee there’s at least one honest server.&nbsp; Then, they divide the users into groups that send duplicate messages to multiple, random chains, which further protects their privacy while speeding things up.</p> <p><strong>Getting to real-time</strong></p> <p>In computer simulations of activity from 2 million users sending messages on a network of 100 servers, XRD was able to get everyone’s messages through in about four minutes. Traditional systems using the same server and user numbers, and providing the same cryptographic security, took one to two hours.</p> <p>“This seems slow in terms of absolute speed in today’s communication world,” Kwon says. “But it’s important to keep in mind that the fastest systems right now [for metadata protection] take hours, whereas ours takes minutes.”</p> <p>Next, the researchers hope to make the network more robust to few users and in instances where servers go offline in the midst of operations, and to speed things up. “Four minutes is acceptable for sensitive messages and emails where two parties’ lives are in danger, but it’s not as natural as today’s internet,” Kwon says. “We want to get to the point where we’re sending metadata-protected messages in near real-time.”</p> In a new metadata-protecting scheme, users send encrypted messages to multiple chains of servers, with each chain mathematically guaranteed to have at least one hacker-free server. Each server decrypts and shuffles the messages in random order, before shooting them to the next server in line. Image: courtesy of the researchersResearch, Computer science and technology, Algorithms, Cyber security, Data, Technology and society, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering MIT Solve announces 2020 global challenges Tech-based solutions sought for challenges in work environments, education for girls and women, maternal and newborn health, and sustainable food. Tue, 25 Feb 2020 16:15:01 -0500 Claire Crowther | MIT Solve <p>On Feb. 25, MIT Solve launched its <a href="">2020 Global Challenges</a>: Good Jobs and Inclusive Entrepreneurship, Learning for Girls and Women, Maternal and Newborn Health, and Sustainable Food Systems, with&nbsp;over $1 million in prize funding&nbsp;available across the challenges.</p> <p>Solve seeks tech-based solutions from social entrepreneurs around the world that address these four challenges. Anyone, anywhere can apply by the June 18 deadline. This year, to guide applicants, Solve created a course with <em>MITx</em> entitled “<a href="">Business and Impact Planning for Social Enterprises</a>,” which introduces core business-model and theory-of-change concepts to early-stage entrepreneurs.</p> <p>Finalists will be invited to attend Solve Challenge Finals on Sept. 20 in New York City during U.N. General Assembly week. At the event, they will pitch their solutions to Solve’s Challenge Leadership Groups, judging panels comprised of industry leaders and MIT faculty. The judges will select the most promising solutions as Solver teams.</p> <p>“Based all over the world, our Solver teams are incredibly diverse and have innovative solutions that turn air pollution into ink, recycle and resell used textiles, crowdsource data on wheelchair accessibility in public spaces, and much more,” says Solve Executive Director Alex Amouyel. “World-changing ideas can come from anywhere, and if you have a relevant solution, we want to hear it.”</p> <p>Solver teams participate in a nine-month program that connects them to the resources they need to scale. To date, Solve has facilitated more than 175 partnerships providing resources such as mentorship, technical expertise, and impact planning. In the past three years, Solve has brokered over $14 million in funding commitments to Solver teams and entrepreneurs.</p> <p>Solve’s challenge design process collects insights and ideas from industry leaders, MIT faculty, and local community voices alike. To develop the 2020 Global Challenges, Solve consulted more than 500 subject matter experts and hosted 14 Challenge Design Workshops in eight countries — in places ranging from Silicon Valley to London to Lagos to Ho Chi Minh City. Solve’s open innovation platform garnered more than 26,000 online votes on challenge themes.</p> <ol> <li> <p>Good Jobs and Inclusive Entrepreneurship:<strong> </strong>How can marginalized populations access and create good jobs and entrepreneurial opportunities for themselves?</p> </li> <li> <p>Learning for Girls and Women:<strong> </strong>How can marginalized girls and young women access quality learning opportunities to succeed?</p> </li> <li> <p>Maternal and Newborn Health:<strong> </strong>How can every pregnant woman, new mother, and newborn access the care they need to survive and thrive?</p> </li> <li> <p>Sustainable Food Systems:<strong> </strong>How can we produce and consume low-carbon, resilient, and nutritious food?</p> </li> </ol> <p>As a marketplace for social impact innovation, Solve’s mission is to solve world challenges. Solve finds promising tech-based social entrepreneurs around the world, then brings together MIT’s innovation ecosystem and a community of members to fund and support these entrepreneurs to help scale their impact. Organizations interested in joining the Solve community can learn more and <a href="">apply for membership here</a>.</p> <div></div> Renewed products consist of upcycled or recycling materials. The Renewal Workshop is an MIT Solver team that works to save textiles from landfill.Photo: The Renewal Workshop MIT Solve, Special events and guest speakers, Global, Technology and society, Innovation and Entrepreneurship (I&E), International development, Artificial intelligence, Learning, Environment, Health, Community, Startups, Crowdsourcing Bringing deep learning to life MIT duo uses music, videos, and real-world examples to teach students the foundations of artificial intelligence. Mon, 24 Feb 2020 14:10:01 -0500 Kim Martineau | MIT Quest for Intelligence <p>Gaby Ecanow loves listening to music, but never considered writing her own until taking <a href="" target="_blank">6.S191</a> (Introduction to Deep Learning). By her second class, the second-year MIT student had composed an original Irish folk song with the help of a recurrent neural network, and was considering how to adapt the model to create her own Louis the Child-inspired dance beats.</p> <p>“It was cool,” she says. “It didn’t sound at all like a machine had made it.”&nbsp;</p> <p>This year, 6.S191 kicked off as usual, with students spilling into the aisles of Stata Center’s Kirsch Auditorium during Independent Activities Period (IAP). But the opening lecture featured a twist: a&nbsp;<a href=";" target="_blank">recorded welcome</a>&nbsp;from former President Barack Obama. The video was quickly revealed to be an AI-generated fabrication, one of many twists that&nbsp;<a href="" target="_blank">Alexander Amini</a>&nbsp;’17 and&nbsp;<a href="" target="_blank">Ava Soleimany</a>&nbsp;’16 introduce throughout their for-credit course to make the equations and code come alive.&nbsp;</p> <div class="cms-placeholder-content-video"></div> <p>As&nbsp;hundreds of their peers look on, Amini and Soleimany take turns at the podium. If they appear at ease, it’s because they know the material cold; they designed the curriculum themselves, and have taught it for the past three years.&nbsp;The course covers the technical foundations of deep learning and its societal implications through lectures and software labs focused on real-world applications. On the final day, students compete for prizes by pitching their own ideas for research projects. In the weeks leading up to class, Amini and Soleimany spend hours updating the labs, refreshing their lectures, and honing their presentations.</p> <p>A branch of&nbsp;machine learning,&nbsp;deep learning harnesses massive data and algorithms&nbsp;modeled loosely on how the brain processes information to make predictions. The class has been credited with helping to spread machine-learning tools into research labs across MIT. That’s by design, says Amini, a graduate student in MIT’s&nbsp;<a href="">Department of Electrical Engineering and Computer Science</a>&nbsp;(EECS), and Soleimany, a graduate student at MIT and Harvard University.</p> <p>Both are using machine learning in their own research — Amini in engineering robots, and Soleimany in developing diagnostic tools for cancer — and they wanted to make sure the curriculum would prepare students to do the same. In addition to the lab on developing a music-generating AI,&nbsp;they offer labs on building a face-recognition model with convolutional neural networks and a bot that uses reinforcement learning to play the vintage Atari video game, Pong.&nbsp;After students master the basics, those taking the class for credit go on to create applications of their own.&nbsp;</p> <p>This year, 23 teams&nbsp;presented projects. Among the prize winners was Carmen&nbsp;Martin, a graduate student in the&nbsp;<a href="" target="_blank">Harvard-MIT Program in Health Sciences and Technology</a>&nbsp;(HST), who&nbsp;proposed using a type of neural net called a graph convolutional network to predict the spread of coronavirus. She combined several data streams: airline ticketing data to measure population fluxes, real-time confirmation of new infections, and a ranking of how well countries are equipped to prevent and respond to&nbsp;a pandemic.&nbsp;</p> <p>“The goal is to train the model to predict cases to guide national governments and the World Health Organization in their recommendations to limit new cases and save lives,” she says.</p> <p>A second prize winner, EECS graduate student Samuel Sledzieski, proposed building a model to predict protein interactions using only their amino acid sequences. Predicting protein behavior is key to designing drug targets, among other clinical applications, and Sledzieski wondered if deep learning could speed up the search for viable protein pairs.&nbsp;</p> <p>“There’s still work to be done, but I’m excited by how far I was able to get in three days,” he says. “Having easy-to-follow examples in TensorFlow and Keras helped me understand how to actually build and train these models myself.”&nbsp;He plans to continue the work in his current lab rotation with&nbsp;<a href="">Bonnie Berger</a>, the Simons Professor of Mathematics in EECS and the&nbsp;<a href="" target="_blank">Computer Science and Artificial Intelligence Laboratory</a>&nbsp;(CSAIL).&nbsp;</p> <p>Each year, students also hear about emerging deep-learning applications from companies sponsoring the course. David Cox, co-director of the&nbsp;<a href="">MIT-IBM Watson AI Lab</a>, covered neuro-symbolic AI, a hybrid approach that combines symbolic programs with deep learning’s expert pattern-matching ability. Alex Wiltschko, a senior researcher at Google Brain, spoke about using a network analysis tool to predict the scent of small molecules. Chuan Li, chief scientific officer at Lambda Labs, discussed neural rendering, a tool for reconstructing and generating graphics scenes. Animesh Garg, a senior researcher at NVIDIA, covered strategies for developing robots that perceive and act more human-like.</p> <p>With 350 students taking the&nbsp;live course each year, and&nbsp;more than a million people who have watched the lectures online, Amini and Soleimany have become prominent ambassadors for deep learning. Yet, it was tennis that first brought them together.&nbsp;</p> <p>Amini competed nationally as a high school student in Ireland&nbsp;and built an award-winning&nbsp;AI model to help amateur and pro tennis players improve their strokes; Soleimany was a two-time captain of the MIT women’s tennis team. They met on the court as undergraduates&nbsp;and discovered they shared a passion for machine learning.&nbsp;</p> <p>After finishing their undergraduate degrees, they decided to challenge themselves and fill what they saw as an increasing need at MIT for a foundational course in deep learning.&nbsp;6.S191 was launched in 2017 by two grad students, Nick Locascio and Harini Suresh, and Amini and Soleimany had a vision for transforming the course into something more. They created a series of software labs, introduced new cutting-edge topics like robust and ethical AI, and added content to appeal to a broad range of students, from computer scientists to aerospace engineers and MBAs.</p> <p>“Alexander and I are constantly brainstorming, and those discussions are key to how 6.S191 and some of our own collaborative research projects have developed,” says Soleimany.&nbsp;</p> <p>They cover one of those research collaborations&nbsp;in class. During the computer vision lab, students learn about algorithmic bias and how to test for and address racial and gender bias in face-recognition tools. The lab is based on an algorithm that Amini and Soleimany developed with their respective advisors,&nbsp;<a href="" target="_blank">Daniela Rus</a>, director of CSAIL, and&nbsp;<a href="" target="_blank">Sangeeta Bhatia</a>, the John J. and Dorothy Wilson Professor of HST and EECS. This year they also covered hot topics in robotics, including recent work of Amini’s on driverless cars.&nbsp;</p> <p>But they don’t plan to stop there.&nbsp;“We’re committed to making 6.S191 the best that it can be, each year we teach it,” says Amini “and that means moving the course forward as deep learning continues to evolve.”&nbsp;</p> A crash course in deep learning organized and taught by grad students Alexander Amini (right) and Ava Soleimany reaches more than 350 MIT students each year; more than a million other people have watched their lectures online over the past three years.Photo: Gretchen ErtlQuest for Intelligence, Independent Activities Period, Electrical engineering and computer science (EECS), Harvard-MIT Health Sciences and Technology, Computer Science and Artificial Intelligence Laboratory (CSAIL), MIT-IBM Watson AI Lab, School of Engineering, Artificial intelligence, Algorithms, Computer science and technology, Machine learning, Classes and programs, Faculty A human-machine collaboration to defend against cyberattacks PatternEx merges human and machine expertise to spot and respond to hacks. Fri, 21 Feb 2020 14:12:18 -0500 Zach Winn | MIT News Office <p>Being a cybersecurity analyst at a large company today is a bit like looking for a needle in a haystack — if that haystack were hurtling toward you at fiber optic speed.</p> <p>Every day, employees and customers generate loads of data that establish a normal set of behaviors. An attacker will also generate data while using any number of techniques to infiltrate the system; the goal is to find that “needle” and stop it before it does any damage.</p> <p>The data-heavy nature of that task lends itself well to the number-crunching prowess of machine learning, and an influx of AI-powered systems have indeed flooded the cybersecurity market over the years. But such systems can come with their own problems, namely a never-ending stream of false positives that can make them more of a time suck than a time saver for security analysts.</p> <p>MIT startup PatternEx starts with the assumption that algorithms can’t protect a system on their own. The company has developed a closed loop approach whereby machine-learning models flag possible attacks and human experts provide feedback. The feedback is then incorporated into the models, improving their ability to flag only the activity analysts care about in the future.</p> <p>“Most machine learning systems in cybersecurity have been doing anomaly detection,” says Kalyan Veeramachaneni, a co-founder of PatternEx and a principal research scientist at MIT. “The problem with that, first, is you need a baseline [of normal activity]. Also, the model is usually unsupervised, so it ends up showing a lot of alerts, and people end up shutting it down. The big difference is that PatternEx allows the analyst to inform the system and then it uses that feedback to filter out false positives.”</p> <p>The result is an increase in analyst productivity. When compared to a generic anomaly detection software program, PatternEx’s Virtual Analyst Platform successfully identified 10 times more threats through the same number of daily alerts, and its advantage persisted even when the generic system gave analysts five times more alerts per day.</p> <p>First deployed in 2016, today the company’s system is being used by security analysts at large companies in a variety of industries along with firms that offer cybersecurity as a service.</p> <p><strong>Merging human and machine approaches to cybersecurity</strong></p> <p>Veeramachaneni came to MIT in 2009 as a postdoc and now directs a research group in the Laboratory for Information and Decision Systems. His work at MIT primarily deals with big data science and machine learning, but he didn’t think deeply about applying those tools to cybersecurity until a brainstorming session with PatternEx co-founders Costas Bassias, Uday Veeramachaneni, and Vamsi Korrapati in 2013.</p> <p>Ignacio Arnaldo, who worked with Veeramachaneni as a postdoc at MIT between 2013 and 2015, joined the company shortly after. Veeramachaneni and Arnaldo knew from their time building tools for machine-learning researchers at MIT that a successful solution would need to seamlessly integrate machine learning with human expertise.</p> <p>“A lot of the problems people have with machine learning arise because the machine has to work side by side with the analyst,” Veeramachaneni says, noting that detected attacks still must be presented to humans in an understandable way for further investigation. “It can’t do everything by itself. Most systems, even for something as simple as giving out a loan, is augmentation, not machine learning just taking decisions away from humans.”</p> <p>The company’s first partnership was with a large online retailer, which allowed the founders to train their models to identify potentially malicious behavior using real-world data. One by one, they trained their algorithms to flag different types of attacks using sources like Wi-Fi access logs, authentication logs, and other user behavior in the network.</p> <p>The early models worked best in retail, but Veeramachaneni knew how much businesses in other industries were struggling to apply machine learning in their operations from his many conversations with company executives at MIT (a subject PatternEx recently published <a href="">a paper</a> on).</p> <p>“MIT has done an incredible job since I got here 10 years ago bringing industry through the doors,” Veeramachaneni says. He estimates that in the past six years as a member of MIT’s Industrial Liaison Program he’s had 200 meetings with members of the private sector to talk about the problems they’re facing. He has also used those conversations to make sure his lab’s research is addressing relevant problems.</p> <p>In addition to enterprise customers, the company began offering its platform to security service providers and teams that specialize in hunting for undetected cyberattacks in networks.</p> <p>Today analysts can build machine learning models through PatternEx’s platform without writing a line of code, lowering the bar for people to use machine learning as part of a larger trend in the industry toward what Veeramachaneni calls the democratization of AI.</p> <p>“There’s not enough time in cybersecurity; it can’t take hours or even days to understand why an attack is happening,” Veeramachaneni says. “That’s why getting the analyst the ability to build and tweak machine learning models &nbsp;is the most critical aspect of our system.”</p> <p><strong>Giving security analysts an army</strong></p> <p>PatternEx’s Virtual Analyst Platform is designed to make security analysts feel like they have an army of assistants combing through data logs and presenting them with the most suspicious behavior on their network.</p> <p>The platform uses machine learning models to go through more than 50 streams of data and identify suspicious behavior. It then presents that information to the analyst for feedback, along with charts and other data visualizations that help the analyst decide how to proceed. After the analyst determines whether or not the behavior is an attack, that feedback is incorporated back into the models, which are updated across PatternEx’s entire customer base.</p> <p>“Before machine learning, someone would catch an attack, probably a little late, they might name it, and then they’ll announce it, and all the other companies will call and find out about it and go in and check their data,” Veeramachaneni says. “For us, if there’s an attack, we take that data, and because we have multiple customers, we have to transfer that in real time to other customer’s data to see if it’s happening with them too. We do that very efficiently on a daily basis.”</p> <p>The moment the system is up and running with new customers, it is able to identify 40 different types of cyberattacks using 170 different prepackaged machine learning models. Arnaldo notes that as the company works to grow those figures, customers are also adding to PatternEx’s model base by building solutions on the platform that address specific threats they’re facing.</p> <p>Even if customers aren’t building their own models on the platform, they can deploy PatternEx’s system out of the box, without any machine learning expertise, and watch it get smarter automatically.</p> <p>By providing that flexibility, PatternEx is bringing the latest tools in artificial intelligence to the people who understand their industries most intimately. It all goes back to the company’s founding principle of empowering humans with artificial intelligence instead of replacing them.</p> <p>“The target users of the system are not skilled data scientists or machine learning experts — profiles that are hard for cybersecurity teams to hire — but rather domain experts already on their payroll that have the deepest understanding of their data and uses cases,” Arnaldo says.</p> PatternEx’s Virtual Analyst Platform uses machine learning models to detect suspicious activity on a network. That activity is then presented to human analysts for feedback that improves the systems’ ability to flag activity analysts care about.Innovation and Entrepreneurship (I&E), Startups, Computer Science and Artificial Intelligence Laboratory (CSAIL), Machine learning, Artificial intelligence, Computer science and technology, Data, Cyber security, MIT Schwarzman College of Computing, Laboratory for Information and Decision Systems (LIDS) A road map for artificial intelligence policy In a Starr Forum talk, Luis Videgaray, director of MIT’s AI Policy for the World Project, outlines key facets of regulating new technologies. Thu, 20 Feb 2020 14:08:04 -0500 Peter Dizikes | MIT News Office <p>The rapid development of artificial intelligence technologies around the globe has led to increasing calls for robust AI policy: laws that let innovation flourish while protecting people from privacy violations, exploitive surveillance, biased algorithms, and more.</p> <p>But the drafting and passing of such laws has been anything but easy.</p> <p>“This is a very complex problem,” Luis Videgaray PhD ’98, director of MIT’s AI Policy for the World Project, said in a lecture on Wednesday afternoon. “This is not something that will be solved in a single report. This has got to be a collective conversation, and it will take a while. It will be years in the making.”</p> <p>Throughout his talk, Videgaray outlined an ambitious vision of AI policy around the globe, one that is sensitive to economic and political dynamics, and grounded in material fairness and democratic deliberation.&nbsp;&nbsp;&nbsp;</p> <p>“Trust is probably the most important problem we have,” Videgaray said.</p> <p>Videgaray’s talk, “From Principles to Implementation: The Challenge of AI Policy Around the World,” was part of the Starr Forum series of public discussions about topics of global concern. The Starr Forum is hosted by MIT’s Center for International Studies. Videgaray gave his remarks to a standing-room crowd of over 150 in MIT’s Building E25.</p> <p>Videgaray, who is also a senior lecturer at the MIT Sloan School of Management, previously served as the finance minister of Mexico from 2012 to 2016, and foreign minister of Mexico from 2017 to 2018. Videgaray has also worked extensively in investment banking.</p> <p><strong>Information lag and media hype</strong></p> <p>In his talk, Videgaray began by outlining several “themes” related to AI that he thinks policymakers should keep in mind. These include government uses of AI; the effects of AI on the economy, including the possibility it could help giant tech firms consolidate market power; social responsibility issues, such as privacy, fairness, and bias; and the implications of AI for democracy, at a time when bots can influence political discussion. Videgaray also noted a “geopolitics” of AI regulation — from China’s comprehensive efforts to control technology to the looser methods used in the U.S.</p> <p>Videgaray observed that it is difficult for AI regulators to stay current with technology.</p> <p>“There’s an information lag,” Videgaray said. “Things that concern computer scientists today might become the concerns of policymakers a few years in the future.”</p> <p>Moreover, he noted, media hype can distort perceptions of AI and its applications. Here Videgaray contrasted the <a href="">recent report</a> of MIT’s Task Force on the Future of Work, which finds uncertainty about how many jobs will be replaced with technology, with a recent television documentary presenting a picture of automated vehicles replacing all truck drivers.</p> <p>“Clearly the evidence is nowhere near [indicating] that all jobs in truck driving, in long-distance driving, are going to be lost,” he said. “That is not the case.”</p> <p>With these general issues in mind, what should policymakers do about AI now? Videgaray offered several concrete suggestions. For starters: Policymakers should no longer just outline general philosophical principles, something that has been done many times, with a general convergence of ideas occurring.</p> <p>“Working on principles has very, very small marginal returns,” Videgaray said. “We can go to the next phase … principles are a necessary but not sufficient condition for AI policy. Because policy is about making hard choices in uncertain conditions.”</p> <p>Indeed, he emphasized, more progress can be made by having many AI policy decisions be particular to specific industries. When it comes to, say, medical diagnostics, policymakers want technology “to be very accurate, but you also want it to be explainable, you want it to be fair, without bias, you want the information to be secure … there are many objectives that can conflict with each other. So, this is all about the tradeoffs.”&nbsp;</p> <p>In many cases, he said, algorithm-based AI tools could go through a rigorous testing process, as required in some other industries: “Pre-market testing makes sense,” Videgaray said. “We do that for drugs, clinical trials, we do that for cars, why shouldn’t we do pre-market testing for algorithms?”</p> <p>But while Videgaray sees value in industry-specific regulations, he is not as keen on having a patchwork of varying state-level AI laws being used to regulate technology in the U.S.</p> <p>“Is this a problem for Facebook, for Google? I don’t think so,” Videgaray said. “They have enough resources to navigate through this complexity. But what about startups? What about students from MIT or Cornell or Stanford that are trying to start something, and would have to go through, at the extreme, 55 [pieces of] legislation?”</p> <p><strong>A collaborative conversation</strong></p> <p>At the event, Videgaray was introduced by Kenneth Oye, a professor of political science at MIT who studies technological regulation, and who asked Videgaray questions after the lecture. Among other things, Oye suggested U.S. states could serve as a useful laboratory for regulatory innovation.</p> <p>“In an area characterized by significant uncertainty, complexity, and controversy, there can be benefits to experimentation, having different models being pursued in different areas to see which works best or worse,” Oye suggested.</p> <p>Videgaray did not necessarily disagree, but emphasized the value of an eventual convergence in regulation. The U.S. banking industry, he noted, also followed this trajectory, until “eventually the regulation we have for finance [became] federal,” rather than determined by states.</p> <p>Prior to his remarks, Videgaray acknowledged some audience members, including his PhD thesis adviser at MIT, James Poterba, the Mitsui Professor of Economics, whom Videgaray called “one of the best teachers, not only in economics but about a lot of things in life.” Mexico’s Consul General in Boston, Alberto Fierro, also attended the event.</p> <p>Ultimately, Videgaray emphasized to the audience, the future of AI policy will be collaborative.</p> <p>“You cannot just go to a computer lab and say, ‘Okay, get me some AI policy,’” he stressed. “This has got to be a collective conversation.”</p> Luis Videgaray, director of MIT’s AI Policy for the World Project, talking at his Starr Forum lecture, hosted by the Center for International Studies, on February 19, 2020.Images: courtesy of Laura Kerwin, Center for International StudiesArtificial intelligence, Law, Ethics, Computer science and technology, Political science, Economics, Special events and guest speakers, Global, Center for International Studies, School of Humanities, Arts, and Social Sciences, Sloan School of Management Artificial intelligence yields new antibiotic A deep-learning model identifies a powerful new drug that can kill many species of antibiotic-resistant bacteria. Thu, 20 Feb 2020 10:59:59 -0500 Anne Trafton | MIT News Office <p>Using a machine-learning algorithm, MIT researchers have identified a powerful new antibiotic compound. In laboratory tests, the drug killed many of the world’s most problematic disease-causing bacteria, including some strains that are resistant to all known antibiotics. It also cleared infections in two different mouse models.</p> <p>The computer model, which can screen more than a hundred million chemical compounds in a matter of days, is designed to pick out potential antibiotics that kill bacteria using different mechanisms than those of existing drugs.</p> <p>“We wanted to develop a platform that would allow us to harness the power of artificial intelligence to usher in a new age of antibiotic drug discovery,” says James Collins, the Termeer Professor of Medical Engineering and Science in MIT’s Institute for Medical Engineering and Science (IMES) and Department of Biological Engineering. “Our approach revealed this amazing molecule which is arguably one of the more powerful antibiotics that has been discovered.”</p> <p>In their new study, the researchers also identified several other promising antibiotic candidates, which they plan to test further. They believe the model could also be used to design new drugs, based on what it has learned about chemical structures that enable drugs to kill bacteria.</p> <p>“The machine learning model can explore, in silico, large chemical spaces that can be prohibitively expensive for traditional experimental approaches,” says Regina Barzilay, the Delta Electronics Professor of Electrical Engineering and Computer Science in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL).</p> <p>Barzilay and Collins, who are faculty co-leads for MIT’s Abdul Latif Jameel Clinic for Machine Learning in Health (J-Clinic), are the senior authors of the study, which appears today in <em>Cell</em>. The first author of the paper is Jonathan Stokes, a postdoc at MIT and the Broad Institute of MIT and Harvard.</p> <p><strong>A new pipeline</strong></p> <p>Over the past few decades, very few new antibiotics have been developed, and most of those newly approved antibiotics are slightly different variants of existing drugs. Current methods for screening new antibiotics are often prohibitively costly, require a significant time investment, and are usually limited to a narrow spectrum of chemical diversity.</p> <p>“We’re facing a growing crisis around antibiotic resistance, and this situation is being generated by both an increasing number of pathogens becoming resistant to existing antibiotics, and an anemic pipeline in the biotech and pharmaceutical industries for new antibiotics,” Collins says.</p> <p>To try to find completely novel compounds, he teamed up with Barzilay, Professor Tommi Jaakkola, and their students Kevin Yang, Kyle Swanson, and Wengong Jin, who have previously developed machine-learning computer models that can be trained to analyze the molecular structures of compounds and correlate them with particular traits, such as the ability to kill bacteria.</p> <p>The idea of using predictive computer models for “in silico” screening is not new, but until now, these models were not sufficiently accurate to transform drug discovery. Previously, molecules were represented as vectors reflecting the presence or absence of certain chemical groups. However, the new neural networks can learn these representations automatically, mapping molecules into continuous vectors which are subsequently used to predict their properties.</p> <p>In this case, the researchers designed their model to look for chemical features that make molecules effective at killing <em>E. coli</em>. To do so, they trained the model on about 2,500 molecules, including about 1,700 FDA-approved drugs and a set of 800 natural products with diverse structures and a wide range of bioactivities.</p> <p>Once the model was trained, the researchers tested it on the Broad Institute’s Drug Repurposing Hub, a library of about 6,000 compounds. The model picked out one molecule that was predicted to have strong antibacterial activity and had a chemical structure different from any existing antibiotics. Using a different machine-learning model, the researchers also showed that this molecule would likely have low toxicity to human cells.</p> <p>This molecule, which the researchers decided to call halicin, after the fictional artificial intelligence system from “2001: A Space Odyssey,” has been previously investigated as possible diabetes drug. The researchers tested it against dozens of bacterial strains isolated from patients and grown in lab dishes, and found that it was able to kill many that are resistant to treatment, including <em>Clostridium difficile</em>, <em>Acinetobacter baumannii</em>, and <em>Mycobacterium tuberculosis</em>. The drug worked against every species that they tested, with the exception of <em>Pseudomonas aeruginosa</em>, a difficult-to-treat lung pathogen.</p> <p>To test halicin’s effectiveness in living animals, the researchers used it to treat mice infected with <em>A. baumannii</em>, a bacterium that has infected many U.S. soldiers stationed in Iraq and Afghanistan. The strain of <em>A. baumannii</em> that they used is resistant to all known antibiotics, but application of a halicin-containing ointment completely cleared the infections within 24 hours.</p> <p>Preliminary studies suggest that halicin kills bacteria by disrupting their ability to maintain an electrochemical gradient across their cell membranes. This gradient is necessary, among other functions, to produce ATP (molecules that cells use to store energy), so if the gradient breaks down, the cells die. This type of killing mechanism could be difficult for bacteria to develop resistance to, the researchers say.</p> <p>“When you’re dealing with a molecule that likely associates with membrane components, a cell can’t necessarily acquire a single mutation or a couple of mutations to change the chemistry of the outer membrane. Mutations like that tend to be far more complex to acquire evolutionarily,” Stokes says.</p> <p>In this study, the researchers found that <em>E. coli</em> did not develop any resistance to halicin during a 30-day treatment period. In contrast, the bacteria started to develop resistance to the antibiotic ciprofloxacin within one to three days, and after 30 days, the bacteria were about 200 times more resistant to ciprofloxacin than they were at the beginning of the experiment.</p> <p>The researchers plan to pursue further studies of halicin, working with a pharmaceutical company or nonprofit organization, in hopes of developing it for use in humans.</p> <p><strong>Optimized molecules</strong></p> <p>After identifying halicin, the researchers also used their model to screen more than 100 million molecules selected from the ZINC15 database, an online collection of about 1.5 billion chemical compounds. This screen, which took only three days, identified 23 candidates that were structurally dissimilar from existing antibiotics and predicted to be nontoxic to human cells.</p> <p>In laboratory tests against five species of bacteria, the researchers found that eight of the molecules showed antibacterial activity, and two were particularly powerful. The researchers now plan to test these molecules further, and also to screen more of the ZINC15 database.</p> <p>The researchers also plan to use their model to design new antibiotics and to optimize existing molecules. For example, they could train the model to add features that would make a particular antibiotic target only certain bacteria, preventing it from killing beneficial bacteria in a patient’s digestive tract.</p> <p>“This groundbreaking work signifies a paradigm shift in antibiotic discovery and indeed in drug discovery more generally,” says Roy Kishony, a professor of biology and computer science at Technion (the Israel Institute of Technology), who was not involved in the study. “Beyond in silica screens, this approach will allow using deep learning at all stages of antibiotic development, from discovery to improved efficacy and toxicity through drug modifications and medicinal chemistry.”</p> <p>The research was funded by the Abdul Latif Jameel Clinic for Machine Learning in Health, the Defense Threat Reduction Agency, the Broad Institute, the DARPA Make-It Program, the Canadian Institutes of Health Research, the Canadian Foundation for Innovation, the Canada Research Chairs Program, the Banting Fellowships Program, the Human Frontier Science Program, the Pershing Square Foundation, the Swiss National Science Foundation, a National Institutes of Health Early Investigator Award, the National Science Foundation Graduate Research Fellowship Program, and a gift from Anita and Josh Bekenstein.</p> MIT researchers used a machine-learning algorithm to identify a drug called halicin that kills many strains of bacteria. Halicin (top row) prevented the development of antibiotic resistance in E. coli, while ciprofloxacin (bottom row) did not.Image: courtesy of the Collins Lab at MITResearch, Biological engineering, Electrical Engineering & Computer Science (eecs), Institute for Medical Engineering and Science (IMES), Computer Science and Artificial Intelligence Laboratory (CSAIL), Broad Institute, School of Engineering, Bacteria, Microbes, Medicine, Health, Machine learning, Artificial intelligence, Algorithms, J-Clinic, National Institutes of Health (NIH), National Science Foundation (NSF) Benjamin Chang: Might technology tip the global scales? MIT graduate student is assessing the impacts of artificial intelligence on military power, with a focus on the US and China. Wed, 19 Feb 2020 14:30:01 -0500 Leda Zimmerman | MIT Political Science <p>The United States and China seem locked in an ever-tightening embrace, superpowers entangled in a web of economic and military concerns. "Every issue critical to world order — whether climate change, terrorism, or trade — is clearly and closely intertwined with U.S.-China relations," says Benjamin Chang, a fourth-year PhD candidate in political science concentrating in international relations and security studies. "Competition between these nations will shape all outcomes anyone cares about in the next 50 years or more."</p> <p>Little surprise, then, that Chang is homing in on this relationship for his thesis, which broadly examines the impact of artificial intelligence on military power. As China and the United States circle each other as rivals and uneasy partners on the global stage, Chang hopes to learn what the integration of artificial intelligence in different domains might mean for the balance of power.</p> <p>"There is a set of questions related to how technology will be used in the world in general, where the U.S. and China are the two actors with the most influence," says Chang. "I want to know, for instance, how AI will affect strategic stability between them."</p> <p><strong>The nuclear balance</strong></p> <p>In the domain of military power, one question Chang has been pursuing is whether the use of AI in nuclear strategy offers a battlefield advantage. "For the U.S., the main issue involves locating China's elusive mobile missile launchers," Chang says. "The U.S. has satellite and other remote sensors that provide too much intelligence for human analysts, but AI, with its image classifiers based on deep learning, could sort through all this data to locate Chinese assets in a timely fashion."</p> <p>While Chang's data draws on publicly available information about each side's military capabilities, these sources can't provide specific numbers for China's nuclear arsenal. "We don't know if China has 250 or 300 nukes, so I design programs to run combat simulations with high and low numbers of weapons to try and isolate the effects of AI on combat outcomes." Chang credits J. Chappell Lawson, Vipin Narang, and Eric Heginbotham — his advisors in international relations and security studies — for helping shape his research methodology.</p> <p>If the United States develops the capacity to locate these mobile nuclear assets quickly, "that could change the battlefield outcome and hold China's arsenal at risk," says Chang. "And if China feels it isn't able to protect its nuclear arsenal, it might have an incentive to use it or lose it."</p> <p>In subsequent research, Chang will examine the impacts of AI on cybersecurity and on autonomous weaponry such as drones.</p> <p><strong>A start in policy debate</strong></p> <p>Pondering international and security issues began early for Chang. "I developed a big interest in these subjects through policy debate, which motivated me to read hundreds of pages and gave me a breadth and depth of knowledge on disparate topics," he says. "Debate exposed me to the study of military affairs and got me interested in America's role in the world generally."</p> <p>Chang's engagement with policy deepened at Princeton University, where he earned his BA summa cum laude from the Woodrow Wilson School of Public and International Affairs. While he knew he wanted to focus on foreign policy of some kind, his special focus on China came fortuitously: He was assigned to a junior seminar where students developed a working paper on "Building the Rule of Law in China." He took a series of Mandarin language courses, and produced a thesis comparing 19th century American nationalist behavior with modern-day Chinese nationalism.</p> <p>By graduation, Chang knew he wanted to aim for a career in national security and policy by way of a graduate school education. But he sought real-world seasoning first: a two-year stint as an analyst at Long Term Strategy Group, a Washington defense research firm. At LTSG, Chang facilitated wargames simulating Asia-Pacific conflicts, and wrote monographs on Chinese foreign policy, nuclear signaling, and island warfare doctrine.</p> <p><strong>Bridging a divide</strong></p> <p>Today, he is applying this expertise. "I'm trying to use my computer science understanding to bridge the gap between people working at a highly technical level of AI, and folks in security studies who are perhaps less familiar with the technology," he says. Propelled by a National Science Foundation Graduate Research Fellowship and a research fellowship with the Center for Security and Emerging Technology, Chang continues with his simulations and is beginning to write up some of his analysis. He thinks some of his findings might prove surprising.</p> <p>"There is an assumption — based on China's vast collection of personal data and surveillance of citizens — that AI is the means by which China will leapfrog the U.S. in military power," says Chang. "But I think this is wrong." In fact, the United States "has much more military-relevant data than China does, because it collects on so many platforms — in the deep ocean, and from satellites — that are a holdover from fighting the Soviet Union."</p> <p>Among Chang's research challenges: the fact that AI is not a mature technology and hasn't been fully implemented in modern militaries. "There's not yet much literature or data to draw on when assessing its impact," he notes. Also, he would like to nail down a good definition of AI for his field. "With current definitions of AI, thinking about its influence is a bit like investigating the effect of explosives on international affairs: you could be talking about nuclear weapons or dynamite and gunpowder," he says. "In my dissertation I'm attempting a scoping of AI so that it's more amenable to good political science analysis."</p> <p>Getting these ideas down on paper will be Chang's job for at least the next year. The writing occasionally feels like a struggle. "Some days I'll sit there and it won't come out, and other days, after a long walk along the Charles, I can write all day, and it feels good."</p> MIT political science PhD candidate Benjamin ChangPhoto: Benjamin ChangPolitical science, School of Humanities Arts and Social Sciences, Artificial intelligence, Security studies and military, China, Policy, Students, Graduate, postdoctoral, Computer science and technology SENSE.nano awards seed grants in optoelectronics, interactive manufacturing The mission of SENSE.nano is to foster the development and use of novel sensors, sensing systems, and sensing solutions. Thu, 13 Feb 2020 16:40:01 -0500 MIT.nano <p>SENSE.nano has announced the recipients of the third annual SENSE.nano seed grants. This year’s grants serve to advance innovations in sensing technologies for augmented and virtual realities (AR/VR) and advanced manufacturing systems.</p> <p>A center of excellence powered by MIT.nano, SENSE.nano received substantial interest in its 2019 call for proposals, making for stiff competition. Proposals were reviewed and evaluated by a committee consisting of industry and academia thought-leaders and were selected for funding following significant discussion. Ultimately, two projects were awarded $75,000 each to further research related to detecting movement in molecules and monitoring machine health.&nbsp;</p> <p>“SENSE.nano strives to&nbsp;convey the breadth and depth of sensing research at MIT," says Brian Anthony, co-leader of SENSE.nano, associate director of MIT.nano, and a principal&nbsp;research scientist in the Department of Mechanical Engineering. “As we work to grow SENSE.nano’s research footing and to attract partners, it is encouraging to know that so much important research — in sensors; sensor systems; and sensor science, engineering — is taking place at the Institute.”</p> <p>The projects receiving grants are:</p> <p><strong>P. Donald Keathley and Karl Berggren: Nanostructured optical-field samplers for visible to near-infrared time-domain spectroscopy</strong></p> <p>Research Scientist Phillip “Donnie” Keathley and Professor Karl Berggren from the Department of Electrical Engineering and Computer Science are developing a field-sampling technique using nanoscale structures and light waves to sense vibrational motion of molecules. Keathley is a member of Berggren’s quantum nanostructures and nanofabrication group in the Research Laboratory of Electronics (RLE). The two are investigating an all-on-chip nanoantenna device for sampling weak sub-femtojoule-level electronic fields, in the near-infrared and visible spectrums.</p> <p>Current technology for sampling these spectra of optical energy requires a large apparatus — there is no compact device with enough sensitivity to detect the low-energy signals. Keathley and Berggren propose using plasmonic nanoantennas for measuring low-energy pulses. This technology could have significant impacts on the medical and food-safety industries by revolutionizing the accurate detection and identification of chemicals and bio-chemicals.</p> <p><strong>Jeehwan Kim: Interactive manufacturing enabled by simultaneous sensing and recognition</strong></p> <p>Jeehwan Kim, associate professor with a dual appointment in mechanical engineering and materials science and engineering, proposes an ultra-sensitive sensor system using neuromorphic chips to improve advanced manufacturing through real-time monitoring of machines. Machine failures compromise productivity and cost. Sensors that can instantly process data to provide real-time feedback would be a valuable tool for preventive maintenance of factory machines.</p> <p>Kim’s group, also part of RLE, aims to develop single-crystalline gallium nitride sensors that, when connected to AI chips, will create a feedback loop with the factory machines. Failure patterns would be recognized by the AI hardware, creating an intelligent manufacturing system that can predict and prevent failures. These sensors will have the sensitivity to navigate noisy factory environments, be small enough to form dense arrays, and have the power efficiency to be used on a large number of manufacturing machines.</p> <p>The mission of SENSE.nano is to foster the development and use of novel sensors, sensing systems, and sensing solutions in order to provide previously unimaginable insight into the condition of our world. Two new calls for seed grant proposals will open later this year in conjunction with the Immersion Lab NCSOFT collaboration and then with the SENSE.nano 2020 symposium.</p> <p>In addition to seed grants and the annual conference, SENSE.nano recently launched Talk SENSE — a monthly series for MIT students to further engage with these topics and connect with experts working in sensing technologies.</p> A center of excellence powered by MIT.nano, SENSE.nano received substantial interest in its 2019 call for proposals, making for stiff competition.Photo: David SellaMIT.nano, Mechanical engineering, Electrical engineering and computer science (EECS), Materials Science and Engineering, Nanoscience and nanotechnology, Awards, honors and fellowships, Augmented and virtual reality, Computer science and technology, Artificial intelligence, Research, Funding, Grants, Sensors, School of Engineering, Research Laboratory of Electronics MIT researchers identify security vulnerabilities in voting app Mobile voting application could allow hackers to alter individual votes and may pose privacy issues for users. Thu, 13 Feb 2020 03:00:00 -0500 Abby Abazorius | MIT News Office <p>In recent years, there has been a growing interest in using internet and mobile technology to increase access to the voting process. At the same time, computer security experts caution that paper ballots are the only secure means of voting.</p> <p>Now, MIT researchers are raising another concern: They say they have uncovered security vulnerabilities in a mobile voting application that was used during the 2018 midterm elections in West Virginia. Their security analysis of the application, called Voatz, pinpoints a number of weaknesses, including the opportunity for hackers to alter, stop, or expose how an individual user has voted. Additionally, the researchers found that Voatz’s use of a third-party vendor for voter identification and verification poses potential privacy issues for users.</p> <p>The findings are described in a new <a href="">technical paper</a> by Michael Specter, a graduate student in MIT’s Department of Electrical Engineering and Computer Science (EECS) and a member of MIT’s <a href="">Internet Policy Research Initiative</a>, and James Koppel, also a graduate student in EECS. The research was conducted under the guidance of Daniel Weitzner, a principal research scientist at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL) and founding director of the Internet Policy Research Initiative.</p> <p>After uncovering these security vulnerabilities, the researchers disclosed their findings to the Department of Homeland Security’s Cybersecurity and Infrastructure Agency (CISA). The researchers, along with the Boston University/MIT Technology Law Clinic, worked in close coordination with election security officials within CISA to ensure that impacted elections officials and the vendor were aware of the findings before the research was made public. This included preparing written summaries of the findings with proof-of-concept code, and direct discussions with affected elections officials on calls arranged by CISA.</p> <p>In addition to its use in the 2018 West Virginia elections, the app was deployed in elections in Denver, Oregon, and Utah, as well as at the 2016 Massachusetts Democratic Convention and the 2016 Utah Republican Convention. Voatz was not used during the 2020 Iowa caucuses.</p> <p>The findings underscore the need for transparency in the design of voting systems, according to the researchers.</p> <p>“We all have an interest in increasing access to the ballot, but in order to maintain trust in our elections system, we must assure that voting systems meet the high technical and operation security standards before they are put in the field,” says Weitzner. “We cannot experiment on our democracy.”&nbsp;&nbsp;&nbsp; &nbsp;</p> <p>“The consensus of security experts is that running a secure election over the internet is not possible today,” adds Koppel. “The reasoning is that weaknesses anywhere in a large chain can give an adversary undue influence over an election, and today’s software is shaky enough that the existence of unknown exploitable flaws is too great a risk to take.”</p> <p><strong>Breaking down the results</strong></p> <p>The researchers were initially inspired to perform a security analysis of Voatz based on Specter’s <a href="">research</a> with Ronald Rivest, Institute Professor at MIT; Neha Narula, director of the MIT Digital Currency Initiative; and Sunoo Park SM ’15, PhD ’18 , exploring the feasibility of using blockchain systems in elections. According to the researchers, Voatz claims to use a permissioned blockchain to ensure security, but has not released any source code or public documentation for how their system operates.</p> <p>Specter, who co-teaches an MIT <a href="">Independent Activities Period course</a> founded by Koppel that is focused on reverse engineering software, broached the idea of reverse engineering Voatz’s application, in an effort to better understand how its system worked. To ensure that they did not interfere with any ongoing elections or expose user records, Specter and Koppel reverse-engineered the application and then created a model of Voatz’s server.</p> <p>They found that an adversary with remote access to the device can alter or discover a user’s vote, and that the server, if hacked, could easily change those votes. “It does not appear that the app’s protocol attempts to verify [genuine votes] with the back-end blockchain,” Specter explains.</p> <p>“Perhaps most alarmingly, we found that a passive network adversary, like your internet service provider, or someone nearby you if you’re on unencrypted Wi-Fi, could detect which way you voted in some configurations of the election. Worse, more aggressive attackers could potentially detect which way you’re going to vote and then stop the connection based on that alone.”</p> <p>In addition to detecting vulnerabilities with Voatz’s voting process, Specter and Koppel found that the app poses privacy issues for users. As the app uses an external vendor for voter ID verification, a third party could potentially access a voter’s photo, driver’s license data, or other forms of identification, if that vendor’s platform isn’t also secure.&nbsp;&nbsp;&nbsp;&nbsp; &nbsp;</p> <p>“Though Voatz’s privacy policy does talk about sending some information to third parties, as far as we can tell the fact that any third party is getting the voter’s driver’s license and selfie isn’t explicitly mentioned,” Specter notes.</p> <p><strong>Calls for increased openness</strong></p> <p>Specter and Koppel say that their findings point to the need for openness when it comes to election administration, in order to ensure the integrity of the election process. Currently, they note, the election process in states that use paper ballots is designed to be transparent, and citizens and political party representatives are given opportunities to observe the voting process.</p> <p>In contrast, Koppel notes, “Voatz’s app and infrastructure were completely closed-source; we were only able to get access to the app itself.&nbsp;&nbsp;&nbsp; &nbsp;</p> <p>“I think this type of analysis is extremely important. Right now, there’s a drive to make voting more accessible, by using internet and mobile-based voting systems. The problem here is that sometimes those systems aren’t made by people who have expertise in keeping voting systems secure, and they’re deployed before they can get proper review,” says Matthew Green, an associate professor at the Johns Hopkins Information Security Institute. In the case of Voatz, he adds, “It looks like there were many good intentions here, but the result lacks key features that would protect a voter and protect the integrity of elections.”</p> <p>Going forward, the researchers caution that software developers should prove their systems are as secure as paper ballots.</p> <p>“The biggest issue is transparency,” says Specter. “When you have part of the election that is opaque, that is not viewable, that is not public, that has some sort of proprietary component, that part of the system is inherently suspect and needs to be put under a lot of scrutiny.”</p> Cyber security, Voting and elections, Computer science and technology, Apps, Technology and society, Internet Policy Research Initiative, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering “Sensorized” skin helps soft robots find their bearings Flexible sensors and an artificial intelligence model tell deformable robots how their bodies are positioned in a 3D environment. Wed, 12 Feb 2020 23:59:59 -0500 Rob Matheson | MIT News Office <p>For the first time, MIT researchers have enabled a soft robotic arm to understand its configuration in 3D space, by leveraging only motion and position data from its own “sensorized” skin.</p> <p>Soft robots constructed from highly compliant materials, similar to those found in living organisms, are being championed as safer, and more adaptable, resilient, and bioinspired alternatives to traditional rigid robots. But giving autonomous control to these deformable robots is a monumental task because they can move in a virtually infinite number of directions at any given moment. That makes it difficult to train planning and control models that drive automation.</p> <p>Traditional methods to achieve autonomous control use large systems of multiple motion-capture cameras that provide the robots feedback about 3D movement and positions. But those are impractical for soft robots in real-world applications.</p> <p>In a paper being published in the journal <em>IEEE Robotics and Automation Letters</em>, the researchers describe a system of soft sensors that cover a robot’s body to provide “proprioception” — meaning awareness of motion and position of its body. That feedback runs into a novel deep-learning model that sifts through the noise and captures clear signals to estimate the robot’s 3D configuration. The researchers validated their system on a soft robotic arm resembling an elephant trunk, that can predict its own position as it autonomously swings around and extends.</p> <p>The sensors can be fabricated using off-the-shelf materials, meaning any lab can develop their own systems, says Ryan Truby, a postdoc in the MIT Computer Science and Artificial Laboratory (CSAIL) who is co-first author on the paper along with CSAIL postdoc Cosimo Della Santina.</p> <p>“We’re sensorizing soft robots to get feedback for control from sensors, not vision systems, using a very easy, rapid method for fabrication,” he says. “We want to use these soft robotic trunks, for instance, to orient and control themselves automatically, to pick things up and interact with the world. This is a first step toward that type of more sophisticated automated control.”</p> <p>One future aim is to help make artificial limbs that can more dexterously handle and manipulate objects in the environment. “Think of your own body: You can close your eyes and reconstruct the world based on feedback from your skin,” says co-author Daniela Rus, director of CSAIL and the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science. “We want to design those same capabilities for soft robots.”</p> <p><strong>Shaping soft sensors</strong></p> <p>A longtime goal in soft robotics has been fully integrated body sensors. Traditional rigid sensors detract from a soft robot body’s natural compliance, complicate its design and fabrication, and can cause various mechanical failures. Soft-material-based sensors are a more suitable alternative, but require specialized materials and methods for their design, making them difficult for many robotics labs to fabricate and integrate in soft robots.</p> <p>While working in his CSAIL lab one day looking for inspiration for sensor materials, Truby made an interesting connection. “I found these sheets of conductive materials used for electromagnetic interference shielding, that you can buy anywhere in rolls,” he says. These materials have “piezoresistive” properties, meaning they change in electrical resistance when strained. Truby realized they could make effective soft sensors if they were placed on certain spots on the trunk. As the sensor deforms in response to the trunk’s stretching and compressing, its electrical resistance is converted to a specific output voltage. The voltage is then used as a signal correlating to that movement.</p> <p>But the material didn’t stretch much, which would limit its use for soft robotics. Inspired by kirigami —&nbsp;a variation of origami that includes making cuts in a material — Truby designed and laser-cut rectangular strips of conductive silicone sheets into various patterns, such as rows of tiny holes or crisscrossing slices like a chain link fence. That made them far more flexible, stretchable, “and beautiful to look at,” Truby says.</p> <p><img alt="" src="/sites/" style="width: 500px; height: 281px;" /></p> <p><img alt="" src="/sites/" style="width: 500px; height: 281px;" /></p> <p><em style="font-size: 10px;">Credit: Ryan L. Truby, MIT&nbsp;CSAIL</em></p> <p>The researchers’ robotic trunk comprises three segments, each with four fluidic actuators (12 total) used to move the arm. They fused one sensor over each segment, with each sensor covering and gathering data from one embedded actuator in the soft robot. They used “plasma bonding,” a technique that energizes a surface of a material to make it bond to another material. It takes roughly a couple hours to shape dozens of sensors that can be bonded to the soft robots using a handheld plasma-bonding device.</p> <p><img alt="" src="/sites/" style="width: 500px; height: 281px;" /></p> <p><span style="font-size:10px;"><em>Credit: Ryan L. Truby, MIT&nbsp;CSAIL</em></span></p> <p><strong>“Learning” configurations</strong></p> <p>As hypothesized, the sensors did capture the trunk’s general movement. But they were really noisy. “Essentially, they’re nonideal sensors in many ways,” Truby says. “But that’s just a common fact of making sensors from soft conductive materials. Higher-performing and more reliable sensors require specialized tools that most robotics labs do not have.”</p> <p>To estimate the soft robot’s configuration using only the sensors, the researchers built a deep neural network to do most of the heavy lifting, by sifting through the noise to capture meaningful feedback signals. The researchers developed a new model to kinematically describe the soft robot’s shape that vastly reduces the number of variables needed for their model to process.</p> <p>In experiments, the researchers had the trunk swing around and extend itself in random configurations over approximately an hour and a half. They used the traditional motion-capture system for ground truth data. In training, the model analyzed data from its sensors to predict a configuration, and compared its predictions to that ground truth data which was being collected simultaneously. In doing so, the model “learns” to map signal patterns from its sensors to real-world configurations. Results indicated, that for certain and steadier configurations, the robot’s estimated shape matched the ground truth.</p> <p>Next, the researchers aim to explore new sensor designs for improved sensitivity and to develop new models and deep-learning methods to reduce the required training for every new soft robot. They also hope to refine the system to better capture the robot’s full dynamic motions.</p> <p>Currently, the neural network and sensor skin are not sensitive to capture subtle motions or dynamic movements. But, for now, this is an important first step for learning-based approaches to soft robotic control, Truby says: “Like our soft robots, living systems don’t have to be totally precise. Humans are not precise machines, compared to our rigid robotic counterparts, and we do just fine.”</p> MIT researchers have created a “sensorized” skin, made with kirigami-inspired sensors, that gives soft robots greater awareness of the motion and position of their bodies.Ryan L. Truby, MIT CSAILResearch, Computer science and technology, Algorithms, Robots, Robotics, Soft robotics, Design, Machine learning, Materials Science and Engineering, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering Automated system can rewrite outdated sentences in Wikipedia articles Text-generating tool pinpoints and replaces specific information in sentences while retaining humanlike grammar and style. Wed, 12 Feb 2020 13:51:56 -0500 Rob Matheson | MIT News Office <p>A system created by MIT researchers could be used to automatically update factual inconsistencies in Wikipedia articles, reducing time and effort spent by human editors who now do the task manually.</p> <p>Wikipedia comprises millions of articles that are in constant need of edits to reflect new information. That can involve article expansions, major rewrites, or more routine modifications such as updating numbers, dates, names, and locations. Currently, humans across the globe volunteer their time to make these edits.&nbsp;&nbsp;</p> <p>In a paper being presented at the AAAI Conference on Artificial Intelligence, the researchers describe a text-generating system that pinpoints and replaces specific information in relevant Wikipedia sentences, while keeping the language similar to how humans write and edit.</p> <p>The idea is that humans would type into an interface an unstructured sentence with updated information, without needing to worry about style or grammar. The system would then search Wikipedia, locate the appropriate page and outdated sentence, and rewrite it in a humanlike fashion. In the future, the researchers say, there’s potential to build a fully automated system that identifies and uses the latest information from around the web to produce rewritten sentences in corresponding Wikipedia articles that reflect updated information.</p> <p>“There are so many updates constantly needed to Wikipedia articles. It would be beneficial to automatically modify exact portions of the articles, with little to no human intervention,” says Darsh Shah, a PhD student in the Computer Science and Artificial Intelligence Laboratory (CSAIL) and one of the lead authors. “Instead of hundreds of people working on modifying each Wikipedia article, then you’ll only need a few, because the model is helping or doing it automatically. That offers dramatic improvements in efficiency.”</p> <p>Many other bots exist that make automatic Wikipedia edits. Typically, those work on mitigating vandalism or dropping some narrowly defined information into predefined templates, Shah says. The researchers’ model, he says, solves a harder artificial intelligence problem: Given a new piece of unstructured information, the model automatically modifies the sentence in a humanlike fashion. “The other [bot] tasks are more rule-based, while this is a task requiring reasoning over contradictory parts in two sentences and generating a coherent piece of text,” he says.</p> <p>The system can be used for other text-generating applications as well, says co-lead author and CSAIL graduate student Tal Schuster. In their paper, the researchers also used it to automatically synthesize sentences in a popular fact-checking dataset that helped reduce bias, without manually collecting additional data. “This way, the performance improves for automatic fact-verification models that train on the dataset for, say, fake news detection,” Schuster says.</p> <p>Shah and Schuster worked on the paper with their academic advisor Regina Barzilay, the Delta Electronics Professor of Electrical Engineering and Computer Science and a professor in CSAIL.</p> <p><strong>Neutrality masking and fusing</strong></p> <p>Behind the system is a fair bit of text-generating ingenuity in identifying contradictory information between, and then fusing together, two separate sentences. It takes as input an “outdated” sentence from a Wikipedia article, plus a separate “claim” sentence that contains the updated and conflicting information. The system must automatically delete and keep specific words in the outdated sentence, based on information in the claim, to update facts but maintain style and grammar. That’s an easy task for humans, but a novel one in machine learning.</p> <p>For example, say there’s a required update to this sentence (in bold): “Fund A considers <strong>28 of their 42</strong> minority stakeholdings in operationally active companies to be of particular significance to the group.” The claim sentence with updated information may read: “Fund A considers <strong>23 of 43</strong> minority stakeholdings significant.” The system would locate the relevant Wikipedia text for “Fund A,” based on the claim. It then automatically strips out the outdated numbers (28 and 42) and replaces them with the new numbers (23 and 43), while keeping the sentence exactly the same and grammatically correct. (In their work, the researchers ran the system on a dataset of specific Wikipedia sentences, not on all Wikipedia pages.)</p> <p>The system was trained on a popular dataset that contains pairs of sentences, in which one sentence is a claim and the other is a relevant Wikipedia sentence. Each pair is labeled in one of three ways: “agree,” meaning the sentences contain matching factual information; “disagree,” meaning they contain contradictory information; or “neutral,” where there’s not enough information for either label. The system must make all disagreeing pairs agree, by modifying the outdated sentence to match the claim. That requires using two separate models to produce the desired output.</p> <p>The first model is a fact-checking classifier — pretrained to label each sentence pair as “agree,” “disagree,” or “neutral” — that focuses on disagreeing pairs. Running in conjunction with the classifier is a custom “neutrality masker” module that identifies which words in the outdated sentence contradict the claim. The module removes the minimal number of words required to “maximize neutrality” — meaning the pair can be labeled as neutral. That’s the starting point: While the sentences don’t agree, they no longer contain obviously contradictory information. The module creates a binary “mask” over the outdated sentence, where a 0 gets placed over words that most likely require deleting, while a 1 goes on top of keepers.</p> <p>After masking, a novel two-encoder-decoder framework is used to generate the final output sentence. This model learns compressed representations of the claim and the outdated sentence. Working in conjunction, the two encoder-decoders fuse the dissimilar words from the claim, by sliding them into the spots left vacant by the deleted words (the ones covered with 0s) in the outdated sentence.</p> <p>In one test, the model scored higher than all traditional methods, using a technique called “SARI” that measures how well machines delete, add, and keep words compared to the way humans modify sentences. They used a dataset with manually edited Wikipedia sentences, which the model hadn’t seen before. Compared to several traditional text-generating methods, the new model was more accurate in making factual updates and its output more closely resembled human writing. In another test, crowdsourced humans scored the model (on a scale of 1 to 5) based on how well its output sentences contained factual updates and matched human grammar. The model achieved average scores of 4 in factual updates and 3.85 in matching grammar.</p> <p><strong>Removing bias</strong></p> <p>The study also showed that the system can be used to augment datasets to eliminate bias when training detectors of “fake news,” a form of propaganda containing disinformation created to mislead readers in order to generate website views or steer public opinion. Some of these detectors train on datasets of agree-disagree sentence pairs to “learn” to verify a claim by matching it to given evidence.</p> <p>In these pairs, the claim will either match certain information with a supporting “evidence” sentence from Wikipedia (agree) or it will be modified by humans to include information contradictory to the evidence sentence (disagree). The models are trained to flag claims with refuting evidence as “false,” which can be used to help identify fake news.</p> <p>Unfortunately, such datasets currently come with unintended biases, Shah says: “During training, models use some language of the human written claims as “give-away” phrases to mark them as false, without relying much on the corresponding evidence sentence. This reduces the model’s accuracy when evaluating real-world examples, as it does not perform fact-checking.”</p> <p>The researchers used the same deletion and fusion techniques from their Wikipedia project to balance the disagree-agree pairs in the dataset and help mitigate the bias. For some “disagree” pairs, they used the modified sentence’s false information to regenerate a fake “evidence” supporting sentence. Some of the give-away phrases then exist in both the “agree” and “disagree” sentences, which forces models to analyze more features. Using their augmented dataset, the researchers reduced the error rate of a popular fake-news detector by 13 percent.</p> <p>“If you have a bias in your dataset, and you’re fooling your model into just looking at one sentence in a disagree pair to make predictions, your model will not survive the real world,” Shah says. “We make models look at both sentences in all agree-disagree pairs.”</p> MIT researchers have created an automated text-generating system that pinpoints and replaces specific information in relevant Wikipedia sentences, while keeping the language similar to how humans write and edit.Image: Christine Daniloff, MITResearch, Computer science and technology, Algorithms, Machine learning, Data, Internet, Crowdsourcing, Social media, Technology and society, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering Maintaining the equipment that powers our world By organizing performance data and predicting problems, Tagup helps energy companies keep their equipment running. Wed, 12 Feb 2020 09:39:37 -0500 Zach Winn | MIT News Office <p>Most people only think about the systems that power their cities when something goes wrong. Unfortunately, many people in the San Francisco Bay Area had a lot to think about recently when their utility company began scheduled power outages in an attempt to prevent wildfires. The decision came after devastating fires last year were found to be the result of faulty equipment, including transformers.</p> <p>Transformers are the links between power plants, power transmission lines, and distribution networks. If something goes wrong with a transformer, entire power plants can go dark. To fix the problem, operators work around the clock to assess various components of the plant, consider disparate data sources, and decide what needs to be repaired or replaced.</p> <p>Power equipment maintenance and failure is such a far-reaching problem it’s difficult to attach a dollar sign to. Beyond the lost revenue of the plant, there are businesses that can’t operate, people stuck in elevators and subways, and schools that can’t open.</p> <p>Now the startup Tagup is working to modernize the maintenance of transformers and other industrial equipment. The company’s platform lets operators view all of their data streams in one place and use machine learning to estimate if and when components will fail.</p> <p>Founded by CEO Jon Garrity ’11 and CTO Will Vega-Brown ’11, SM ’13 —&nbsp;who recently completed his PhD program in MIT’s Department of Mechanical Engineering and will be graduating this month — Tagup is currently being used by energy companies to monitor approximately 60,000 pieces of equipment around North America and Europe. That includes transformers, offshore wind turbines, and reverse osmosis systems for water filtration, among other things.</p> <p>“Our mission is to use AI to make the machines that power the world safer, more reliable, and more efficient,” Garrity says.</p> <p><strong>A light bulb goes on</strong></p> <p>Vega-Brown and Garrity crossed paths in a number of ways at MIT over the years. As undergraduates, they took a few of the same courses, with Vega-Brown double majoring in mechanical engineering and physics and Garrity double majoring in economics and physics. They were also fraternity brothers as well as teammates on the football team.</p> <p>Garrity was&nbsp;first exposed&nbsp;to entrepreneurship as an undergraduate in MIT’s Energy Ventures class and in the Martin Trust Center for Entrepreneurship.&nbsp;Later, when Garrity returned to campus while attending Harvard Business School and Vega-Brown was pursuing his doctorate, they were again classmates in MIT’s New Enterprises course.</p> <p>Still, the founders didn’t think about starting a company until 2015, after Garrity had worked at GE Energy and Vega-Brown was well into his PhD work at MIT’s Computer Science and Artificial Intelligence Laboratory.</p> <p>At GE, Garrity discovered an intriguing business model through which critical assets like jet engines were leased by customers — in this case airlines — rather than purchased, and manufacturers held responsibility for remotely monitoring and maintaining them. The arrangement allowed GE and others to leverage their engineering expertise while the customers focused on their own industries.</p> <p>"When I worked at GE, I always wondered: Why isn’t this service available for any equipment type? The answer is economics.” Garrity says. “It is expensive to set up a remote monitoring center, to instrument the equipment in the field, to staff the 50 or more engineering subject matter experts, and to provide the support required to end customers. The cost of equipment failure, both in terms of business interruption and equipment breakdown, must be enormous to justify the high average fixed cost."</p> <p>“We realized two things,” Garrity continues. “With the increasing availability of sensors and cloud infrastructure, we can dramatically reduce the cost [of monitoring critical assets] from the infrastructure and communications side. And, with new machine-learning methods, we can increase the productivity of engineers who review equipment data manually.”</p> <p>That realization led to Tagup, though it would take time to prove the founders’ technology. “The problem with using AI for industrial applications is the lack of high-quality data,” Vega-Brown explains. “Many of our customers have giant datasets, but the information density in industrial data is often quite low. That means we need to be very careful in how we hunt for signal and validate our models, so that we can reliably make accurate forecasts and predictions.”</p> <p>The founders leveraged their MIT ties to get the company off the ground. They received guidance from MIT’s Venture Mentoring Service, and Tagup was in the first cohort of startups accepted into the MIT Industrial Liaison Program’s (ILP) STEX 25 accelerator, which connects high potential startups with members of industry. Tagup has since secured several customers through ILP, and those early partnerships helped the company train and validate some of its machine-learning models.</p> <p><strong>Making power more reliable</strong></p> <p>Tagup’s platform combines all of a customer’s equipment data into one sortable master list that displays the likelihood of each asset causing a disruption. Users can click on specific assets to see charts of historic data and trends that feed into Tagup’s models.</p> <p>The company doesn’t deploy any sensors of its own. Instead, it combines customers’ real-time sensor measurements with other data sources like maintenance records and machine parameters to improve its proprietary machine-learning models.</p> <p>The founders also began with a focused approach to building their system. Transformers were one of the first types of equipment they worked with, and they’ve expanded to other groups of assets gradually.</p> <p>Tagup’s first deployment was in August of 2016 with a power plant that faces the Charles River close to MIT’s campus. Just a few months after it was installed, Garrity was at a meeting overseas when he got a call from the plant manager about a transformer that had just gone offline unexpectedly. From his phone, Garrity was able to inspect real-time data from the transformer&nbsp;and give the manager the information he needed to restart the system. Garrity says it saved the plant about 26 hours of downtime and $150,000 in revenue.</p> <p>“These are really catastrophic events in terms of business outcomes,” Garrity says, noting transformer failures are estimated to cost $23 billion annually.</p> <p>Since then they’ve secured partnerships with several large utility companies, including National Grid and Consolidated Edison Company of New York.</p> <p>Down the line, Garrity and Vega-Brown are excited about using machine learning to control the operation of equipment. For example, a machine could manage itself in the same way an autonomous car can sense an obstacle and steer around it.</p> <p>Those capabilities have major implications for the systems that ensure the lights go on when we flip switches at night.</p> <p>“Where it gets really exciting is moving toward optimization,” Garrity says. Vega-Brown agrees, adding, “Enormous amounts of power and water are wasted because there aren't enough experts to tune the controllers on every industrial machine in the world. If we can use AI to capture some of the expert knowledge in an algorithm, we can cut inefficiency and improve safety at scale.”</p> Tagup's industrial equipment monitoring platform is currently being used by energy companies to monitor approximately 60,000 pieces of equipment around North America and Europe. That includes transformers, offshore wind turbines, and reverse osmosis systems for water filtration.Innovation and Entrepreneurship (I&E), Startups, Alumni/ae, Computer Science and Artificial Intelligence Laboratory (CSAIL), Mechanical engineering, School of Engineering, Machine learning, Energy, Artificial intelligence Bridging the gap between human and machine vision Researchers develop a more robust machine-vision architecture by studying how human vision responds to changing viewpoints of objects. Tue, 11 Feb 2020 16:40:01 -0500 Kris Brewer | Center for Brains, Minds and Machines <p>Suppose you look briefly from a few feet away at a person you have never met before. Step back a few paces and look again. Will you be able to recognize her face? “Yes, of course,” you probably are thinking. If this is true, it would mean that our visual system, having seen a single image of an object such as a specific face, recognizes it robustly despite changes to the object’s position and scale, for example. On the other hand, we know that state-of-the-art classifiers, such as vanilla deep networks, will fail this simple test.</p> <p>In order to recognize a specific face under a range of transformations, neural networks need to be trained with many examples of the face under the different conditions. In other words, they can achieve invariance through memorization, but cannot do it if only one image is available. Thus, understanding how human vision can pull off this remarkable feat is relevant for engineers aiming to improve their existing classifiers. It also is important for neuroscientists modeling the primate visual system with deep networks. In particular, it is possible that the invariance with one-shot learning exhibited by biological vision requires a rather different computational strategy than that of deep networks.&nbsp;</p> <p>A new paper by MIT PhD candidate in electrical engineering and computer science Yena Han and colleagues in <em>Nature Scientific Reports</em> entitled “Scale and translation-invariance for novel objects in human vision” discusses how they study this phenomenon more carefully to create novel biologically inspired networks.</p> <p>"Humans can learn from very few examples, unlike deep networks. This is a huge difference with vast implications for engineering of vision systems and for understanding how human vision really works," states co-author Tomaso Poggio — director of the Center for Brains, Minds and Machines (CBMM) and the Eugene McDermott Professor of Brain and Cognitive Sciences at MIT. "A key reason for this difference is the relative invariance of the primate visual system to scale, shift, and other transformations. Strangely, this has been mostly neglected in the AI community, in part because the psychophysical data were so far less than clear-cut. Han's work has now established solid measurements of basic invariances of human vision.”</p> <p>To differentiate invariance rising from intrinsic computation with that from experience and memorization, the new study measured the range of invariance in one-shot learning. A one-shot learning task was performed by presenting Korean letter stimuli to human subjects who were unfamiliar with the language. These letters were initially presented a single time under one specific condition and tested at different scales or positions than the original condition. The first experimental result is that — just as you guessed — humans showed significant scale-invariant recognition after only a single exposure to these novel objects. The second result is that the range of position-invariance is limited, depending on the size and placement of objects.</p> <p>Next, Han and her colleagues performed a comparable experiment in deep neural networks designed to reproduce this human performance. The results suggest that to explain invariant recognition of objects by humans, neural network models should explicitly incorporate built-in scale-invariance. In addition, limited position-invariance of human vision is better replicated in the network by having the model neurons’ receptive fields increase as they are further from the center of the visual field. This architecture is different from commonly used neural network models, where an image is processed under uniform resolution with the same shared filters.</p> <p>“Our work provides a new understanding of the brain representation of objects under different viewpoints. It also has implications for AI, as the results provide new insights into what is a good architectural design for deep neural networks,” remarks Han, CBMM researcher and lead author of the study.</p> <p>Han and Poggio were joined by Gemma Roig and Gad Geiger in the work.</p> Yena Han (left) and Tomaso Poggio stand with an example of the visual stimuli used in a new psychophysics study.Photo: Kris BrewerCenter for Brains Minds and Machines, Brain and cognitive sciences, Machine learning, Artificial intelligence, Computer vision, Research, School of Science, Computer science and technology, Electrical Engineering & Computer Science (eecs), School of Engineering Brainstorming energy-saving hacks on Satori, MIT’s new supercomputer Three-day hackathon explores methods for making artificial intelligence faster and more sustainable. Tue, 11 Feb 2020 11:50:01 -0500 Kim Martineau | MIT Quest for Intelligence <p>Mohammad Haft-Javaherian planned to spend an hour at the&nbsp;<a href="">Green AI Hackathon</a>&nbsp;— just long enough to get acquainted with MIT’s new supercomputer,&nbsp;<a href="">Satori</a>. Three days later, he walked away with $1,000 for his winning strategy to shrink the carbon footprint of artificial intelligence models trained to detect heart disease.&nbsp;</p> <p>“I never thought about the kilowatt-hours I was using,” he says. “But this hackathon gave me a chance to look at my carbon footprint and find ways to trade a small amount of model accuracy for big energy savings.”&nbsp;</p> <p>Haft-Javaherian was among six teams to earn prizes at a hackathon co-sponsored by the&nbsp;<a href="">MIT Research Computing Project</a>&nbsp;and&nbsp;<a href="">MIT-IBM Watson AI Lab</a> Jan. 28-30. The event was meant to familiarize students with Satori, the computing cluster IBM&nbsp;<a href="">donated</a> to MIT last year, and to inspire new techniques for building energy-efficient AI models that put less planet-warming carbon dioxide into the air.&nbsp;</p> <p>The event was also a celebration of Satori’s green-computing credentials. With an architecture designed to minimize the transfer of data, among other energy-saving features, Satori recently earned&nbsp;<a href="">fourth place</a>&nbsp;on the Green500 list of supercomputers. Its location gives it additional credibility: It sits on a remediated brownfield site in Holyoke, Massachusetts, now the&nbsp;<a href="">Massachusetts Green High Performance Computing Center</a>, which runs largely on low-carbon hydro, wind and nuclear power.</p> <p>A postdoc at MIT and Harvard Medical School, Haft-Javaherian came to the hackathon to learn more about Satori. He stayed for the challenge of trying to cut the energy intensity of his own work, focused on developing AI methods to screen the coronary arteries for disease. A new imaging method, optical coherence tomography, has given cardiologists a new tool for visualizing defects in the artery walls that can slow the flow of oxygenated blood to the heart. But even the experts can miss subtle patterns that computers excel at detecting.</p> <p>At the hackathon, Haft-Javaherian ran a test on his model and saw that he could cut its energy use eight-fold by reducing the time Satori’s graphics processors sat idle. He also experimented with adjusting the model’s number of layers and features, trading varying degrees of accuracy for lower energy use.&nbsp;</p> <p>A second team, Alex Andonian and Camilo Fosco, also won $1,000 by showing they could train a classification model nearly 10 times faster by optimizing their code and losing a small bit of accuracy. Graduate students in the Department of Electrical Engineering and Computer Science (EECS), Andonian and Fosco are currently training a classifier to tell legitimate videos from AI-manipulated fakes, to compete in Facebook’s&nbsp;<a href="">Deepfake Detection Challenge</a>. Facebook launched the contest last fall to crowdsource ideas for stopping the spread of misinformation on its platform ahead of the 2020 presidential election.</p> <p>If a technical solution to deepfakes is found, it will need to run on millions of machines at once, says Andonian. That makes energy efficiency key. “Every optimization we can find to train and run more efficient models will make a huge difference,” he says.</p> <p>To speed up the training process, they tried streamlining their code and lowering the resolution of their 100,000-video training set by eliminating some frames. They didn’t expect a solution in three days, but Satori’s size worked in their favor. “We were able to run 10 to 20 experiments at a time, which let us iterate on potential ideas and get results quickly,” says Andonian.&nbsp;</p> <p>As AI continues to improve at tasks like reading medical scans and interpreting video, models have grown bigger and more calculation-intensive, and thus, energy intensive. By one&nbsp;<a href="">estimate</a>, training a large language-processing model produces nearly as much carbon dioxide as the cradle-to-grave emissions from five American cars. The footprint of the typical model is modest by comparison, but as AI applications proliferate its environmental impact is growing.&nbsp;</p> <p>One way to green AI, and tame the exponential growth in demand for training AI, is to build smaller models. That’s the approach that a third hackathon competitor, EECS graduate student Jonathan Frankle, took. Frankle is looking for signals early in the training process that point to subnetworks within the larger, fully-trained network that can do the same job.&nbsp;The idea builds on his award-winning&nbsp;<a href="">Lottery Ticket Hypothesis</a>&nbsp;paper from last year that found a neural network could perform with 90 percent fewer connections if the right subnetwork was found early in training.</p> <p>The hackathon competitors were judged by John Cohn, chief scientist at the MIT-IBM Watson AI Lab, Christopher Hill, director of MIT’s Research Computing Project, and Lauren Milechin, a research software engineer at MIT.&nbsp;</p> <p>The judges recognized four&nbsp;other teams: Department of Earth, Atmospheric and Planetary Sciences (EAPS) graduate students Ali Ramadhan,&nbsp;Suyash Bire, and James Schloss,&nbsp;for adapting the programming language Julia for Satori; MIT Lincoln Laboratory postdoc Andrew Kirby, for adapting code he wrote as a graduate student to Satori using a library designed for easy programming of computing architectures; and Department of Brain and Cognitive Sciences graduate students Jenelle Feather and Kelsey Allen, for applying a technique that drastically simplifies models by cutting their number of parameters.</p> <p>IBM developers were on hand to answer questions and gather feedback.&nbsp;&nbsp;“We pushed the system — in a good way,” says Cohn. “In the end, we improved the machine, the documentation, and the tools around it.”&nbsp;</p> <p>Going forward, Satori will be joined in Holyoke by&nbsp;<a href="">TX-Gaia</a>, Lincoln Laboratory’s new supercomputer.&nbsp;Together, they will provide feedback on the energy use of their workloads. “We want to raise awareness and encourage users to find innovative ways to green-up all of their computing,” says Hill.&nbsp;</p> Several dozen students participated in the Green AI Hackathon, co-sponsored by the MIT Research Computing Project and MIT-IBM Watson AI Lab. Photo panel: Samantha SmileyQuest for Intelligence, MIT-IBM Watson AI Lab, Electrical engineering and computer science (EECS), EAPS, Lincoln Laboratory, Brain and cognitive sciences, School of Engineering, School of Science, Algorithms, Artificial intelligence, Computer science and technology, Data, Machine learning, Software, Climate change, Awards, honors and fellowships, Hackathon, Special events and guest speakers Drawing daily doodles: Chalk of the Day brightens MIT Chalk of the Day, an MIT student group, draws beautiful daily works of art on the chalk wall in Building 32. Mon, 10 Feb 2020 16:10:01 -0500 Julia Newman | Division of Student Life <p>The Ray and Maria Stata Center is an architectural staple of MIT’s campus. Inside the angled walls and modern exterior lives the Computer Science and Artificial Intelligence Laboratory (<a href="" target="_blank">CSAIL</a>), the Laboratory for Information and Decision Systems (<a href="" target="_blank">LIDS</a>), and the <a href="" target="_blank">Department of Linguistics and Philosophy</a>. It's also a central hub for conferences, lunch meetings, and regular events like <a href="" target="_blank">Choose to Reuse</a>. Building 32 also houses the canvas used by student group Chalk of the Day to share daily works of art.</p> <p><a href="" target="_blank">Chalk of the Day</a> was started in 2015 by Benjamin Chan ’17 as a way to give back to the MIT community through inspirational messages and doodles. Today, Chalk of the Day remains a tight-knit group of friends who craft new pieces of art that are visible to passers-by for a day, memorialized on the group's <a href="" target="_blank">Instagram account</a> — and then erased every night.</p> <div class="cms-placeholder-content-video"></div> <p>Priscilla Wong, a chalker who finished her coursework in computer science and engineering last fall and will be graduating in May, says she began chalking as a way to find an escape from the typical routine at MIT. Each semester, students schedule and claim a day based on their availability. Wong and her chalking partner Jessica Xu, a junior in mechanical engineering, have chalked before, and last semester, they made sure they shared free Tuesday mornings to continue their tradition of making art together.</p> <p>Using a pointillist technique, Wong taps the chalk repeatedly against the board to create a snow effect as she discusses the ways in which chalk is an unusual medium. “Some of the most difficult things about chalking are also what makes it the most interesting,” she says. “If you chalk over a really big area it ends up snowing down on everything below. Sometimes it’s an effect you want to achieve.” Her chalk partner, Jessica Xu, adds “for the most part we come up with techniques on our own.” Wong echoes how there’s a learning curve and the way they learn is simply by chalking.</p> <p>The works of art span from inspirational quotes to more political works, like a drawing in response to the Australian wildfires: a mama and child koala sit in a tree with the text “save us” above, the letters connecting like a crossword puzzle to configure AUS for Australia. The art is often incredibly detailed: One homage to the film “Up” displayed the characters lifted by a house tied to balloons with the message “adventure is out there,” while another featured a hummingbird eating nectar from a blossoming flower.</p> <p>Sarah Wu, a senior mathematics major, has been chalking since her first year at MIT, when she was looking for more artsy things to do around campus. She compares chalking with solving a math problem: Both require a level of creativity, but approaching a blank canvas is a totally different process and engages a different part of her mind. Chalking is a way to relax and de-stress for Wu: “Normally, I’m always thinking about the next assignment or the next test, but this is an opportunity where once a week I can actually remove myself from that and try to focus only on the art I’m making and the things I’m contributing to the community.” She and her chalking partner, Charleen Wang, a senior in electrical engineering and computer science, worked on a lettering piece that reads “Catch your breath, take your time,” filled with snowflakes mimicking the weather conditions outside.</p> <p>Wang shares a similar sentiment to Wu about the importance of Chalk of the Day in her routine. “I think sometimes I forget to engage in a more creative side of me. I learned a lot about how to put in other priorities that I might be forgetting into my schedule. It’s not all about grades,” she says. She likes how temporary chalk is as a medium. “I feel more free to try different things because it’s not something so permanent like pen or painting.” Every day is an opportunity for chalk artists to try something new, create a new work of art, and feel empowered to think outside of the box.</p> <p>Chalking helps students de-stress, but more than that, their artwork spreads positivity and inspiration to the entire MIT community. Passersby “send it to their boyfriend or girlfriend or friend or mother. I like that it has an impact that is beyond Stata or MIT,” Wong reflects. Chalk of the Day members hope that sharing the daily chalk-works encourages others to be more creative in their everyday lives.</p> Jessica Xu and Priscilla Wong stand in front of their finished chalk art in the Stata Center.Photo: Jeff Saint DicStudent life, Community, Arts, Computer Science and Artificial Intelligence Laboratory (CSAIL), Students, Alumni/ae, Electrical Engineering & Computer Science (eecs), School of Engineering, School of Humanities Arts and Social Sciences, Laboratory for Information and Decision Systems (LIDS), Clubs and activities Hey Alexa! Sorry I fooled you ... MIT’s new system TextFooler can trick the types of natural-language-processing systems that Google uses to help power its search results, including audio for Google Home. Fri, 07 Feb 2020 11:20:01 -0500 Rachel Gordon | MIT CSAIL <p>A human can likely tell the difference between a turtle and a rifle. Two years ago, Google’s AI wasn’t so <a href="">sure</a>. For quite some time, a subset of computer science research has been dedicated to better understanding how machine-learning models handle these “adversarial” attacks, which are inputs deliberately created to trick or fool machine-learning algorithms.&nbsp;</p> <p>While much of this work has focused on <a href="">speech</a> and <a href="">images</a>, recently, a team from MIT’s <a href="">Computer Science and Artificial Intelligence Laboratory</a> (CSAIL) tested the boundaries of text. They came up with “TextFooler,” a general framework that can successfully attack natural language processing (NLP) systems — the types of systems that let us interact with our Siri and Alexa voice assistants — and “fool” them into making the wrong predictions.&nbsp;</p> <p>One could imagine using TextFooler for many applications related to internet safety, such as email spam filtering, hate speech flagging, or “sensitive” political speech text detection — which are all based on text classification models.&nbsp;</p> <p>“If those tools are vulnerable to purposeful adversarial attacking, then the consequences may be disastrous,” says Di Jin, MIT PhD student and lead author on a new paper about TextFooler. “These tools need to have effective defense approaches to protect themselves, and in order to make such a safe defense system, we need to first examine the adversarial methods.”&nbsp;</p> <p>TextFooler works in two parts: altering a given text, and then using that text to test two different language tasks to see if the system can successfully trick machine-learning models.&nbsp;&nbsp;</p> <p>The system first identifies the most important words that will influence the target model’s prediction, and then selects the synonyms that fit contextually. This is all while maintaining grammar and the original meaning to look “human” enough, until the prediction is altered.&nbsp;</p> <p>Then, the framework is applied to two different tasks — text classification, and entailment (which is the relationship between text fragments in a sentence), with the goal of changing the classification or invalidating the entailment judgment of the original models.&nbsp;</p> <p>In one example, TextFooler’s input and output were:</p> <p>“The characters, cast in impossibly contrived situations, are totally estranged from reality.”&nbsp;</p> <p>“The characters, cast in impossibly engineered circumstances, are fully estranged from reality.”&nbsp;</p> <p>In this case, when testing on an NLP model, it gets the example input right, but then gets the modified input wrong.&nbsp;</p> <p>In total, TextFooler successfully attacked three target models, including “BERT,” the popular open-source NLP model. It fooled the target models with an accuracy of over 90 percent to under 20 percent, by changing only 10 percent of the words in a given text. The team evaluated success on three criteria: changing the model's prediction for classification or entailment; whether it looked similar in meaning to a human reader, compared with the original example; and whether the text looked natural enough.&nbsp;</p> <p>The researchers note that while attacking existing models is not the end goal, they hope that this work will help more abstract models generalize to new, unseen data.&nbsp;</p> <p>“The system can be used or extended to attack any classification-based NLP models to test their robustness,” says Jin. “On the other hand, the generated adversaries can be used to improve the robustness and generalization of deep-learning models via adversarial training, which is a critical direction of this work.”&nbsp;</p> <p>Jin wrote the paper alongside MIT Professor Peter Szolovits, Zhijing Jin of the University of Hong Kong, and Joey Tianyi Zhou of A*STAR, Singapore. They will present the paper at the AAAI Conference on Artificial Intelligence in New York.&nbsp;</p> CSAIL PhD student Di Jin led the development of the TextFooler system.Photo: Jason Dorfman/MIT CSAILComputer Science and Artificial Intelligence Laboratory (CSAIL), Computer science and technology, Machine learning, Algorithms, Data, Natural language processing, Artificial intelligence, Electrical Engineering & Computer Science (eecs), School of Engineering, Technology and society Researchers develop a roadmap for growth of new solar cells Starting with higher-value niche markets and then expanding could help perovskite-based solar panels become competitive with silicon. Thu, 06 Feb 2020 10:57:11 -0500 David L. Chandler | MIT News Office <p>Materials called perovskites show strong potential for a new generation of solar cells, but they’ve had trouble gaining traction in a market dominated by silicon-based solar cells. Now, a study by researchers at MIT and elsewhere outlines a roadmap for how this promising technology could move from the laboratory to a significant place in the global solar market.</p> <p>The “technoeconomic” analysis shows that by starting with higher-value niche markets and gradually expanding, solar panel manufacturers could avoid the very steep initial capital costs that would be required to make perovskite-based panels directly competitive with silicon for large utility-scale installations at the outset. Rather than making a prohibitively expensive initial investment, of hundreds of millions or even billions of dollars, to build a plant for utility-scale production, the team found that starting with more specialized applications could be accomplished for more realistic initial capital investment on the order of $40 million.</p> <p>The results are described in a paper in the journal <em>Joule</em> by MIT postdoc Ian Mathews, research scientist Marius Peters, professor of mechanical engineering Tonio Buonassisi, and five others at MIT, Wellesley College, and Swift Solar Inc.</p> <p>Solar cells based on perovskites — a broad category of compounds characterized by a certain arrangement of their molecular structure — could provide dramatic improvements in solar installations. Their constituent materials are inexpensive, and they could be manufactured in a roll-to-roll process like printing a newspaper, and printed onto lightweight and flexible backing material. This could greatly reduce costs associated with transportation and installation, although they still require further work to improve their durability. Other promising new solar cell materials are also under development in labs around the world, but none has yet made inroads in the marketplace.</p> <p>“There have been a lot of new solar cell materials and companies launched over the years,” says Mathews, “and yet, despite that, silicon remains the dominant material in the industry and has been for decades.”</p> <p>Why is that the case? “People have always said that one of the things that holds new technologies back is that the expense of constructing large factories to actually produce these systems at scale is just too much,” he says. “It’s difficult for a startup to cross what’s called ‘the valley of death,’ to raise the tens of millions of dollars required to get to the scale where this technology might be profitable in the wider solar energy industry.”</p> <p>But there are a variety of more specialized solar cell applications where the special qualities of perovskite-based solar cells, such as their light weight, flexibility, and potential for transparency, would provide a significant advantage, Mathews says. By focusing on these markets initially, a startup solar company could build up to scale gradually, leveraging the profits from the premium products to expand its production capabilities over time.</p> <p>Describing the literature on perovskite-based solar cells being developed in various labs, he says, “They’re claiming very low costs. But they’re claiming it once your factory reaches a certain scale. And I thought, we’ve seen this before — people claim a new photovoltaic material is going to be cheaper than all the rest and better than all the rest. That’s great, except we need to have a plan as to how we actually get the material and the technology to scale.”</p> <p>As a starting point, he says, “We took the approach that I haven’t really seen anyone else take: Let’s actually model the cost to manufacture these modules as a function of scale. So if you just have 10 people in a small factory, how much do you need to sell your solar panels at in order to be profitable? And once you reach scale, how cheap will your product become?”</p> <p>The analysis confirmed that trying to leap directly into the marketplace for rooftop solar or utility-scale solar installations would require very large upfront capital investment, he says. But “we looked at the prices people might get in the internet of things, or the market in building-integrated photovoltaics. People usually pay a higher price in these markets because they’re more of a specialized product. They’ll pay a little more if your product is flexible or if the module fits into a building envelope.” Other potential niche markets include self-powered microelectronics devices.</p> <p>Such applications would make the entry into the market feasible without needing massive capital investments. “If you do that, the amount you need to invest in your company is much, much less, on the order of a few million dollars instead of tens or hundreds of millions of dollars, and that allows you to more quickly develop a profitable company,” he says.</p> <p>“It’s a way for them to prove their technology, both technically and by actually building and selling a product and making sure it survives in the field,” Mathews says, “and also, just to prove that you can manufacture at a certain price point.”</p> <p>Already, there are a handful of startup companies working to try to bring perovskite solar cells to market, he points out, although none of them yet has an actual product for sale. The companies have taken different approaches, and some seem to be embarking on the kind of step-by-step growth approach outlined by this research, he says. “Probably the company that’s raised the most money is a company called Oxford PV, and they’re looking at tandem cells,” which incorporate both silicon and perovskite cells to improve overall efficiency. Another company is one started by Joel Jean PhD ’17 (who is also a co-author of this paper) and others, called Swift Solar, which is working on flexible perovskites. And there’s a company called Saule Technologies, working on printable perovskites.</p> <p>Mathews says the kind of technoeconomic analysis the team used in its study could be applied to a wide variety of other new energy-related technologies, including rechargeable batteries and other storage systems, or other types of new solar cell materials.</p> <p>“There are many scientific papers and academic studies that look at how much it will cost to manufacture a technology once it’s at scale,” he says. “But very few people actually look at how much does it cost at very small scale, and what are the factors affecting economies of scale? And I think that can be done for many technologies, and it would help us accelerate how we get innovations from lab to market.”</p> <p>The research team also included MIT alumni Sarah Sofia PhD ’19 and Sin Cheng Siah PhD ’15, Wellesley College student Erica Ma, and former MIT postdoc Hannu Laine. The work was supported by the European Union’s Horizon 2020 research and innovation program, the Martin Family Society for Fellows of Sustainability, the U.S. Department of Energy, Shell, through the MIT Energy Initiative, and the Singapore-MIT Alliance for Research and Technology.</p> Perovskites, a family of materials defined by a particular kind of molecular structure as illustrated here, have great potential for new kinds of solar cells. A new study from MIT shows how these materials could gain a foothold in the solar marketplace.Image: Christine Daniloff, MITResearch, School of Engineering, Energy, Solar, Nanoscience and nanotechnology, Materials Science and Engineering, Mechanical engineering, National Science Foundation (NSF), Renewable energy, Alternative energy, Sustainability, Artificial intelligence, Machine learning, MIT Energy Initiative, Singapore-MIT Alliance for Research and Technology (SMART) A college for the computing age With the initial organizational structure in place, the MIT Schwarzman College of Computing moves forward with implementation. Tue, 04 Feb 2020 12:30:01 -0500 Terri Park | MIT Schwarzman College of Computing <p>The mission of the MIT Stephen A. Schwarzman College of Computing is to address the opportunities and challenges of the computing age — from hardware to software to algorithms to artificial intelligence (AI) — by transforming the capabilities of academia in three key areas: supporting the rapid evolution and growth of computer science and AI; facilitating collaborations between computing and other disciplines; and focusing on social and ethical responsibilities of computing through combining technological approaches and insights from social science and humanities, and through engagement beyond academia.</p> <p>Since starting his position in August 2019, Daniel Huttenlocher, the inaugural dean of the MIT Schwarzman College of Computing, has been working with many stakeholders in designing the initial organizational structure of the college. Beginning with the <a href="" target="_blank">College of Computing Task Force Working Group reports</a> and feedback from the MIT community, the structure has been developed through an iterative process of draft plans yielding a <a href="" target="_blank">26-page document</a> outlining the initial academic organization of the college that is designed to facilitate the college mission through improved coordination and evolution of existing computing programs at MIT, improved collaboration in computing across disciplines, and development of new cross-cutting activities and programs, notably in the social and ethical responsibilities of computing.</p> <p>“The MIT Schwarzman College of Computing is both bringing together existing MIT programs in computing and developing much-needed new cross-cutting educational and research programs,” says Huttenlocher. “For existing programs, the college helps facilitate coordination and manage the growth in areas such as computer science, artificial intelligence, data systems and society, and operations research, as well as helping strengthen interdisciplinary computing programs such as computational science and engineering. For new areas, the college is creating cross-cutting platforms for the study and practice of social and ethical responsibilities of computing, for multi-departmental computing education, and for incubating new interdisciplinary computing activities.”</p> <p>The following existing departments, institutes, labs, and centers are now part of the college:</p> <ul> <li>Department of Electrical Engineering and Computer (EECS), which has been <a href="" target="_self">reorganized</a> into three overlapping sub-units of electrical engineering (EE), computer science (CS), and artificial intelligence and decision-making (AI+D), and is jointly part of the MIT Schwarzman College of Computing and School of Engineering;</li> <li>Operations Research Center (ORC), which is jointly part of the MIT Schwarzman College of Computing and MIT Sloan School of Management;</li> <li>Institute for Data, Systems, and Society (IDSS), which will be increasing its focus on the societal aspects of its mission while also continuing to support statistics across MIT, and including the Technology and Policy Program (TPP) and Sociotechnical Systems Research Center (SSRC);</li> <li>Center for Computational Science Engineering (CCSE), which is being renamed from the Center for Computational Engineering and broadening its focus in the sciences;</li> <li>Computer Science and Artificial Intelligence Laboratory (CSAIL);</li> <li>Laboratory for Information and Decision Systems (LIDS); and</li> <li>Quest for Intelligence.</li> </ul> <p>With the initial structure in place, Huttenlocher, the college leadership team, and the leaders of the academic units that are part of the college, in collaboration with departments in all five schools, are actively moving forward with curricular and programmatic development, including the launch of two new areas, the Common Ground for Computing Education and the Social and Ethical Responsibilities of Computing (SERC). Still in the early planning stages, these programs are the aspects of the college that are designed to cut across lines and involve a number of departments throughout MIT. Other programs are expected to be introduced as the college continues to take shape.</p> <p>“The college is an Institute-wide entity, working with and across all five schools,” says Anantha Chandrakasan, dean of the School of Engineering and the Vannevar Bush Professor of Electrical Engineering and Computer Science, who was part of the task force steering committee. “Its continued growth and focus depend greatly on the input of our MIT community, a process which began over a year ago. I’m delighted that Dean Huttenlocher and the college leadership team have engaged the community for collaboration and discussion around the plans for the college.”</p> <p>With these organizational changes, students, faculty, and staff in these units are members of the college, and in some cases, jointly with a school, as will be those who are engaged in the new cross-cutting activities in SERC and Common Ground. “A question we get frequently,” says Huttenlocher, “is how to apply to the college. As is the case throughout MIT, undergraduate admissions are handled centrally, and graduate admissions are handled by each individual department or graduate program.”<strong> </strong></p> <p><strong>Advancing computing</strong></p> <p>Despite the unprecedented growth in computing, there remains substantial unmet demand for expertise. In academia, colleges and universities worldwide are faced with oversubscribed programs in computer science and the constant need to keep up with rapidly changing materials at both the graduate and undergraduate level.</p> <p>According to Huttenlocher, the computing fields are evolving at a pace today that is beyond the capabilities of current academic structures to handle. “As academics, we pride ourselves on being generators of new knowledge, but academic institutions themselves don’t change that quickly. The rise of AI is probably the biggest recent example of that, along with the fact that about 40 percent of MIT undergraduates are majoring in computer science, where we have 7 percent of the MIT faculty.”</p> <p>In order to help meet this demand, MIT is increasing its academic capacity in computing and AI with 50 new faculty positions — 25 will be core computing positions in CS, AI, and related areas, and 25 will be shared jointly with departments. Searches are now active to recruit core faculty in CS and AI+D, and for joint faculty with MIT Philosophy, the Department of Brain and Cognitive Sciences, and several interdisciplinary institutes.</p> <p>The new shared faculty searches will largely be conducted around the concept of “clusters” to build capacity at MIT in important computing areas that cut across disciplines, departments, and schools. Huttenlocher, the provost, and the five school deans will work to identify themes based on input from departments so that recruiting can be undertaken during the next academic year.</p> <p><strong>Cross-cutting collaborations in computing</strong></p> <p>Building on the history of strong faculty participation in interdepartmental labs, centers, and initiatives, the MIT Schwarzman College of Computing provides several forms of membership in the college based on cross-cutting research, teaching, or external engagement activities. While computing is affecting intellectual inquiry in almost every discipline, Huttenlocher is quick to stress that “it’s bi-directional.” He notes that existing collaborations across various schools and departments, such as MIT Digital Humanities, as well as opportunities for new such collaborations, are key to the college mission because in the same way that “computing is changing thinking in the disciplines; the disciplines are changing the way people do computing.”</p> <p>Under the leadership of Asu Ozdaglar, the deputy dean of academics and department head of EECS, the college is developing the Common Ground for Computing Education, an interdepartmental teaching collaborative that will facilitate the offering of computing classes and coordination of computing-related curricula across academic units.</p> <p>The objectives of this collaborative are to provide opportunities for faculty across departments to work together, including co-teaching classes, creating new undergraduate majors or minors such as in AI+D, as well as facilitating undergraduate blended degrees such as 6-14 (Computer Science, Economics, and Data Science), 6-9 (Computation and Cognition), 11-6 (Urban Science and Planning with Computer Science), 18-C (Mathematics with Computer Science), and others.</p> <p>“It is exciting to bring together different areas of computing with methodological and substantive commonalities as well as differences around one table,” says Ozdaglar. “MIT faculty want to collaborate in topics around computing, but they are increasingly overwhelmed with teaching assignments and other obligations. I think the college will enable the types of interactions that are needed to foster new ideas.”</p> <p>Thinking about the impact on the student experience, Ozdaglar expects that the college will help students better navigate the computing landscape at MIT by creating clearer paths. She also notes that many students have passions beyond computer science, but realize the need to be adept in computing techniques and methodologies in order to pursue other interests, whether it be political science, economics, or urban science. “The idea for the college is to educate students who are fluent in computation, but at the same time, creatively apply computing with the methods and questions of the domain they are mostly interested in.”</p> <p>For Deputy Dean of Research Daniela Rus, who is also the director of CSAIL and the Andrew and Erna Viterbi Professor in EECS, developing research programs “that bring together MIT faculty and students from different units to advance computing and to make the world better through computing” is a top priority. She points to the recent launch of the <a href="" target="_self">MIT Air Force AI Innovation Accelerator</a>, a collaboration between the MIT Schwarzman College of Computing and the U.S. Air Force focused on AI, as an example of the types of research projects the college can facilitate.</p> <p>“As humanity works to solve problems ranging from climate change to curing disease, removing inequality, ensuring sustainability, and eliminating poverty, computing opens the door to powerful new solutions,” says Rus. “And with the MIT Schwarzman College as our foundation, I believe MIT will be at the forefront of those solutions. Our scholars are laying theoretical foundations of computing and applying those foundations to big ideas in computing and across disciplines.”</p> <p><strong>Habits of mind and action</strong></p> <p>A critically important cross-cutting area is the Social and Ethical Responsibilities of Computing, which will facilitate the development of responsible “habits of mind and action” for those who create and deploy computing technologies, and the creation of technologies in the public interest.</p> <p>“The launch of the MIT Schwarzman College of Computing offers an extraordinary new opportunity for the MIT community to respond to today’s most consequential questions in ways that serve the common good,” says Melissa Nobles, professor of political science, the Kenan Sahin Dean of the MIT School of Humanities, Arts, and Social Sciences, and co-chair of the Task Force Working Group on Social Implications and Responsibilities of Computing.</p> <p>“As AI and other advanced technologies become ubiquitous in their influence and impact, touching nearly every aspect of life, we have increasingly seen the need to more consciously align powerful new technologies with core human values — integrating consideration of societal and ethical implications of new technologies into the earliest stages of their development. Asking, for example, of every new technology and tool: Who will benefit? What are the potential ecological and social costs? Will the new technology amplify or diminish human accomplishments in the realms of justice, democracy, and personal privacy?</p> <p>“As we shape the college, we are envisioning an MIT culture in which all of us are equipped and encouraged to think about such implications. In that endeavor, MIT’s humanistic disciplines will serve as deep resources for research, insight, and discernment. We also see an opportunity for advanced technologies to help solve political, economic, and social issues that trouble today’s world by integrating technology with a humanistic analysis of complex civilizational issues — among them climate change, the future of work, and poverty, issues that will yield only to collaborative problem-solving. It is not too much to say that human survival may rest on our ability to solve these problems via collective intelligence, designing approaches that call on the whole range of human knowledge.”</p> <p>Julie Shah, an associate professor in the Department of Aeronautics and Astronautics and head of the Interactive Robotics Group at CSAIL, who co-chaired the working group with Nobles and is now a member of the college leadership, adds that “traditional technologists aren’t trained to pause and envision the possible futures of how technology can and will be used. This means that we need to develop new ways of training our students and ourselves in forming new habits of mind and action so that we include these possible futures into our design.”</p> <p>The associate deans of Social and Ethical Responsibilities of Computing, Shah and David Kaiser, the Germeshausen Professor of the History of Science and professor of physics, are designing a systemic framework for SERC that will not only effect change in computing education and research at MIT, but one that will also inform policy and practice in government and industry. Activities that are currently in development include multi-disciplinary curricula embedded in traditional computing and AI courses across all levels of instruction, the commission and curation of a series of case studies that will be modular and available to all via MIT’s open access channels, active learning projects, cross-disciplinary monthly convenings, public forums, and more.&nbsp;</p> <p>“A lot of how we’ve been thinking about SERC components is building capacity with what we already have at the Institute as a very important first step. And that means how do we get people interacting in ways that can be a little bit different than what has been familiar, because I think there are a lot of shared goals among the MIT community, but the gears aren’t quite meshing yet. We want to further support collaborations that might cut across lines that otherwise might not have had much traffic between them,” notes Kaiser.</p> <p><strong>Just the beginning</strong></p> <p>While he’s excited by the progress made so far, Huttenlocher points out there will continue to be revisions made to the organizational structure of the college. “We are at the very beginning of the college, with a tremendous amount of excellence at MIT to build on, and with some clear needs and opportunities, but the landscape is changing rapidly and the college is very much a work in progress.”</p> <p>The college has other initiatives in the planning stages, such as the Center for Advanced Studies of Computing that will host fellows from inside and outside of MIT on semester- or year-long project-oriented programs in focused topic areas that could seed new research, scholarly, educational, or policy work. In addition, Huttenlocher is planning to launch a search for an assistant or associate dean of equity and inclusion, once the Institute Community and Equity Officer is in place, to focus on improving and creating programs and activities that will help broaden participation in computing classes and degree programs, increase the&nbsp;diversity&nbsp;of top faculty candidates in computing fields, and ensure that faculty search and graduate admissions processes have diverse slates of candidates and interviews.</p> <p>“The typical academic approach would be to wait until it’s clear what to do, but that would be a mistake. The way we’re going to learn is by trying and by being more flexible. That may be a more general attribute of the new era we’re living in, he says. “We don’t know what it’s going to look like years from now, but it’s going to be pretty different, and MIT is going to be shaping it.”</p> <p>The MIT Schwarzman College of Computing will be hosting a community forum on Wednesday, Feb. 12 at 2 p.m. in Room 10-250. Members from the MIT community are welcome to attend to learn more about the initial organizational structure of the college.</p> MIT Schwarzman College of Computing leadership team (left to right) David Kaiser, Daniela Rus, Dan Huttenlocher, Julie Shah, and Asu Ozdaglar Photo: Sarah BastilleMIT Schwarzman College of Computing, School of Engineering, Computer Science and Artificial Intelligence Laboratory (CSAIL), Laboratory for Information and Decision Systems (LIDS), Quest for Intelligence, Philosophy, Brain and cognitive sciences, Digital humanities, School of Humanities Arts and Social Sciences, Artificial intelligence, Operations research, Aeronautical and astronautical engineering, Electrical Engineering & Computer Science (eecs), IDSS, Ethics, Administration, Classes and programs A smart surface for smart devices External system improves phones’ signal strength 1,000 percent, without requiring extra antennas. Mon, 03 Feb 2020 10:30:01 -0500 Adam Conner-Simons | CSAIL <p>We’ve heard it for years: 5G is coming.&nbsp;</p> <p>And yet, while high-speed 5G internet has indeed slowly been rolling out in a smattering of countries across the globe, many barriers remain that have prevented widespread adoption.</p> <p>One issue is that we can’t get faster internet speeds without more efficient ways of delivering wireless signals. The general trend has been to simply add antennas to either the transmitter (i.e., Wi-Fi access points and cell towers) or the receiver (such as a phone or laptop). But that’s grown difficult to do as companies increasingly produce smaller and smaller devices, including a new wave of “internet of things” systems.</p> <p>Researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL)&nbsp;looked at the problem recently and wondered if people&nbsp;have had things completely backwards this whole time. Rather than focusing on the transmitters and receivers, what if we could amplify the signal by adding antennas to an external surface in the environment itself?</p> <p>That’s the idea behind the CSAIL team's new system RFocus, a software-controlled “smart surface” that uses more than 3,000 antennas to maximize the strength of the signal at the receiver. Tests showed that RFocus could improve the average signal strength by a factor of almost 10. Practically speaking, the platform is also very cost-effective, with each antenna costing only a few cents. The antennas are inexpensive because they don’t process the signal at all; they merely control how it is reflected. Lead author Venkat Arun says that the project represents what is, to the team’s knowledge, the largest number of antennas ever used for a single communication link.</p> <p>While the system could serve as another form of WiFi range extender, the researchers say&nbsp;its most valuable use could be in the network-connected homes and factories of the future.&nbsp;</p> <p>For example, imagine a warehouse with hundreds of sensors for monitoring machines and inventory. MIT Professor Hari Balakrishnan says that systems for that type of scale would normally be prohibitively expensive and/or power-intensive, but could be possible with a low-power interconnected system that uses an approach like RFocus.</p> <p>“The core goal here was to explore whether we can use elements in the environment and arrange them to direct the signal in a way that we can actually control,” says Balakrishnan, senior author on a new paper about RFocus that will be presented next month at the USENIX Symposium on Networked Systems Design and Implementation (NSDI) in Santa Clara, California. “If you want to have wireless devices that transmit at the lowest possible power, but give you a good signal, this seems to be one extremely promising way to do it.”</p> <p>RFocus is a two-dimensional surface composed of thousands of antennas that can each either let the signal through or reflect it. The state of the elements is set by a software controller that the team developed with the goal of maximizing the signal strength at a receiver.<em>&nbsp;</em></p> <p>“The biggest challenge was determining how to configure the antennas to maximize signal strength without using any additional sensors, since the signals we measure are very weak,”&nbsp; says PhD student Venkat Arun, lead author of the new paper alongside Balakrishnan. “We ended up with a technique that is surprisingly robust.”</p> <p>The&nbsp;researchers aren’t the first to explore the possibility of improving internet speeds using the external environment. A team at Princeton University led by <a href="">Professor Kyle Jamieson</a> proposed a similar scheme for the specific situation of people using computers on either side of a wall. Balakrishnan says that the goal with RFocus was to develop an even more low-cost approach that could be used in a wider range of scenarios.&nbsp;</p> <p>“Smart surfaces give us literally thousands of antennas to play around with,” says Jamieson, who was not involved in the RFocus project. “The best way of controlling all these antennas, and navigating the massive search space that results when you imagine all the possible antenna configurations, are just two really challenging open problems.”</p> Venkat Arun of MIT stands in front of the prototype of RFocus, a software-controlled “smart surface” that uses more than 3,000 antennas to maximize the strength of the signal at the receiver.Photo: Jason Dorfman/CSAILComputer Science and Artificial Intelligence Laboratory (CSAIL), Electrical engineering and computer science (EECS), Research, School of Engineering, Wireless, internet of things, Data, Mobile devices, Internet, Networks Giving cryptocurrency users more bang for their buck Routing scheme boosts efficiency in networks that help speed up blockchain transactions. Thu, 30 Jan 2020 13:43:32 -0500 Rob Matheson | MIT News Office <p>A new cryptocurrency-routing scheme co-invented by MIT researchers can boost the efficiency — and, ultimately, profits — of certain networks designed to speed up notoriously slow blockchain transactions.&nbsp;&nbsp;</p> <p>Cryptocurrencies hold promise for peer-to-peer financial transactions, potentially making banks and credit cards obsolete. But there’s a scalability issue: Bitcoin, for instance, processes only a handful of transactions per second, while major credit cards process hundreds or thousands. That’s because the blockchain — the digital ledger cryptocurrencies are built on — takes a really long time to process transactions.&nbsp;</p> <p>A new solution is “payment channel networks” (PCNs), where transactions are completed with minimal involvement from the blockchain. Pairs of PCN users form off-blockchain escrow accounts with a dedicated amount of money, forming a large, interconnected network of joint accounts. Users route payments through these &nbsp;accounts, only pinging the blockchain to establish and close the accounts, which speeds things up dramatically. Accounts can also collect a tiny fee when transactions get routed through them.</p> <p>Inefficient routing schemes, however, slow down even these fast solutions. They deplete users’ balances in these accounts frequently, forcing them to invest a lot of money in each account or frequently rebalance their accounts on the blockchain. In a paper being presented next month at the USENIX Symposium on Networked Systems Design and Implementation, the researchers introduce “Spider,” a more efficient routing scheme that lets users invest only a fraction of funds in each account and process roughly four times more transactions before rebalancing on the blockchain.</p> <p>“It’s important to have balanced, high-throughput routing in PCNs to ensure the money that users put into joint accounts is used efficiently,” says first author Vibhaalakshmi Sivaraman, a graduate student in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “This should be efficient and a lucrative business. That means routing as many transactions as possible, with as little funds as possible, to give PCNs the best bang for their buck.”</p> <p>Joining Sivaraman on the paper are former postdoc Shaileshh Bojja Venkatakrishnan, CSAIL graduate students Parimarjan Negi and Lei Yang, and Mohammad Alizadeh, an associate professor of electrical engineering and computer science and a CSAIL researcher; Radhika Mittal of the University of Illinois at Urbana-Champaign; and Kathleen Ruan and Giulia Fanti of Carnegie Mellon University.</p> <p><strong>Packet payments</strong></p> <p>PCNs rely heavily on bidirectional joint accounts — where both parties can receive and send money — so money can be routed between any users. User B can have a joint account with user A, while also linking separately to user C. Users A and C are not directly connected, but user A can send money to user C via the A-B and B-C joint accounts.</p> <p>To exchange funds, each party must approve and update the balances in their joint accounts. Payments can only be routed on channels with sufficient funds to handle the transactions, causing major issues.</p> <p>Traditional schemes send transactions along the shortest path possible, without being aware of any given user’s balance or the rate of sending on that account. This can cause one of the users in the joint account to handle too many transactions and drop to a zero balance, making it unable to route further transactions. What’s more, users can only send a payment in full. If a user wants to send, say, 10 bitcoins, current schemes try to push the full amount on the shortest path possible. If that path can’t support all 10 bitcoins at once, they’ll search for the next shortest path, and so on — which can slow down or completely fail the transaction.</p> <p>Inspired by a technique for internet communications called packet switching, Spider splits each full transaction into smaller “packets” that are sent across different channels at different rates. This lets the scheme route chunks of these large payments through potentially low-funded accounts. Each packet is then far more likely to reach its destination without slowing down the network or being rejected in any given account for its size.</p> <p>“Shortest-path routing can cause imbalances between accounts that deplete key payment channels and paralyze the system,” Sivaraman says. “Routing money in a way that the funds of both users in each joint account are balanced allows us to reuse the same initial funds to support as many transactions as possible.”</p> <p><br /> <strong>All queued up</strong></p> <p>Another innovation was creating queues at congested accounts. If an account can’t handle incoming transactions that require it to send money, instead of rejecting them, it queues them up. Then, it waits for any transactions that will replenish its funds — within a reasonable time frame — to be able to process those transactions.</p> <p>“If you’re waiting on a queue, but I send you funds within the next second, you can then use any of those funds to send your waiting transactions,” Sivaraman says.</p> <p>The researchers also adopted an algorithm —&nbsp;built by Alizadeh and other researchers&nbsp;— that monitors data center congestion to identify queueing delays at congested accounts. This helps control the rate of transactions. Say user A sends funds to user C through user B, which has a long queue. The receiver C sends the sender A, along with the payment confirmation, one bit of information representing the transaction’s wait time at user B. If it’s too long, user A routes fewer transactions through user B. As the queueing time decreases, account A routes more transactions through B. In this manner, by monitoring the queues alone, Spider is able to ensure that the rate of transactions is both balanced and as high as possible.</p> <p>Ultimately, the more balanced the routing of PCNs, the smaller the capacity required — meaning, overall funds across all joint accounts — for high transaction throughput. In PCN simulations, Spider processed 95 percent of all transactions using only 25 percent of the capacity needed in traditional schemes.</p> <p>The researchers also ran tests on tricky transactions called “DAGs,” which are one-directional payments where one user inevitably runs out of funds and needs to rebalance on the blockchain. A key metric for the performance of PCNs on DAG transactions is the number of off-chain transactions enabled for each transaction on the blockchain. In this regard, Spider is able to process eight times as many off-chain transactions for each transaction on-chain. In contrast, traditional schemes only support twice as many off-chain transactions.</p> <p>“Even with extremely frequent rebalancing, traditional schemes can’t process all DAG transactions. But with very low-frequency rebalancing, Spider can complete them all,” Sivaraman says.</p> <p>Next, the researchers are making Spider more robust to DAG transactions, which can cause bottlenecks. They’re also exploring data privacy issues and ways to incentivize users to use Spider.</p> Spider, a new cryptocurrency-routing scheme, splits each full transaction into smaller “packets” that are sent across different channels at different rates.Image: Chelsea Turner, MITResearch, Computer science and technology, Algorithms, Cyber security, Technology and society, Networks, Finance, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering At halfway point, SuperUROP scholars share their research results In a lively poster session, more than 100 undergraduates discuss their yearlong research projects on everything from machine learning to political geography. Wed, 29 Jan 2020 14:25:01 -0500 Kathryn O'Neill | Department of Electrical Engineering and Computer Science <p>MIT undergraduates are rolling up their sleeves to address major problems in the world, conducting research on topics ranging from nursing care to money laundering to the spread of misinformation about climate change — work highlighted at the most recent SuperUROP Showcase.</p> <p>The event, which took place on the Charles M. Vest Student Street in the Stata Center in December 2019, marked the halfway point in the Advanced Undergraduate Research Opportunities Program (better known as “SuperUROP”). The yearlong program gives MIT students firsthand experience in conducting research with close faculty mentorship. Many participants receive scholar titles recognizing the program’s industry sponsors, individual donors, and other contributors.</p> <p>This year, 102 students participated in SuperUROP, with many of their projects focused on applying computer science technologies, such as machine learning, to challenges in fields ranging from robotics to health care. Almost all presented posters of their work at the December showcase, explaining research to fellow students, faculty members, alumni, sponsors, and other guests.</p> <p>“Every year, this program gets more and more impressive,” says Anantha P. Chandrakasan, dean of the School of Engineering and Vannevar Bush Professor of Electrical Engineering and Computer Science. “What’s especially noteworthy is the incredible breadth of projects and how articulate students are in talking about their work. Their presentation skills seem pretty remarkable.”</p> <p>SuperUROP, administered by the Department of Electrical Engineering and Computer Science (EECS), includes a two-term course, 6.UAR (Undergraduate Advanced Research), designed to teach students research skills, including how to design an experiment and communicate results.</p> <p>“What’s different about SuperUROP [compared to other research opportunities offered to undergraduates] is the companion class that guides you through the necessary writing and speaking,” says Anis Ehsani, a senior majoring in EECS and mathematics, whose project centered on the geometry of drawing political districts. “If I want to pursue a research career, it’s nice to have those skills,” adds Ehsani, an MIT EECS/Nutanix SuperUROP scholar.</p> <p><strong>Beyond the lab and classroom</strong></p> <p>Participants present their work at showcases in the fall and spring, and they are expected to produce prototypes or publication-worthy results by the end of the year.</p> <p>“All these presentations help keep us on track with our projects,” says Weitung Chen, an EECS junior whose project focuses on automating excavation for mining applications. He explains that the inspiration for his SuperUROP work was a real-world problem he faced when trying to build a startup in automated food preparation. Scooping tofu, it turns out, is surprisingly difficult to automate. At the showcase, Chen — an MIT EECS/Angle SuperUROP scholar — explained that he is trying to create a simulation than can be used to train machines to scoop materials autonomously. “I feel really accomplished having this poster and presentation,” he said.</p> <p>Launched by EECS in 2012, SuperUROP has expanded across the Institute over the past several years.</p> <p>Adam Berinsky, the Mitsui Professor of Political Science, is working with SuperUROP students for the first time this year, an experience he’s enjoying. “What’s really cool is being able to give undergraduates firsthand experience in real research,” he says. He’s been able to tap students for the computer science skills he needs for his work, while providing them with a deep dive into the social sciences.</p> <p>Madeline Abrahams, an MIT/Tang Family FinTech SuperUROP scholar, says she especially appreciates the program’s flexibility: “I could explore my interdisciplinary interests,” she says. A computer science and engineering major who is also passionate about political science, Abrahams is working with Berinsky to investigate the spread of misinformation related to climate change via algorithmic aggregation platforms.</p> <p>Nicholas Bonaker also enjoyed the freedom of pursuing his SuperUROP project. “I’ve been able to take the research in the direction I want,” says Bonaker, a junior in EECS, who has developed a new algorithm he hopes will improve an assistive technology developed by his advisor, EECS Associate Professor Tamara Broderick.</p> <p><strong>Exploring new directions in health care</strong></p> <p>Bonaker said he particularly values the health-care focus of his project, which centers on creating better communications software for people living with severe motor impairments. “It feels like I’m doing something that can help people — using things I learned in class,” says Bonaker. He is among this year’s MIT EECS/CS+HASS SuperUROP scholars, whose projects combine computer science with the humanities, arts, or social sciences. &nbsp;</p> <p>Many of this year’s SuperUROP students are working on health-care applications. For example, Fatima Gunter-Rahman, a junior in EECS and biology, is examining Alzheimer’s data, and Sabrina Liu, an EECS junior and MIT EECS/Takeda SUperUROP scholar, is investigating noninvasive ways to monitor the heartrates of dental patients. Justin Lim, a senior math major, is using data analytics to try to determine the optimal treatment for chronic diseases like diabetes. “I like the feeling that my work would have real-world impact,” says Lim, an MIT EECS/Hewlett Foundation SuperUROP scholar. “It’s been very satisfying.”</p> <p>Dhamanpreet Kaur, a junior majoring in math and computer science and molecular biology, is using machine learning to determine the characteristics of patients who are readmitted to hospitals following their discharge to skilled nursing facilities. The work aims to predict who might benefit most from expensive telehealth systems that enable clinicians to monitor patients remotely. The project has given Kaur the chance to work with a multidisciplinary team of professors and doctors. “I find that aspect fascinating,” says Kaur, also an MIT EECS/Takeda SuperUROP scholar.</p> <p>As attendees bustled through the two-hour December showcase, some of the most enthusiastic visitors were industry sponsors, including Larry Bair ’84, SM ’86, a director at Advanced Micro Devices. “I’m always amazed at what undergraduates are doing,” he says, noting that his company has been sponsoring SuperUROPs for the last few years.</p> <p>“It’s always interesting to see what’s going on at MIT,” says Tom O’Dwyer, an MIT research affiliate and the former director of technology at Analog Devices, another industry sponsor. O’Dwyer notes that supporting SuperUROP can help companies with recruitment. “The whole high-tech business runs on smart people,” he says. “SuperUROPs can lead to internships and employment.”</p> <p>SuperUROP also exposes students to the work of academia, which can underscore a key difference between classwork and research: Research results are unpredictable.</p> <p>Junior math major Lior Hirschfeld, for example, compared the effectiveness of different machine learning methods used to test molecules for potential in drug development. “None of them performed exceptionally well,” he says.</p> <p>That might appear to be a poor result, but Hirschfeld notes that it’s important information for those who are using and trusting those tests today. “It shows you may not always know where you are going when you start a project,” says Hirschfeld, also an MIT EECS/Takeda SuperUROP scholar.</p> <p>EECS senior Kenneth Acquah had a similar experience with his SuperUROP project, which focuses on finding a technological way to combat money laundering with Bitcoin. “We’ve tried a bunch of things but mostly found out what doesn’t work,” he says.</p> <p>Still, Acquah says, he values the SuperUROP experience, including the chance to work in MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL). "I get a lot more supervision, more one-on-one time with my mentor," the MIT/EECS Tang Family FinTech SuperUROP scholar says. "And working in CSAIL has given me access to state-of-the-art materials."</p> Madeline Abrahams, an EECS senior and MIT/Tang Family FinTech SuperUROP scholar, presents her work investigating the spread of misinformation related to climate change via algorithmic aggregation platforms at the SuperUROP Showcase. Photo: Gretchen ErtlElectrical engineering and computer science (EECS), School of Engineering, SuperUROP, Political science, School of Humanities Arts and Social Sciences, Computer Science and Artificial Intelligence Laboratory (CSAIL), Aeronautical and astronautical engineering, Chemical engineering, Civil and environmental engineering, Urban studies and planning, School of Architecture and Planning, Students, Research, Undergraduate, Classes and programs, Special events and guest speakers Engineers design bionic “heart” for testing prosthetic valves, other cardiac devices Device made of heart tissue and a robotic pumping system beats like the real thing. Wed, 29 Jan 2020 14:00:00 -0500 Jennifer Chu | MIT News Office <p>As the geriatric population is expected to balloon in the coming decade, so too will rates of heart disease in the United States. The demand for prosthetic heart valves and other cardiac devices — a market that is valued at more than $5 billion dollars today — is predicted to rise by almost 13 percent in the next six years.</p> <p>Prosthetic valves are designed to mimic a real, healthy heart valve in helping to circulate blood through the body. However, many of them have issues such as leakage around the valve, and engineers working to improve these designs must test them repeatedly, first in simple benchtop simulators, then in animal subjects, before reaching human trials — an arduous and expensive process.</p> <p>Now engineers at MIT and elsewhere have developed a bionic “heart” that offers a more realistic model for testing out artificial valves and other cardiac devices.</p> <p>The device is a real biological heart whose tough muscle tissue has been replaced with a soft robotic matrix of artificial heart muscles, resembling bubble wrap. The orientation of the artificial muscles mimics the pattern of the heart’s natural muscle fibers, in such a way that when the researchers remotely inflate the bubbles, they act together to squeeze and twist the inner heart, similar to the way a real, whole heart beats and pumps blood.</p> <p>With this new design, which they call a “biorobotic hybrid heart,” the researchers envision that device designers and engineers could iterate and fine-tune designs more quickly by testing on the biohybrid heart, significantly reducing the cost of cardiac device development.</p> <p>“Regulatory testing of cardiac devices requires many fatigue tests and animal tests,” says Ellen Roche, assistant professor of mechanical engineering at MIT. “[The new device] could realistically represent what happens in a real heart, to reduce the amount of animal testing or iterate the design more quickly.”</p> <p>Roche and her colleagues have published their results today in the journal <em>Science Robotics.</em> Her co-authors are lead author and MIT graduate student Clara Park, along with Yiling Fan, Gregor Hager, Hyunwoo Yuk, Manisha Singh, Allison Rojas, and Xuanhe Zhao at MIT, along with collaborators from Nanyang Technology University, the Royal College of Surgeons in Dublin, Boston’s Children’s Hospital, Harvard Medical School, and Massachusetts General Hospital.</p> <p><img alt="" src="/sites/" style="width: 500px; height: 615px;" /></p> <p><em><span style="font-size:10px;">The structure of the biorobotic hybrid heart under magnetic resonance imaging. Credit: Christopher T. Nguyen</span></em></p> <p><strong>“Mechanics of the heart”</strong></p> <p>Before coming to MIT, Roche worked briefly in the biomedical industry, helping to test cardiac devices on artificial heart models in the lab.</p> <p>“At the time I didn’t feel any of these benchtop setups were representative of both the anatomy and the physiological biomechanics of the heart,” Roche recalls. “There was an unmet need in terms of device testing.”</p> <p>In separate research as part of her doctoral work at Harvard University, she developed a soft, robotic, implantable sleeve, designed to wrap around a whole, live heart, to help it pump blood in patients suffering from heart failure.</p> <p>At MIT, she and Park wondered if they could combine the two research avenues, to develop a hybrid heart: a heart that is made partly of chemically preserved, explanted heart tissue and partly of soft artificial actuators that help the heart pump blood. Such a model, they proposed, should be a more realistic and durable environment in which to test cardiac devices, compared with models that are either entirely artificial but do not capture the heart’s complex anatomy, or are made from a real explanted heart, requiring highly controlled conditions to keep the tissue alive.</p> <p>The team briefly considered wrapping a whole, explanted heart in a soft robotic sleeve, similar to Roche’s previous work, but realized the heart’s outer muscle tissue, the myocardium, quickly stiffened when removed from the body. Any robotic contraction by the sleeve would fail to translate sufficiently to the heart within.</p> <p>Instead, the team looked for ways to design a soft robotic matrix to replace the heart’s natural muscle tissue, in both material and function. They decided to try out their idea first on the heart’s left ventricle, one of four chambers in the heart, which pumps blood to the rest of the body, while the right ventricle uses less force to pump blood to the lungs.</p> <p>“The left ventricle is the harder one to recreate given its higher operating pressures, and we like to start with the hard challenges,” Roche says.</p> <p><strong>The heart, unfurled</strong></p> <p>The heart normally pumps blood by squeezing and twisting, a complex combination of motions that is a result of the alignment of muscle fibers along the outer myocardium that covers each of the heart’s ventricles. The team planned to fabricate a matrix of artificial muscles resembling inflatable bubbles, aligned in the orientations of the natural cardiac muscle. But copying these patterns by studying a ventricle’s three-dimensional geometry proved extremely challenging.</p> <p>They eventually came across the helical ventricular myocardial band theory, the idea that cardiac muscle is essentially a large helical band that wraps around each of the heart’s ventricles. This theory is still a subject of debate by some researchers, but Roche and her colleagues took it as inspiration for their design. Instead of trying to copy the left ventricle’s muscle fiber orientation from a 3D perspective, the team decided to remove the ventricle’s outer muscle tissue and unwrap it to form a long, flat band — a geometry that should be far easier to recreate. In this case, they used the cardiac tissue from an explanted pig heart.</p> <p>In collaboration with co-lead author Chris Nguyen at MGH, the researchers used diffusion tensor imaging, an advanced technique that typically tracks how water flows through white matter in the brain, to map the microscopic fiber orientations of a left ventricle’s unfurled, two-dimensional muscle band. They then fabricated a matrix of artificial muscle fibers made from thin air tubes, each connected to a series of inflatable pockets, or bubbles, the orientation of which they patterned after the imaged muscle fibers.</p> <p><img alt="" src="/sites/" style="width: 500px; height: 380px;" /></p> <p><em><span style="font-size:10px;">Motion of the biorobotic hybrid heart mimics the pumping motion of the heart under echocardiography. Credit: Mossab Saeed</span></em></p> <p>The soft matrix consists of two layers of silicone, with a water-soluble layer between them to prevent the layers from sticking, as well as two layers of laser-cut paper, which ensures that the bubbles inflate in a specific orientation.</p> <p>The researchers also developed a new type of bioadhesive to glue the bubble wrap to the ventricle’s real, intracardiac tissue. While adhesives exist for bonding biological tissues to each other, and and for materials like silicone to each other, the team &nbsp;realized few soft adhesives do an adequate job of gluing together biological tissue with synthetic materials, silicone in particular.</p> <p>So Roche collaborated with Zhao, associate professor of mechanical engineering at MIT, who specializes in developing hydrogel-based adhesives. The new adhesive, named TissueSil, was made by functionalizing silicone in a chemical cross-linking process, to bond with components in heart tissue. The result was a viscous liquid that the researchers brushed onto the soft robotic matrix. They also brushed the glue onto a new explanted pig heart that had its left ventricle removed but its endocardial structures preserved. When they wrapped the artificial muscle matrix around this tissue, the two bonded tightly.</p> <p>Finally, the researchers placed the entire hybrid heart in a mold that they had previously cast of the original, whole heart, and filled the mold with silicone to encase the hybrid heart in a uniform covering — a step that produced a form similar to a real heart and ensured that the robotic bubble wrap fit snugly around the real ventricle.</p> <p>“That way, you don’t lose transmission of motion from the synthetic muscle to the biological tissue,” Roche says.</p> <p>When the researchers pumped air into the bubble wrap at frequencies resembling a naturally beating heart, and imaged the bionic heart’s response, it contracted in a manner similar to the way a real heart moves to pump blood through the body.</p> <p>Ultimately, the researchers hope to use the bionic heart as a realistic environment to help designers test cardiac devices such as prosthetic heart valves.</p> <p>“Imagine that a patient before cardiac device implantation could have their heart scanned, and then clinicians could tune the device to perform optimally in the patient well before the surgery,” says Nyugen. “Also, with further tissue engineering, we could potentially see the biorobotic hybrid heart be used as an artificial heart — a very needed potential solution given the global heart failure epidemic where millions of people are at the mercy of a competitive heart transplant list.”</p> <p>This research was supported in part by the National Science Foundation.</p> A preserved heart muscle (1) is removed and replaced with a soft synthetic matrix (2). The two structures (inner cardiac tissue and synthetic matrix) (3) are bonded using a newly developed adhesive, TissueSil (4). The resulting piece is the biohybrid heart containing the preserved intracardiac structures and synthetic heart muscle (5).Image: Clara ParkBioengineering and biotechnology, Medicine, Health sciences and technology, Civil and environmental engineering, Institute for Medical Engineering and Science (IMES), Mechanical engineering, Research, Robotics, School of Engineering, School of Science, National Science Foundation (NSF) Demystifying artificial intelligence Doctoral candidate Natalie Lao wants to show that anyone can learn to use AI to make a better world. Wed, 29 Jan 2020 13:55:01 -0500 Kim Martineau | MIT Quest for Intelligence <p><a href="">Natalie Lao</a>&nbsp;was set on becoming an electrical engineer, like her parents, until she stumbled on course 6.S192 (<a href="">Making Mobile Apps</a>), taught by Professor <a href="">Hal Abelson</a>. Here was a blueprint for turning a smartphone into a tool for finding clean drinking water, or sorting pictures of faces, or doing just about anything. “I thought, I wish people knew building tech could be like this,” she said on a recent afternoon, taking a break from writing her dissertation.</p> <p>After shifting her focus as an MIT undergraduate&nbsp;to computer science, Lao joined Abelson’s lab, which was busy spreading its&nbsp;<a href="">App Inventor</a>&nbsp;platform and do-it-yourself philosophy to high school students around the world. App Inventor set Lao on her path to making it easy for anyone, from farmers to factory workers, to understand AI, and use it to improve their lives. Now in the third and final year of her PhD at MIT, Lao is also the co-founder of an AI startup to fight fake news, and the co-producer of a series of machine learning tutorials. It’s all part of her mission to help people find the creator and free thinker within.&nbsp;</p> <p>“She just radiates optimism and enthusiasm,” says Abelson, the Class of 1922 Professor in the Department of Electrical Engineering and Computer Science (EECS). “She’s a natural leader who knows how to get people excited and organized.”&nbsp;</p> <p>Lao was immersed in App Inventor, building modules to teach students to build face recognition models and store data in the cloud. Then, in 2016, the surprise election of Donald Trump to U.S. president forced her to think more critically about technology. She was less upset by Trump the politician as by revelations that social media-fueled propaganda and misinformation had tilted the race in Trump’s favor.</p> <p>When a friend, Elan Pavlov, a then-EECS postdoc, approached Lao about an idea he had for building a platform to combat fake news she was ready to dive in. Having grown up in rural, urban, and suburban parts of Tennessee and Ohio, Lao was used to hearing a range of political views. But now, social platforms were filtering those voices, and amplifying polarizing, often inaccurate, content. Pavlov’s idea stood out for its focus on identifying the people (and bots) spreading misinformation and disinformation, rather than the content itself.&nbsp;</p> <p>Lao recruited two friends,&nbsp;<a href="">Andrew Tsai</a>&nbsp;and&nbsp;<a href="">Keertan Kini</a>, to help build out the platform. They would later name it&nbsp;<a href="">HINTS</a>, or Human Interaction News Trustworthiness System, after an early page-ranking algorithm called HITS.&nbsp;</p> <p>In a demo last fall, Lao and Tsai highlighted a network of Twitter accounts that had shared conspiracy theories tied to the murder of Saudi journalist Jamal Khashoggi under the hashtag #khashoggi. When they looked at what else those accounts had shared, they found streams of other false and misleading news. Topping the list was the incorrect claim that then-U.S. Congressman Beto O’Rourke had funded a caravan of migrants headed for the U.S. border.</p> <p>The HINTS team hopes that by flagging the networks that spread fake news, social platforms will move faster to remove fake accounts and contain the propagation of misinformation.</p> <p>“Fake news doesn’t have any impact in a vacuum — real people have to read it and share it,” says Lao. “No matter what your political views, we’re concerned about facts and democracy. There’s fake news being pushed on both sides and it’s making the political divide even worse.”</p> <p>The HINTS team is now working with its first client, a media analytics firm based in Virginia. As CEO, Lao has called on her experience as a project manager from internships at GE, Google, and Apple, where, most recently, she led the rollout of the iPhone XR display screen. “I’ve never met anyone as good at managing people and tech,” says Tsai, an EECS master’s student who met Lao as a lab assistant for Abelson’s course 6.S198 (<a href="">Deep Learning Practicum</a>), and is now CTO of HINTS.</p> <p>As HINTS was getting off the ground, Lao co-founded a second startup,&nbsp;<a href="">ML Tidbits</a>, with EECS graduate student&nbsp;<a href="">Harini Suresh</a>. While learning to build AI models, both women grew frustrated by the tutorials on YouTube. “They were full of formulas, with very few pictures,” she says. “Even if the material isn’t that hard, it looks hard!”&nbsp;</p> <p>Convinced they could do better, Lao and Suresh reimagined a menu of intimidating topics like unsupervised learning and model-fitting as a set of inviting side dishes. Sitting cross-legged on a table, as if by a cozy fire, Lao and Suresh put viewers at ease with real-world anecdotes, playful drawings, and an engaging tone. Six more videos, funded by&nbsp;<a href="">MIT Sandbox</a>&nbsp;and the&nbsp;MIT-IBM Watson AI Lab, are planned for release this spring.&nbsp;</p> <p>If her audience learns one thing from ML Tidbits, Lao says, she hopes it’s that anyone can learn the basic underpinnings of AI. “I want them to think, ‘Oh, this technology isn't just something that professional computer scientists or mathematicians can touch. I can learn it too. I can form educated opinions and join discussions about how it should be used and regulated.’ ”</p> A PhD student in the MIT Department of Electrical Engineering and Computer Science, Natalie Lao has co-founded startups aimed at democratizing artificial intelligence and using AI to protect democracy by fighting false and misleading information.Photo: Andrew TsaiQuest for Intelligence, MIT-IBM Watson AI Lab, School of Engineering, Computer science and technology, Technology and society, STEM education, K-12 education, Apps, Invention, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical engineering and computer science (EECS), Artificial intelligence, Graduate, postdoctoral, Profile, education, Education, teaching, academics Gift will allow MIT researchers to use artificial intelligence in a biomedical device Device developed within the Department of Civil and Environmental Engineering has the potential to replace damaged organs with lab-grown ones. Wed, 29 Jan 2020 10:50:03 -0500 Maria Iacobo | Department of Civil and Environmental Engineering <p>Researchers in the MIT Department of Civil and Environmental Engineering (CEE) have received a gift to advance their work on a device designed to position living cells for growing human organs using acoustic waves. The Acoustofluidic Device Design with Deep Learning is being supported by Natick, Massachusetts-based MathWorks, a leading developer of mathematical computing software.</p> <p>“One of the fundamental problems in growing cells is how to move and position them without damage,” says John R. Williams, a professor in CEE. “The devices we’ve designed are like acoustic tweezers.”</p> <p>Inspired by the complex and beautiful patterns in the sand made by waves, the researchers' approach is to use sound waves controlled by machine learning to design complex cell patterns. The pressure waves generated by acoustics in a fluid gently move and position the cells without damaging them.</p> <p>The engineers developed a computer simulator to create a variety of device designs, which were then fed to an AI platform to understand the relationship between device design and cell positions.</p> <p>“Our hope is that, in time, this AI platform will create devices that we couldn’t have imagined with traditional approaches,” says Sam Raymond, who recently completed his doctorate working with Williams on this project. Raymond’s thesis title, "Combining Numerical Simulation and Machine Learning," explored the application of machine learning in computational engineering.</p> <p>“MathWorks and MIT have a 30-year long relationship that centers on advancing innovations in engineering and science,” says P.J. Boardman, director of MathWorks. “We are pleased to support Dr. Williams and his team as they use new methodologies in simulation and deep learning to realize significant scientific breakthroughs.”</p> <p>Williams and Raymond collaborated with researchers at the University of Melbourne and the Singapore University of Technology and Design on this project.</p> Patterns of suspended 1-micrometer polystyrene particles in channel shapes are given by a new AI platform (interior boundary highlighted in white). An applied surface acoustic wave propagates from left to right. Research, Artificial intelligence, Civil and environmental engineering, Sustainability, School of Engineering, Biomedicine, Bioengineering and biotechnology, Funding, Tissue engineering Technique reveals whether models of patient risk are accurate Computer scientists’ new method could help doctors avoid ineffective or unnecessarily risky treatments. Thu, 23 Jan 2020 05:00:00 -0500 Anne Trafton | MIT News Office <p>After a patient has a heart attack or stroke, doctors often use risk models to help guide their treatment. These models can calculate a patient’s risk of dying based on factors such as the patient’s age, symptoms, and other characteristics.</p> <p>While these models are useful in most cases, they do not make accurate predictions for many patients, which can lead doctors to choose ineffective or unnecessarily risky treatments for some patients.</p> <p>“Every risk model is evaluated on some dataset of patients, and even if it has high accuracy, it is never 100 percent accurate in practice,” says Collin Stultz, a professor of electrical engineering and computer science at MIT and a cardiologist at Massachusetts General Hospital. “There are going to be some patients for which the model will get the wrong answer, and that can be disastrous.”</p> <p>Stultz and his colleagues from MIT, IBM Research, and the University of Massachusetts Medical School have now developed a method that allows them to determine whether a particular model’s results can be trusted for a given patient. This could help guide doctors to choose better treatments for those patients, the researchers say.</p> <p>Stultz, who is also a professor of health sciences and technology, a member of MIT’s Institute for Medical Engineering and Sciences and Research Laboratory of Electronics, and an associate member of the Computer Science and Artificial Intelligence Laboratory, is the senior author of the <a href="" target="_blank">new study</a>. MIT graduate student Paul Myers is the lead author of the paper, which appears today in <em>Digital Medicine</em>.</p> <p><strong>Modeling risk</strong></p> <p>Computer models that can predict a patient’s risk of harmful events, including death, are used widely in medicine. These models are often created by training machine-learning algorithms to analyze patient datasets that include a variety of information about the patients, including their health outcomes.</p> <p>While these models have high overall accuracy, “very little thought has gone into identifying when a model is likely to fail,” Stultz says. “We are trying to create a shift in the way that people think about these machine-learning models. Thinking about when to apply a model is really important because the consequence of being wrong can be fatal.”</p> <p>For instance, a patient at high risk who is misclassified would not receive sufficiently aggressive treatment, while a low-risk patient inaccurately determined to be at high risk could receive unnecessary, potentially harmful interventions.</p> <p>To illustrate how the method works, the researchers chose to focus on a widely used risk model called the GRACE risk score, but the technique can be applied to nearly any type of risk model. GRACE, which stands for Global Registry of Acute Coronary Events, is a large dataset that was used to develop a risk model that evaluates a patient’s risk of death within six months after suffering an acute coronary syndrome (a condition caused by decreased blood flow to the heart). The resulting risk assessment is based on age, blood pressure, heart rate, and other readily available clinical features.</p> <p>The researchers’ new technique generates an “unreliability score” that ranges from 0 to 1. For a given risk-model prediction, the higher the score, the more unreliable that prediction. The unreliability score is based on a comparison of the risk prediction generated by a particular model, such as the GRACE risk-score, with the prediction produced by a different model that was trained on the same dataset. If the models produce different results, then it is likely that the risk-model prediction for that patient is not reliable, Stultz says.</p> <p>“What we show in this paper is, if you look at patients who have the highest unreliability scores — in the top 1 percent — the risk prediction for that patient yields the same information as flipping a coin,” Stultz says. “For those patients, the GRACE score cannot discriminate between those who die and those who don’t. It’s completely useless for those patients.”</p> <p>The researchers’ findings also suggested that the patients for whom the models don’t work well tend to be older and to have a higher incidence of cardiac risk factors.</p> <p>One significant advantage of the method is that the researchers derived a formula that tells how much two predictions would disagree, without having to build a completely new model based on the original dataset.&nbsp;</p> <p>“You don’t need access to the training dataset itself in order to compute this unreliability measurement, and that’s important because there are privacy issues that prevent these clinical datasets from being widely accessible to different people,” Stultz says.</p> <p><strong>Retraining the model</strong></p> <p>The researchers are now designing a user interface that doctors could use to evaluate whether a given patient’s GRACE score is reliable. In the longer term, they also hope to improve the reliability of risk models by making it easier to retrain models on data that include more patients who are similar to the patient being diagnosed.</p> <p>“If the model is simple enough, then retraining a model can be fast. You could imagine a whole suite of software integrated into the electronic health record that would automatically tell you whether a particular risk score is appropriate for a given patient, and then try to do things on the fly, like retrain new models that&nbsp;might be more appropriate,” Stultz says.</p> <p>The research was funded by the MIT-IBM Watson AI Lab. Other authors of the paper include MIT graduate student Wangzhi Dai; Kenney Ng, Kristen Severson, and Uri Kartoun of the Center for Computational Health at IBM Research; and Wei Huang and Frederick Anderson of the Center for Outcomes Research at the University of Massachusetts Medical School.</p> Researchers have developed a method that allows them to determine whether a particular risk model’s results can be trusted for a given patient.Research, Health care, Medicine, Machine learning, Artificial intelligence, Electrical Engineering & Computer Science (eecs), Health sciences and technology, Research Laboratory for Electronics, Computer Science and Artificial Intelligence Laboratory (CSAIL), Institute for Medical Engineering and Science (IMES), School of Engineering Using artificial intelligence to enrich digital maps Model tags road features based on satellite images, to improve GPS navigation in places with limited map data. Thu, 23 Jan 2020 00:00:00 -0500 Rob Matheson | MIT News Office <p>A model invented by researchers at MIT and Qatar Computing Research Institute (QCRI) that uses satellite imagery to tag road features in digital maps could help improve GPS navigation.&nbsp;&nbsp;</p> <p>Showing drivers more details about their routes can often help them navigate in unfamiliar locations. Lane counts, for instance, can enable a GPS system to warn drivers of diverging or merging lanes. Incorporating information about parking spots can help drivers plan ahead, while mapping bicycle lanes can help cyclists negotiate busy city streets. Providing updated information on road conditions can also improve planning for disaster relief.</p> <p>But creating detailed maps is an expensive, time-consuming process done mostly by big companies, such as Google, which send<s>s</s> vehicles around with cameras strapped to their hoods to capture video and images of an area’s roads. Combining that with other data can create accurate, up-to-date maps. Because this process is expensive, however, some parts of the world are ignored.</p> <p>A solution is to unleash machine-learning models on satellite images — which are easier to obtain and updated fairly regularly — to automatically tag road features. But roads can be occluded by, say, trees and buildings, making it a challenging task. In a <a href="">paper</a> being presented at the Association for the Advancement of Artificial Intelligence conference, the MIT and QCRI researchers describe “RoadTagger,” which uses a combination of neural network architectures to automatically predict the number of lanes and road types (residential or highway) behind obstructions.</p> <p>In testing RoadTagger on occluded roads from digital maps of 20 U.S. cities, the model counted lane numbers with 77 percent accuracy and inferred road types with 93 percent accuracy. The researchers are also planning to enable RoadTagger to predict other features, such as parking spots and bike lanes.</p> <p>“Most updated digital maps are from places that big companies care the most about. If you’re in places they don’t care about much, you’re at a disadvantage with respect to the quality of map,” says co-author Sam Madden, a professor in the Department of Electrical Engineering and Computer Science (EECS) and a researcher in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “Our goal is to automate the process of generating high-quality digital maps, so they can be available in any country.”</p> <p>The paper’s co-authors are CSAIL graduate students Songtao He, Favyen Bastani, and Edward Park; EECS undergraduate student Satvat Jagwani; CSAIL professors Mohammad Alizadeh and Hari Balakrishnan; and QCRI researchers Sanjay Chawla, Sofiane Abbar, and Mohammad Amin Sadeghi.</p> <p><strong>Combining CNN and GNN</strong></p> <p>Qatar, where QCRI is based, is “not a priority for the large companies building digital maps,” Madden says. Yet, it’s constantly building new roads and improving old ones, especially in preparation for hosting the 2022 FIFA World Cup.</p> <p>“While visiting Qatar, we’ve had experiences where our Uber driver can’t figure out how to get where he’s going, because the map is so off,” Madden says. “If navigation apps don’t have the right information, for things such as lane merging, this could be frustrating or worse.”</p> <p>RoadTagger relies on a novel combination of a convolutional neural network (CNN) — commonly used for images-processing tasks — and a graph neural network (GNN). GNNs model relationships between connected nodes in a graph and have become popular for analyzing things like social networks and molecular dynamics. The model is “end-to-end,” meaning it’s fed only raw data and automatically produces output, without human intervention.</p> <p>The CNN takes as input raw satellite images of target roads. The GNN breaks the road into roughly 20-meter segments, or “tiles.” Each tile is a separate graph node, connected by lines along the road. For each node, the CNN extracts road features and shares that information with its immediate neighbors. Road information propagates along the whole graph, with each node receiving some information about road attributes in every other node. If a certain tile is occluded in an image, RoadTagger uses information from all tiles along the road to predict what’s behind the occlusion.</p> <p>This combined architecture represents a more human-like intuition, the researchers say. Say part of a four-lane road is occluded by trees, so certain tiles show only two lanes. Humans can easily surmise that a couple lanes are hidden behind the trees. Traditional machine-learning models — say, just a CNN — extract features only of individual tiles and most likely predict the occluded tile is a two-lane road.</p> <p>“Humans can use information from adjacent tiles to guess the number of lanes in the occluded tiles, but networks can’t do that,” He says. “Our approach tries to mimic the natural behavior of humans, where we capture local information from the CNN and global information from the GNN to make better predictions.”</p> <p><strong>Learning weights&nbsp;&nbsp;&nbsp; </strong></p> <p>To train and test RoadTagger, the researchers used a real-world map dataset, called OpenStreetMap, which lets users edit and curate digital maps around the globe. From that dataset, they collected confirmed road attributes from 688 square kilometers of maps of 20 U.S. cities — including Boston, Chicago, Washington, and Seattle. Then, they gathered the corresponding satellite images from a Google Maps dataset.</p> <p>In training, RoadTagger learns weights — which assign varying degrees of importance to features and node connections —&nbsp;of the CNN and GNN. The CNN extracts features from pixel patterns of tiles and the GNN propagates the learned features along the graph. From randomly selected subgraphs of the road, the system learns to predict the road features at each tile. In doing so, it automatically learns which image features are useful and how to propagate those features along the graph. For instance, if a target tile has unclear lane markings, but its neighbor tile has four lanes with clear lane markings and shares the same road width, then the target tile is likely to also have four lanes. In this case, the model automatically learns that the road width is a useful image feature, so if two adjacent tiles share the same road width, they’re likely to have the same lane count.</p> <p>Given a road not seen in training from OpenStreetMap, the model breaks the road into tiles and uses its learned weights to make predictions. Tasked with predicting a number of lanes in an occluded tile, the model notes that neighboring tiles have matching pixel patterns and, therefore, a high likelihood to share information. So, if those tiles have four lanes, the occluded tile must also have four.</p> <p>In another result, RoadTagger accurately predicted lane numbers in a dataset of synthesized, highly challenging road disruptions. As one example, an overpass with two lanes covered a few tiles of a target road with four lanes. The model detected mismatched pixel patterns of the overpass, so it ignored the two lanes over the covered tiles, accurately predicting four lanes were underneath.</p> <p>The researchers hope to use RoadTagger to help humans rapidly validate and approve continuous modifications to infrastructure in datasets such as OpenStreetMap, where many maps don’t contain lane counts or other details. A specific area of interest is Thailand, Bastani says, where roads are constantly changing, but there are few if any updates in the dataset.</p> <p>“Roads that were once labeled as dirt roads have been paved over so are better to drive on, and some intersections have been completely built over. There are changes every year, but digital maps are out of date,” he says. “We want to constantly update such road attributes based on the most recent imagery.”</p> An AI model developed at MIT and Qatar Computing Research Institute that uses only satellite imagery to automatically tag road features in digital maps could improve GPS navigation, especially in countries with limited map data.Image: Google Maps/MIT NewsResearch, Computer science and technology, Algorithms, Transportation, Cities, Automobiles, Machine learning, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering, Artificial intelligence How to verify that quantum chips are computing correctly A new method determines whether circuits are accurately executing complex operations that classical computers can’t tackle. Mon, 13 Jan 2020 10:59:03 -0500 Rob Matheson | MIT News Office <p>In a step toward practical quantum computing, researchers from MIT, Google, and elsewhere have designed a system that can verify when quantum chips have accurately performed complex computations that classical computers can’t.</p> <p>Quantum chips perform computations using quantum bits, called “qubits,” that can represent the two states corresponding to classic binary bits — a 0 or 1 — or a “quantum superposition” of both states simultaneously. The unique superposition state can enable quantum computers to solve problems that are practically impossible for classical computers, potentially spurring breakthroughs in material design, drug discovery, and machine learning, among other applications.</p> <p>Full-scale quantum computers will require millions of qubits, which isn’t yet feasible. In the past few years, researchers have started developing “Noisy Intermediate Scale Quantum” (NISQ) chips, which contain around 50 to 100 qubits. That’s just enough to demonstrate “quantum advantage,” meaning the NISQ chip can solve certain algorithms that are intractable for classical computers. Verifying that the chips performed operations as expected, however, can be very inefficient. The chip’s outputs can look entirely random, so it takes a long time to simulate steps to determine if everything went according to plan.</p> <p>In a paper published today in <em>Nature Physics</em>, the researchers describe a novel protocol to efficiently verify that an NISQ chip has performed all the right quantum operations. They validated their protocol on a notoriously difficult quantum problem running on custom quantum photonic chip.</p> <p>“As rapid advances in industry and academia bring us to the cusp of quantum machines that can outperform classical machines, the task of quantum verification becomes time critical,” says first author Jacques Carolan, a postdoc in the Department of Electrical Engineering and Computer Science (EECS) and the Research Laboratory of Electronics (RLE). “Our technique provides an important tool for verifying a broad class of quantum systems. Because if I invest billions of dollars to build a quantum chip, it sure better do something interesting.”</p> <p>Joining Carolan on the paper are researchers from EECS and RLE at MIT, as well from the Google Quantum AI Laboratory, Elenion Technologies, Lightmatter, and Zapata Computing. &nbsp;</p> <p><strong>Divide and conquer</strong></p> <p>The researchers’ work essentially traces an output quantum state generated by the quantum circuit back to a known input state. Doing so reveals which circuit operations were performed on the input to produce the output. Those operations should always match what researchers programmed. If not, the researchers can use the information to pinpoint where things went wrong on the chip.</p> <p>At the core of the new protocol, called “Variational Quantum Unsampling,” lies a “divide and conquer” approach, Carolan says, that breaks the output quantum state into chunks. “Instead of doing the whole thing in one shot, which takes a very long time, we do this unscrambling layer by layer. This allows us to break the problem up to tackle it in a more efficient way,” Carolan says.</p> <p>For this, the researchers took inspiration from neural networks — which solve problems through many layers of computation —&nbsp;to build a novel “quantum neural network” (QNN), where each layer represents a set of quantum operations.</p> <p>To run the QNN, they used traditional silicon fabrication techniques to build a 2-by-5-millimeter NISQ chip with more than 170 control parameters — tunable circuit components that make manipulating the photon path easier. Pairs of photons are generated at specific wavelengths from an external component and injected into the chip. The photons travel through the chip’s phase shifters — which change the path of the photons — interfering with each other. This produces a random quantum output state —&nbsp;which represents what would happen during computation. The output is measured by an array of external photodetector sensors.</p> <p>That output is sent to the QNN. The first layer uses complex optimization techniques to dig through the noisy output to pinpoint the signature of a single photon among all those scrambled together. Then, it “unscrambles” that single photon from the group to identify what circuit operations return it to its known input state. Those operations should match exactly the circuit’s specific design for the task. All subsequent layers do the same computation — removing from the equation any previously unscrambled photons — until all photons are unscrambled.</p> <p>As an example, say the input state of qubits fed into the processor was all zeroes. The NISQ chip executes a bunch of operations on the qubits to generate a massive, seemingly randomly changing number as output. (An output number will constantly be changing as it’s in a quantum superposition.) The QNN selects chunks of that massive number. Then, layer by layer, it determines which operations revert each qubit back down to its input state of zero. If any operations are different from the original planned operations, then something has gone awry. Researchers can inspect any mismatches between the expected output to input states, and use that information to tweak the circuit design.</p> <p><strong>Boson “unsampling”</strong></p> <p>In experiments, the team successfully ran a popular computational task used to demonstrate quantum advantage, called “boson sampling,” which is usually performed on photonic chips. In this exercise, phase shifters and other optical components will manipulate and convert a set of input photons into a different quantum superposition of output photons. Ultimately, the task is to calculate the probability that a certain input state will match a certain output state. That will essentially be a sample from some probability distribution.</p> <p>But it’s nearly impossible for classical computers to compute those samples, due to the unpredictable behavior of photons. It’s been theorized that NISQ chips can compute them fairly quickly. Until now, however, there’s been no way to verify that quickly and easily, because of the complexity involved with the NISQ operations and the task itself.</p> <p>“The very same properties which give these chips quantum computational power makes them nearly impossible to verify,” Carolan says.</p> <p>In experiments, the researchers were able to “unsample” two photons that had run through the boson sampling problem on their custom NISQ chip — and in a fraction of time it would take traditional verification approaches.</p> <p>“This is an excellent paper that employs a nonlinear quantum neural network to learn the unknown unitary operation performed by a black box,” says Stefano Pirandola, a professor of computer science who specializes in quantum technologies at the University of York. “It is clear that this scheme could be very useful to verify the actual gates that are performed by a quantum circuit — [for example] by a NISQ processor. From this point of view, the scheme serves as an important benchmarking tool for future quantum engineers. The idea was remarkably implemented on a photonic quantum chip.”</p> <p>While the method was designed for quantum verification purposes, it could also help capture useful physical properties, Carolan says. For instance, certain molecules when excited will vibrate, then emit photons based on these vibrations. By injecting these photons into a photonic chip, Carolan says, the unscrambling technique could be used to discover information about the quantum dynamics of those molecules to aid in bioengineering molecular design. It could also be used to unscramble photons carrying quantum information that have accumulated noise by passing through turbulent spaces or materials. &nbsp;</p> <p>“The dream is to apply this to interesting problems in the physical world,” Carolan says.</p> Researchers from MIT, Google, and elsewhere have designed a novel method for verifying when quantum processors have accurately performed complex computations that classical computers can’t. They validate their method on a custom system (pictured) that’s able to capture how accurately a photonic chip (“PNP”) computed a notoriously difficult quantum problem.Image: Mihika PrabhuResearch, Computer science and technology, Algorithms, Light, Photonics, Nanoscience and nanotechnology, Quantum computing, electronics, Machine learning, Artificial intelligence, Research Laboratory of Electronics, Electrical Engineering & Computer Science (eecs), School of Engineering How well can computers connect symptoms to diseases? Models that map these relationships based on patient data require fine-tuning for certain conditions, study shows. Thu, 09 Jan 2020 17:10:38 -0500 Rob Matheson | MIT News Office <p>A new MIT study finds “health knowledge graphs,” which show relationships between symptoms and diseases and are intended to help with clinical diagnosis, can fall short for certain conditions and patient populations. The results also suggest ways to boost their performance.</p> <p>Health knowledge graphs have typically been compiled manually by expert clinicians, but that can be a laborious process. Recently, researchers have experimented with automatically generating these knowledge graphs from patient data. The MIT team has been studying how well such graphs hold up across different diseases and patient populations.</p> <p>In a paper presented at the Pacific Symposium on Biocomputing 2020, the researchers evaluated automatically generated health knowledge graphs based on real datasets comprising more than 270,000 patients with nearly 200 diseases and more than 770 symptoms.</p> <p>The team analyzed how various models used electronic health record (EHR) data, containing medical and treatment histories of patients, to automatically “learn” patterns of disease-symptom correlations. They found that the models performed particularly poorly for diseases that have high percentages of very old or young patients, or high percentages of male or female patients — but that choosing the right data for the right model, and making other modifications, can improve performance.</p> <p>The idea is to provide guidance to researchers about the relationship between dataset size, model specification, and performance when using electronic health records to build health knowledge graphs. That could lead to better tools to aid physicians and patients with medical decision-making or to search for new relationships between diseases and symptoms.</p> <p>“In the last 10 years, EHR use has skyrocketed in hospitals, so there’s an enormous amount of data that we hope to mine to learn these graphs of disease-symptom relationships,” says first author Irene Y. Chen, a graduate student in the Department of Electrical Engineering and Computer Science (EECS). “It is essential that we closely examine these graphs, so that they can be used as the first steps of a diagnostic tool.”</p> <p>Joining Chen on the paper are Monica Agrawal, a graduate student in MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL); Steven Horng of Beth Israel Deaconess Medical Center (BIDMC); and EECS Professor David Sontag, who is a member of CSAIL and the Institute for Medical Engineering and Science, and head of the Clinical Machine Learning Group.</p> <p><strong>Patients and diseases</strong></p> <p>In health knowledge graphs, there are hundreds of nodes, each representing a different disease and symptom. Edges (lines) connect disease nodes, such as “diabetes,” with correlated symptom nodes, such as “excessive thirst.” Google famously launched its own version in 2015, which was manually curated by several clinicians over hundreds of hours and is considered the gold standard. When you Google a disease now, the system displays associated symptoms.</p> <p>In a 2017 <em>Nature Scientific Reports</em> paper, Sontag, Horng, and other researchers leveraged data from the same 270,00 patients in their current study — which came from the emergency department at BIDMC between 2008 and 2013 — to build health knowledge graphs. They used three model structures to generate the graphs, called logistic regression, naive Bayes, and noisy OR. Using data provided by Google, the researchers compared their automatically generated health knowledge graph with the Google Health Knowledge Graph (GHKG). The researchers’ graph performed very well.</p> <p>In their new work, the researchers did a rigorous error analysis to determine which specific patients and diseases the models performed poorly for. Additionally, they experimented with augmenting the models with more data, from beyond the emergency room.</p> <p>In one test, they broke the data down into subpopulations of diseases and symptoms. For each model, they looked at connecting lines between diseases and all possible symptoms, and compared that with the GHKG. In the paper, they sort the findings into the 50 bottom- and 50 top-performing diseases. Examples of low performers are polycystic ovary syndrome (which affects women), allergic asthma (very rare), and prostate cancer (which predominantly affects older men). High performers are the more&nbsp;common diseases and conditions, such as heart arrhythmia and plantar fasciitis, which is tissue swelling along the feet.</p> <p>They found the noisy OR model was the most robust against error overall for nearly all of the diseases and patients. But accuracy decreased among all models for patients that have many co-occurring diseases and co-occurring symptoms, as well as patients that are very young or above the age of 85. Performance also suffered for patient populations with very high or low percentages of any sex.</p> <p><a name="_gjdgxs"></a>Essentially, the researchers hypothesize, poor performance is caused by patients and diseases that have outlier predictive performance, as well as potential unmeasured confounders. Elderly patients, for instance, tend to enter hospitals with more diseases and related symptoms than younger patients. That means it’s difficult for the models to correlate specific diseases with specific symptoms, Chen says. “Similarly,” she adds, “young patients don’t have many diseases or as many symptoms, and if they have a rare disease or symptom, it doesn’t present in a normal way the models understand.”</p> <p><strong>Splitting data</strong></p> <p>The researchers also collected much more patient data and created three distinct datasets of different granularity to see if that could improve performance. For the 270,000 visits used in the original analysis, the researchers extracted the full EHR history of the 140,804 unique patients, tracking back a decade, with around 7.4 million annotations total from various sources, such as physician notes.</p> <p>Choices in the dataset-creation process impacted the model performance as well. One of the datasets aggregates each of the 140,400 patient histories as one data point each. Another dataset treats each of the 7.4 million annotations as a separate data point. A final one creates “episodes” for each patient, defined as a continuous series of visits without a break of more than 30 days, yielding a total of around 1.4 million episodes.</p> <p>Intuitively, a dataset where the full patient history is aggregated into one data point should lead to greater accuracy since the entire patient history is considered. Counterintuitively, however, it also caused the naive Bayes model&nbsp;to perform more poorly for some diseases. “You assume the more intrapatient information, the better, with machine-learning models. But these models are dependent on the granularity of the data you feed them,” Chen says. “The type of model you use could get overwhelmed.”</p> <p>As expected, feeding the model demographic information can also be effective. For instance, models can use that information to exclude all male patients for, say, predicting cervical cancer. And certain diseases far more common for elderly patients can be eliminated in younger patients.</p> <p>But, in another surprise, the demographic information didn’t boost performance for the most successful model, so collecting that data may be unnecessary. That’s important, Chen says, because compiling data and training models on the data can be expensive and time-consuming. Yet, depending on the model, using scores of data may not actually improve performance.</p> <p>Next, the researchers hope to use their findings to build a robust model to deploy in clinical settings. Currently, the health knowledge graph learns relations between diseases and symptoms but does not give a direct prediction of disease from symptoms. “We hope that any predictive model and any medical knowledge graph would be put under a stress test so that clinicians and machine-learning researchers can confidently say, ‘We trust this as a useful diagnostic tool,’” Chen says.</p> A variety of new diagnostic models can analyze patient data and real-time symptoms to predict if a given patient has a particular disease. Research, Computer science and technology, Algorithms, Machine learning, Health, Health care, Disease, Medicine, Institute for Medical Engineering and Science (IMES), Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering Three from MIT are named 2020 fellows of the IEEE Two staff members from Lincoln Laboratory and a professor in the School of Engineering are recognized for their influential research. Wed, 08 Jan 2020 16:25:01 -0500 Dorothy Ryan | Lincoln Laboratory <p>Among the newly selected 2020 class of fellows of the Institute of Electrical and Electronics Engineers (IEEE) are three members of the MIT community: Hari Balakrishnan, the Fujitsu Chair Professor in the MIT Department of Electrical Engineering and Computer Science, and Richard Lippmann and Daniel Rabideau, members of the technical staff at MIT Lincoln Laboratory. The <a href="" target="_blank">IEEE</a>, the world's largest technical professional organization, confers the rank of fellow on senior members whose work has advanced innovation in their respective fields and has furthered the IEEE mission to foster the development of technology to benefit society.</p> <p><strong>Hari Balakrishnan</strong></p> <p>Balakrishnan was elevated to fellow for his contributions to the design and application of mobile sensing systems. These contributions include advances in mobile and sensor computing, internet congestion control and routing, overlay networks and peer-to-peer systems, and data management. His current research interests are in networking, sensing, and perception for a world of mobile devices connected to cloud services running in large data centers.</p> <p>In 2010, Balakrishnan cofounded Cambridge Mobile Telematics (CMT), which develops mobile sensing and artificial intelligence techniques to change driver behavior and make roads safer. This startup leveraged work on MIT's CarTel project that investigated how cars could be utilized as reliable mobile sensors. Today, this smartphone-centric telematics and analytics service provider has programs in more than 20 countries. Balakrishnan was previously a founding advisor to Meraki, which originated from an MIT wireless mesh networking project and was acquired by Cisco in 2012. He also cofounded StreamBase Systems (acquired by TIBCO) and was a network algorithms consultant for Sandburst (acquired by Broadcom). Like CMT, all these companies were spinoffs of MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), where Balakrishnan is a principal investigator.</p> <p>He has received several MIT honors for excellence in research and teaching, including the Harold E. Edgerton Faculty Achievement Award, the Ruth and Joel Spira Award for Excellence in Teaching, the Burgess and Elizabeth Jamieson Prize for Excellence in Teaching, the Junior Bose Award, and the Eta Kappa Nu Best Instructor Award.</p> <p>Balakrishnan is a member of the National Academy of Engineering and the American Academy of Arts and Sciences. He is a fellow of the Association for Computing Machinery, a Sloan Research Fellow, and an ACM dissertation award winner. His best-paper awards include the IEEE Communication Society’s William R. Bennett Prize and the IEEE's Foundations of Computer Science Test of Time Awards from IEEE special-interest groups on communications and computer networks, operating systems, management of data, mobility, and embedded networked sensor systems.</p> <p><strong>Richard Lippmann</strong></p> <p>During his 37-year career at Lincoln Laboratory, Lippmann has been responsible for groundbreaking work in two emerging fields: neural networks and cybersecurity. His research on neural networks resulted in the publication of the highly influential paper, "An Introduction to Computing with Neural Nets." This 1987 paper concisely explained the applicability of a neural-network approach to a variety of problems, won the first <em>IEEE Signal Processing Magazine</em> award, led to a global increase in research into neural networks, and has been cited more than 11,000 times. Lippmann's work has been instrumental in helping scientists accept, understand, and apply neural networks; for example, neural networks are now recognized as the most effective approach for speech recognition.</p> <p>Lippmann also led the design and development of the first quantitative, objective, and repeatable evaluations of the performance of computer intrusion-detection systems. Developed for the Defense Advanced Research Projects Agency in 1998-99, his approach looked at not only the probability that a network intrusion occurred, but also the probability that normal network traffic was tagged as an intrusion. This innovative approach allowed researchers to both better evaluate intrusion detection systems and apply modern machine-learning approaches. The datasets developed were made available to the cybersecurity community. They have been widely used and cited in more than 3,100 papers</p> <p>In the cybersecurity area, Lippmann and colleagues at Lincoln Laboratory developed security metrics that accurately estimate risk from important cyber threats and modeled the ways that adversaries progress through large enterprise networks. This collaborative work resulted in NetSPA (Network Security Planning Architecture), a software tool that creates attack graphs, has received two patents, and is used commercially.</p> <p>"Rich was the guiding technical force behind almost all the good technical ideas and impactful accomplishments of the laboratory’s first five years of cybersecurity activities," says Marc Zissman, an associate head of Lincoln Laboratory's Cyber Security and Information Sciences Division. "Everything good we did was something he suggested and did."</p> <p>Lippmann, a Lincoln Laboratory Fellow in the Cyber Security and Information Sciences Division, is currently investigating automated approaches to machine learning. He was a founding board member of the Neural Information Processing Systems conference; has given many talks on neural networks, including traveling around the world as an IEEE distinguished lecturer; served as the program chair for the 2008 Recent Advances in Intrusion Detection Conference; and has authored or coauthored more than 100 papers, reports, or books.</p> <p><strong>Daniel Rabideau</strong></p> <p>A nationally recognized expert in radar technology, Rabideau has made diverse contributions to the field. His work on modern adaptive signal-processing techniques has helped revolutionize radar capabilities in intelligence, surveillance, and reconnaissance applications. For example, he developed novel algorithms for space-time adaptive processing to mitigate clutter in radar returns from airborne moving target indication systems. He later extended his work in adaptive signal processing to the problem of terrain-scattered interference.</p> <p>Rabideau performed pioneering work on digital-array radar and its offshoot, multiple-input, multiple-output (MIMO) radar, that expanded the capabilities of radar systems. The paper he coauthored on MIMO technology, "Ubiquitous MIMO Multifunction Digital Array Radar," has been cited almost 400 times, influencing the advancement of MIMO techniques, which have become accepted as a fundamental topic in radar textbooks and have been applied to military radar systems and to commercial products — for example, automotive radars used for autonomous driving.</p> <p>"Dan has unparalleled intuition when it comes to radar technology. He is an incredibly creative engineer with an innate curiosity which propels him toward tackling the most difficult problems," says Jennifer Watson, the leader of the Lincoln Laboratory Airborne Radar Systems and Techniques Group, in which Rabideau is currently an assistant leader.</p> <p>Rabideau also developed important technology for the U.S. Navy's surface radar, including novel waveform techniques and advancements to digital-array radar, and participated in the study that culminated in the Navy's road map for its surface assets. He worked on algorithm development and hardware requirements to enable advanced capabilities for U.S. Navy airborne radar systems.</p> <p>During his 24-year tenure at Lincoln Laboratory, he has helped build the laboratory's substantial portfolio of programs focused on bringing electronic protection capability to radars used by the Navy and the U.S. Air Force. He has led the development of new systems and architectures with resilience to electronic attack, and has thus become a subject-matter expert on electric protection who is consulted by program managers from various government agencies.</p> <p>Rabideau's involvement in the larger radar community includes participating in studies, such as the Discoverer II space-based radar study, and serving as a member of the IEEE Aerospace and Electronic Systems Society Radar Systems Panel and as a reviewer and session chair for the Tri-Service Symposium and IEEE Radar Conferences, most recently as the technical chair for the 2019 IEEE Radar Conference held in Boston, Massachusetts. An author or coauthor of more than 60 publications, he has served often as a technical reviewer for several IEEE journals, including the <em>IEEE Transactions on Signal Processing</em> and the <em>IEEE Transactions on Aerospace and Electronic Systems.</em></p> <p>Newly elevated fellows can choose to be recognized at whichever of the 2020 IEEE conferences each wishes.</p> Left to right: Hari Balakrishnan, Richard Lippmann, and Daniel Rabideau have been named 2020 IEEE Fellows.Photos: MIT School of Engineering and MIT Lincoln LaboratoryLincoln Laboratory, School of Engineering, IEEE, Computer Science and Artificial Intelligence Laboratory (CSAIL), Awards, honors and fellowships, Cyber security, Computing, Radar, Faculty, Staff, Electrical Engineering & Computer Science (eecs) Finding the true potential of algorithms Using mathematical theory, Virginia Williams coaxes algorithms to run faster or proves they’ve hit their maximum speed. Tue, 07 Jan 2020 00:00:00 -0500 Rob Matheson | MIT News Office <p>Each semester, Associate Professor Virginia Vassilevska Williams tries to impart one fundamental lesson to her computer-science undergraduates: Math is the foundation of everything.</p> <p>Often, students come into Williams’ class, 6.006 (Introduction to Algorithms), wanting to dive into advanced programming that power the latest, greatest computing techniques. Her lessons instead focus on how algorithms are designed around core mathematical models and concepts. &nbsp;</p> <p>“When taking an algorithms class, many students expect to program a lot and perhaps use deep learning, but it’s very mathematical and has very little programming,” says Williams, the Steven G. (1968) and Renee Finn Career Development Professor who recently earned tenure in the Department of Electrical Engineering and Computer Science. “We don’t have much time together in class (only two hours a week), but I hope in that time they get to see a little of the beauty of math — because math allows you to see how and why everything works together. It really is a beautiful thing.”</p> <p>Williams’ life is very much shaped by math. As a child of two mathematician parents, she fell in love with the subject early on. But even though she excelled in the subject, her high school classes focused on German, writing, and biology. Returning to her first love in college and beyond, she applied her math skills to make waves in computer science.</p> <p>In highly influential work, Williams in 2012 improved an algorithm for “<a href="">matrix multiplication</a>” —&nbsp;a fundamental operation across computer science — that was thought to be the fastest iteration for 24 years. Years later, she co-founded an emerging field called “fine-grained complexity,” which seeks to explain, in part, how fast certain algorithms can solve various problems.</p> <p>In matrix multiplication, her work has now shifted slightly to showing that existing techniques “cannot do better,” she says. “We couldn’t improve the performance of our own algorithms anymore, so we came up with ways to explain why we couldn’t and why other methods can’t improve the performance either.”</p> <p><strong>Winding path to math</strong></p> <p>Growing up in Sofia, Bulgaria, Williams loved math and was a gifted student. But her parents often reminded her the mathematician’s life wasn’t exactly glamorous —especially when trying to find faculty gigs in the same area for two people. They sometimes traveled where work took them.</p> <p>That included a brief odyssey around the U.S. as a child. The first stop was Laramie, Wyoming. Her parents were visiting professors at the University of Wyoming, while Williams initially struggled through fourth grade because of the language barrier. “I didn’t really speak English, and was thrown into this school. My brother and I learned English watching the Disney channel, which was pretty fun,” says Williams, who today speaks Bulgarian, English, German, and some Russian.</p> <p>The next stop was Los Angeles — right around the time of the Rodney King riots. “The house on the other side of our street was set on fire,” Williams recalls. “Those were some very strange memories of L.A.”</p> <p>Returning to Bulgaria after two years, Williams decided to “explore her options” outside math by enrolling in the German Language High School in Sofia, the country’s top high school at the time, where she studied the German language, literature, history, and other humanities subjects. But, when it came to applying to colleges, she could never shake her first love. “I really tried to like the humanities, and what I learned is very helpful to me nowadays. But those subjects were very hard for me. My brain just doesn’t work that way,” she says. “I went back to what I like.”</p> <p><strong>Transfixed by algorithms</strong></p> <p>In 1999, Williams enrolled in Caltech. In her sophomore year, she became smitten by an exciting new field: computer science. “I took my first programming course, and I loved it,” she says.</p> <p>She became transfixed by matrix multiplication algorithms, which have some heavy-duty math at their core. These algorithms compute multiple arrays of numbers corresponding to some data and output a single combined matrix of some target values. Applications are wide-ranging, including computer graphics, product design, artificial intelligence, and biotechnology.</p> <p>As a PhD student at Carnegie Mellon, and beyond, she published <a href="">numerous papers</a>, on topics such as developing fast matrix multiplication algorithms in special algebraic structures, with applications including flight scheduling and network routing. After earning her PhD, she took on a series of postdoc and researcher positions at the Institute for Advanced Study, the University of California at Berkeley, and Stanford University, where she landed a faculty position in 2013 teaching courses on algorithms.</p> <p>In 2012, she developed a new algorithm that was faster than the Coppersmith–Winograd algorithm, which had reigned supreme in matrix multiplication since the 1980s. Williams’ method reduced the number of steps required to multiply matrices. Her algorithm is only slightly slower than the current record-holder.</p> <p><strong>Dealing with complexity</strong></p> <p>Between 2010 and 2015, Williams and her husband, Ryan Williams, who is also an MIT professor, became main founders of “fine-grained complexity.” The older field of “computational complexity” finds provably efficient algorithms and algorithms that are probably inefficient, based on some threshold of computational steps they take to solve a problem.</p> <p>Fine-grained complexity groups problems together by computational equivalence to better prove if algorithms are truly optimal or not. For instance, two problems may appear very different in what they solve and how many steps algorithms take to solve them. But fine-grained complexity shows such problems are secretly the same. Therefore, if an algorithm exists for one problem that uses fewer steps, then there must exist an algorithm for the other problem that uses fewer steps, and vice versa. On the flip side, if there exists a provably optimal algorithm for one problem, then all equivalent problems must have optimal algorithms. If someone ever finds a much faster algorithm for one problem, all the equivalent problems can be solved faster.</p> <p>Since co-launching the field, “it’s ballooned,” Williams says. “For most theoretical computer science conferences, you can now submit your paper under the heading ‘fine-grained complexity.’”</p> <p>In 2017, Williams came to MIT, where she says she has found impassioned, likeminded researchers. Many graduate students and colleagues, for instance, are working in topics related to fine-grained complexity. In turn, her students have introduced her to other subjects, such as cryptography, where she’s now introducing ideas from fine-grained complexity.</p> <p>She also sometimes studies “computational social choice,” a field that caught her eye during graduate school. Her work focuses on examining the computational complexity needed to rig sports games, voting schemes, and other systems where competitors are placed in paired brackets. If someone knows, for instance, which player will win in paired match-ups, a tournament organizer can place all players in specific positions in the initial seeding to ensure a certain player wins it all.</p> <p>Simulating all the possible combinations to rig these schemes can be very computationally complex. But Williams, an avid tennis player, authored a 2010 <a href="">paper</a> that found it’s fairly simple to rig a single-elimination tournament so a certain player wins, depending on accurate predictions for match-up winners and other factors.</p> <p>This year she co-wrote a <a href="">paper</a> that showed a tournament organizer could arrange an initial seeding and bribe certain top players — within a specific budget —&nbsp;to ensure a favorite player wins the tournament. “When I need a break from my usual work, I work in this field,” Williams says. “It’s a fun change of pace.”</p> <p>Thanks to the ubiquity of computing today, Williams’ graduate students often enter her classroom far more experienced in computer science than she was at their age. But to help steer them down a distinct path, she draws inspiration from her own college experiences, getting hooked on specific topics she still pursues today.</p> <p>“In order to do good research, you have to obsess over a problem,” Williams says. “I want them to find something in my course they can obsess over.”</p> Virginia WilliamsImage: Jared CharneyResearch, Computer science and technology, Algorithms, Profile, Faculty, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering, Mathematics Tool predicts how fast code will run on a chip Machine-learning system should enable developers to improve computing efficiency in a range of applications. Mon, 06 Jan 2020 00:00:00 -0500 Rob Matheson | MIT News Office <p>MIT researchers have invented a machine-learning tool that predicts how fast computer chips will execute code from various applications.&nbsp;&nbsp;</p> <p>To get code to run as fast as possible, developers and compilers — programs that translate programming language into machine-readable code — typically use performance models that run the code through a simulation of given chip architectures.&nbsp;</p> <p>Compilers use that information to automatically optimize code, and developers use it to tackle performance bottlenecks on the microprocessors that will run it. But performance models for machine code are handwritten by a relatively small group of experts and are not properly validated. As a consequence, the simulated performance measurements often deviate from real-life results.&nbsp;</p> <p>In series of conference papers, the researchers describe a novel machine-learning pipeline that automates this process, making it easier, faster, and more accurate. In a&nbsp;<a href="">paper</a>&nbsp;presented at the International Conference on Machine Learning in June, the researchers presented Ithemal, a neural-network model that trains on labeled data in the form of “basic blocks” — fundamental snippets of computing instructions — to automatically predict how long it takes a given chip to execute previously unseen basic blocks. Results suggest Ithemal performs far more accurately than traditional hand-tuned models.&nbsp;</p> <p>Then, at the November IEEE International Symposium on Workload Characterization, the researchers&nbsp;<a href="">presented</a>&nbsp;a benchmark suite of basic blocks from a variety of domains, including machine learning, compilers, cryptography, and graphics that can be used to validate performance models. They pooled more than 300,000 of the profiled blocks into an open-source dataset called BHive.&nbsp;During their evaluations, Ithemal predicted how fast Intel chips would run code even better than a performance model built by Intel itself.&nbsp;</p> <p>Ultimately, developers and compilers can use the tool to generate code that runs faster and more efficiently on an ever-growing number of diverse and “black box” chip designs.&nbsp;“Modern computer processors are opaque, horrendously complicated, and difficult to understand. It is also incredibly challenging to write computer code that executes as fast as possible for these processors,” says co-author on all three papers Michael Carbin, an assistant professor in the Department of Electrical Engineering and Computer Science (EECS) and a researcher in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “This tool is a big step forward toward fully modeling the performance of these chips for improved efficiency.”</p> <p>Most recently, in a&nbsp;<a href="">paper</a>&nbsp;presented at the NeurIPS conference in December, the team proposed a new technique to automatically generate compiler optimizations.&nbsp;&nbsp;Specifically, they automatically generate an algorithm, called Vemal, that converts certain code into vectors, which can be used for parallel computing. Vemal outperforms hand-crafted vectorization algorithms used in the LLVM compiler — a popular compiler used in the industry.</p> <p><strong>Learning from data</strong></p> <p>Designing performance models by hand can be “a black art,” Carbin says. Intel provides extensive documentation of more than 3,000 pages describing its chips’ architectures. But there currently exists only a small group of experts who will build performance models that simulate the execution of code on those architectures.&nbsp;</p> <p>“Intel’s documents are neither error-free nor complete, and Intel will omit certain things, because it’s proprietary,” says co-author on all three papers Charith Mendis, a graduate student in EECS and CSAIL. “However, when you use data, you don’t need to know the documentation. If there’s something hidden you can learn it directly from the data.”</p> <p>To do so, the researchers clocked the average number of cycles a given microprocessor takes to compute basic block instructions — basically, the sequence of boot-up, execute, and shut down — without human intervention. Automating the process enables rapid profiling of hundreds of thousands or millions of blocks.&nbsp;</p> <p><strong>Domain-specific architectures</strong></p> <p>In training, the Ithemal model analyzes millions of automatically profiled basic blocks to learn exactly how different chip architectures will execute computation. Importantly, Ithemal takes raw text as input and does not require manually adding features to the input data. In testing, Ithemal can be fed previously unseen basic blocks and a given chip, and will generate a single number indicating how fast the chip will execute that code.&nbsp;</p> <p>The researchers found Ithemal cut error rates in accuracy —&nbsp;meaning the difference between the predicted speed versus real-world speed —&nbsp;by 50 percent over traditional hand-crafted models. Further,&nbsp;in their next&nbsp;paper, they showed that&nbsp;Ithemal’s error rate was 10 percent, while the Intel performance-prediction model’s error rate was 20 percent on a variety of basic blocks across multiple different domains.</p> <p>The tool now makes it easier to quickly learn performance speeds for any new chip architectures, Mendis says. For instance, domain-specific architectures, such as Google’s new Tensor Processing Unit used specifically for neural networks, are now being built but aren’t widely understood. “If you want to train a model on some new architecture, you just collect more data from that architecture, run it through our profiler, use that information to train Ithemal, and now you have a model that predicts performance,” Mendis says.</p> <p>Next, the researchers are studying methods to make models interpretable. Much of machine learning is a black box, so it’s not really clear why a particular model made its predictions. “Our model is saying it takes a processor, say, 10 cycles to execute a basic block. Now, we’re trying to figure out why,” Carbin says. “That’s a fine level of granularity that would be amazing for these types of tools.”</p> <p>They also hope to use Ithemal to enhance the performance of Vemal even further and achieve better performance automatically.</p> MIT researchers have built a new benchmark tool that can accurately predict how long it takes given code to execute on a computer chip, which can help programmers tweak the code for better performance.Research, Computer science and technology, Algorithms, Machine learning, Data, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Engineering Bringing artificial intelligence and MIT to middle school classrooms MIT researchers and collaborators have developed an open-source curriculum to teach young students about ethics and artificial intelligence. Mon, 30 Dec 2019 10:45:01 -0500 Stefanie Koperniak | MIT Open Learning <p>In the age of Alexa, YouTube recommendations, and Spotify playlists, artificial intelligence has become a way of life, improving marketing and advertising, e-commerce, and more. But what are the ethical implications of technology that collects and learns personal information? How should society navigate these issues and shape the future?</p> <p>A new curriculum designed for middle school students aims to help them understand just that at an early age, as they grow up surrounded by the technology. The open-source educational material, designed by an MIT team and piloted at this year’s Massachusetts STEM Week this past fall, teaches students how AI systems are designed, how they can be used to influence the public — and also how to use them to be successful in jobs of the future.</p> <p>During Mass STEM Week in October, middle schools across the commonwealth replaced their regular curriculum with an immersive week of hands-on learning led by a team including Cynthia Breazeal, associate professor of media arts and sciences at MIT; Randi Williams ’18, graduate research assistant in the <a href="">Personal Robots Group</a> at the MIT Media Lab; and the nonprofit organization <a href="">i2 Learning</a>.</p> <p>“Preparing students for the future means having them engage in technology through hands-on activities. We provide students with tools and conceptual frameworks where we want them to engage with our materials as conscientious designers of AI-enabled technologies,” Breazeal says. “As they think through designing a solution to address a problem in their community, we get them to think critically about the ethical implications of the technology.”</p> <p>Three years ago, the Personal Robots Group began a program around teaching AI concepts to preschoolers. This effort then broadened into planning learning experiences for more children, and the group developed a curriculum geared toward middle school students. Last spring, an AI curriculum was shared with teachers and piloted in Somerville, Massachusetts, to determine which activities resonated the most in the classrooms.</p> <p>“We want to make a curriculum in which middle-schoolers can build and use AI — and, more importantly, we want them to take into account the societal impact of any technology,” says Williams.</p> <p>This curriculum, <a href="">How to Train Your Robot</a>, was first piloted at an i2 summer camp in Boston before being presented to teachers from local schools during Mass STEM Week. The teachers, many of whom had little familiarity with STEM subjects, also participated in two days of professional development training to prepare them to deliver more than 20 class hours of AI content to their students. The curriculum ran in three schools across six classrooms.</p> <p>The AI curriculum incorporates the work of Blakeley Hoffman Payne, a graduate research assistant in the Personal Robots Group, whose research focuses on the ethics of artificial intelligence and how to teach children to design, use, and think about AI. Students participated in discussions and creative activities, designing robot companions and using machine learning to solve real-world problems they have observed. At the end of the week, students share their inventions with their communities.</p> <p>“AI is an area that is becoming increasingly important in people’s lives,” says Ethan Berman, founder of i2 Learning and MIT parent. “This curriculum is very relevant to both students and teachers. Beyond just being a class on technology, it focuses on what it means to be a global citizen.”</p> <p>The creative projects provided opportunities for students to consider problems from a variety of angles, including thinking about issues of bias ahead of time, before a system is designed. For example, for one project that focused on sign language, the student trained her algorithm for understanding sign language around students of a wide range of skin tones, and incorporated adults, too — considering potential algorithmic bias to inform the design of the system.</p> <p>Another group of students built a “library robot,” designed to help find and retrieve a book for people with mobility challenges. Students had to think critically about why and how this might be helpful, and also to consider the job of a librarian and how this would impact a librarian’s work. They considered how a robot that finds and retrieves books might be able to free up more of a librarian’s time to actually help people and find information for them.</p> <p>Some of the current opportunities include scaling for more classrooms and schools, and also incorporating some other disciplines. There is interest in incorporating social studies, math, science, art, and music by finding ways to weave these other subjects into the AI projects. The main focus is on experiential learning that impacts how students think about AI.</p> <p>“We hope students walk away with a different understanding of AI and how it works in the world,” says Williams, “and that they feel empowered to play an important role in shaping the technology.”</p> MIT researchers piloted a new curriculum to teach middle school students about ethics and artificial intelligence at the second annual Mass STEM Week, a statewide event to encourage more young people to explore science, technology, engineering, and math studies and careers.Photo courtesy of i2 Learning.Office of Open Learning, Media Lab, Quest for Intelligence, STEM education, K-12 education, Education, teaching, academics, Artificial intelligence, Robotics, Schol of Architecture and Planning Finding a good read among billions of choices As natural language processing techniques improve, suggestions are getting speedier and more relevant. Fri, 20 Dec 2019 12:55:01 -0500 Kim Martineau | MIT Quest for Intelligence <p>With billions of books, news stories, and documents online, there’s never been a better time to be reading — if you have time to sift through all the options. “There’s a ton of text on the internet,” says&nbsp;<a href="">Justin Solomon</a>, an assistant professor at MIT. “Anything to help cut through all that material is extremely useful.”</p> <p>With the&nbsp;<a href="">MIT-IBM Watson AI Lab</a>&nbsp;and his&nbsp;<a href="">Geometric Data Processing Group</a>&nbsp;at MIT, Solomon recently presented a new technique for cutting through massive amounts of text at the&nbsp;<a href="">Conference on Neural Information Processing Systems</a> (NeurIPS). Their method combines three popular text-analysis tools — topic modeling, word embeddings, and optimal transport — to deliver better, faster results than competing methods on a popular benchmark for classifying documents.</p> <p>If an algorithm knows what you liked in the past, it can scan the millions of possibilities for something similar. As natural language processing techniques improve, those “you might also like” suggestions are getting speedier and more relevant.&nbsp;</p> <p>In the method presented at NeurIPS, an algorithm summarizes a collection of, say, books, into topics based on commonly-used words in the collection. It then divides each book into its five to 15 most important topics, with an estimate of how much each topic contributes to the book overall.&nbsp;</p> <p>To compare books, the researchers use two other tools: word embeddings, a technique that turns words into lists of numbers to reflect their similarity in popular usage, and optimal transport, a framework for calculating the most efficient way of moving objects — or data points — among multiple destinations.&nbsp;</p> <p>Word embeddings make it possible to leverage optimal transport twice: first to compare topics within the collection as a whole, and then, within any pair of books, to measure how closely common themes overlap.&nbsp;</p> <p>The technique works especially well when scanning large collections of books and lengthy documents. In the study, the researchers offer the example of Frank Stockton’s “The Great War Syndicate,” a 19th&nbsp;century American novel that anticipated the rise of nuclear weapons. If you’re looking for a similar book, a topic model would help to identify the dominant themes shared with other books — in this case, nautical, elemental, and martial.&nbsp;</p> <p>But a topic model alone wouldn’t identify Thomas Huxley’s 1863 lecture,&nbsp;“<a href="">The Past Condition of Organic Nature</a>,” as a good match. The writer was a champion of Charles Darwin’s theory of evolution, and his lecture, peppered with mentions of fossils and sedimentation, reflected emerging ideas about geology. When the themes in Huxley’s lecture are matched with Stockton’s novel via optimal transport, some cross-cutting motifs emerge: Huxley’s geography, flora/fauna, and knowledge themes map closely to Stockton’s nautical, elemental, and martial themes, respectively.</p> <p>Modeling books by their representative topics, rather than individual words, makes high-level comparisons possible. “If you ask someone to compare two books, they break each one into easy-to-understand concepts, and then compare the concepts,” says the study’s lead author&nbsp;<a href="">Mikhail Yurochkin</a>, a researcher at IBM.&nbsp;</p> <p>The result is faster, more accurate comparisons, the study shows. The researchers compared 1,720 pairs of books in the Gutenberg Project dataset in one second — more than 800 times faster than the next-best method.</p> <p>The technique also does a better job of accurately sorting documents than rival methods — for example, grouping books in the Gutenberg dataset by author, product reviews on Amazon by department, and BBC sports stories by sport. In a series of visualizations, the authors show that their method neatly clusters documents by type.</p> <p>In addition to categorizing documents quickly and more accurately, the method offers a window into the model’s decision-making process. Through the list of topics that appear, users can see why the model is recommending a document.</p> <p>The study’s other authors are&nbsp;<a href="">Sebastian Claici</a>&nbsp;and&nbsp;<a href="">Edward Chien</a>, a graduate student and a postdoc, respectively, at MIT’s Department of Electrical Engineering and Computer Science and Computer Science and Artificial Intelligence Laboratory, and&nbsp;<a href="">Farzaneh Mirzazadeh</a>, a researcher at IBM.</p> In a new study, researchers at MIT and IBM combine three popular text-analysis tools — topic modeling, word embeddings, and optimal transport — to compare thousands of documents per second. Here, they show that their method (left) clusters newsgroup posts by category more tightly than a competing method. Image courtesy of the researchers.Quest for Intelligence, MIT-IBM Watson AI Lab, Electrical engineering and computer science (EECS), Computer Science and Artificial Intelligence Laboratory (CSAIL), School of Engineering, Algorithms, Artificial intelligence, Computer science and technology, Data, Machine learning, Natural language processing When machine learning packs an economic punch Study: After eBay improved its translation software, international commerce increased sharply. Fri, 20 Dec 2019 10:04:08 -0500 Peter Dizikes | MIT News Office <p>A new study co-authored by an MIT economist shows that improved translation software can significantly boost international trade online — a notable case of machine learning having a clear impact on economic activity.</p> <p>The research finds that after eBay improved its automatic translation program in 2014, commerce shot up by 10.9 percent among pairs of countries where people could use the new system.&nbsp; &nbsp;</p> <p>“That’s a striking number. To have it be so clear in such a short amount of time really says a lot about the power of this technology,” says Erik Brynjolfsson, an MIT economist and co-author of a new paper detailing the results.</p> <p>To put the results in perspective, he adds, consider that physical distance is, by itself, also a significant barrier to global commerce. The 10.9 percent change generated by eBay’s new translation software increases trade by the same amount as “making the world 26 percent smaller, in terms of its impact on the goods that we studied,” he says.</p> <p>The paper, “Does Machine Translation Affect International Trade? Evidence from a Large Digital Platform,” appears in the December issue of <em>Management Science</em>. The authors are Brynjolfsson, who is the Schussel Family Professor of Management Science at the MIT Sloan School of Management, and Xiang Hui and Meng Liu, who are both assistant professors in the Olin Business School at Washington University in St. Louis.</p> <p><strong>Just cause</strong></p> <p>To conduct the study, the scholars examined what happened after eBay, in 2014, introduced its new eBay Machine Translation (eMT) system — a proprietary machine-learning program that, by several objective measures, significantly improved translation quality on eBay’s site. The new system initially was focused on English-Spanish translations, to facilitate trade between the United States and Latin America</p> <p>Previously, eBay had used Bing Translator to render the titles of objects for sale. By one evaluation measure, called the Human Acceptance Rate (HAR), in which three experts accept or reject translations, the eMT system increased the number of acceptable Spanish-language item titles on eBay from 82 percent to 90 percent.</p> <p>Using administrative data from eBay, the researchers then examined the volume of trade on the platform, within countries, after the eMT system went into use. Other factors being equal, the study showed that the new translation system not only had an effect on sales, but that trade increased by 1.06 percent for each additional word in the titles of items on eBay.</p> <p>That is a substantial change for a commerce platform on which, as the paper notes, items for sale often have long, descriptive titles such as “Diamond-Cut Stackable Thin Wedding Ring New .925 Sterling Silver Band Sizes 4-12,” or “Alpine Swiss Keira Women’s Trench Coast Double Breasted Wool Jacket Belted.” In those cases, making the translation clearer helps potential buyers understand exactly what they might be purchasing.</p> <p>Given the study’s level of specificity, Brynjolfsson calls it “a really fortunate natural experiment, with a before-and-after that sharply distinguished what happened when you had machine translation and when you didn’t.”</p> <p>The structure of the study, he adds, has enabled the researchers to say with confidence that the new eBay program, and not outside factors, directly generated the change in trade volume among affected countries.</p> <p>“In economics, it’s often hard to do causal analyses and prove that A caused B, not just that A was associated with B,” says Brynjolfsson. “But in this case, I feel very comfortable using causal language and saying that improvement in machine translation caused the increase in international trade.”</p> <p><strong>Larger puzzle: The productivity issue</strong></p> <p>The genesis of the paper stems from an ongoing question about new technology and economic productivity. While many forms of artificial intelligence have been developed and expanded in the last couple of decades, the impact of AI, including things like machine-translation systems, has not been obvious in economics statistics.</p> <p>“There’s definitely some amazing progress in the core technologies, including in things like natural language processing and translation,” Brynjolfsson says. “But what’s been lacking has been evidence of an economic impact, or business impact. So that’s a bit of a puzzle.”</p> <p>When looking to see if an economic impact for various forms of AI could be measured, Brynjolfsson, Hui, and Liu thought machine translation “made sense, because it’s a relatively straightforward implementation,” Brynjolfsson adds. That is, better translations could influence economic activity, at least on eBay, without any other changes in technology occurring.</p> <p>In this vein, the findings fit with a larger postulation Brynjolfsson has developed in recent years — that the adoption of AI technologies produces a “J-curve” in productivity. As Brynjolfsson has previously written, broad-ranging AI technologies nonetheless “require significant complementary investments, including business process redesign, co-invention of new products and business models, and investments in human capital” to have a large economic impact.</p> <p>As a result, when AI technologies are introduced, productivity may appear to slow down, and when the complementary technologies are developed, productivity may appear to take off — in the “J-curve” shape.</p> <p>So while Brynjolfsson believes the results of this study are clear, he warns against generalizing too much on the basis of this finding about the impact of machine learning and other forms of AI on economic activity. Every case is different, and AI will not always produce such notable changes by itself.</p> <p>“This was a case where not a lot of other changes had to happen in order for the technology to benefit the company,” Brynjolfsson says. “But in many other cases, much more complicated, complementary changes are needed. That’s why, in most cases with machine learning, it takes longer for the benefits to be delivered.”</p> A study co-authored by an MIT economist shows that an improved, automated language-translation system significantly boosted commerce on eBay’s website.Sloan School of Management, Business and management, Machine learning, Artificial intelligence, Economics, Technology and society, Social sciences, Innovation and Entrepreneurship (I&E) Exploring hip hop history with art and technology With its centerpiece exhibit for the forthcoming Universal Hip Hop Museum, an MIT team uses artificial intelligence to explore the rich history of hip hop music. Fri, 20 Dec 2019 09:00:00 -0500 Suzanne Day | Office of Open Learning <p>A new museum is coming to New York City in 2023, the year of hip-hop’s 50th birthday, and an MIT team has helped to pave the way for the city to celebrate the legacy of this important musical genre — by designing unique creative experiences at the intersection of art, learning, and contemporary technology.</p> <p>With “The [R]evolution of Hip Hop Breakbeat Narratives,” a team led by D. Fox Harrell, professor of digital media and artificial intelligence and director of the MIT Center for Advanced Virtuality, has created an art installation that takes museum-goers on an interactive, personalized journey through hip hop history.</p> <p>The installation served as the centerpiece of an event held this month by leaders of the highly anticipated Universal Hip Hop Museum (UHHM), which will officially open in just a few years in the Bronx — the future home of the UHHM, and where many agree that the genre of hip hop music originated.</p> <p>“Hip hop is much more than a musical genre. It is a global phenomenon, with a rich history and massive social and cultural impact, with local roots in the Bronx,” Harrell says. “As an educational center, the Universal Hip Hop Museum will have the power to connect people to the surrounding community.”</p> <p>Harrell’s immersive art installation takes museum-goers on a journey through hip hop culture and history, from the 1970s to the present. However, not everyone experiences the installation in the same way. Using a computational model of users’ preferences and artificial intelligence technologies to drive interaction, the team of artists and computer scientists from the Center for Advanced Virtuality has created layered, personalized virtual experiences.</p> <p>When approaching the exhibit, museum-goers are greeted by “The Elementals,” or novel characters named after the five elements of hip hop (MC, DJ, Breakdance, Graffiti Art, and Knowledge) that guide users and ask key questions — “What is your favorite hip hop song?” or “Which from this pair of lyrics do you like the most?” Based on those answers, the Elementals take users through their own personalized narrative of hip hop history.</p> <p>Harrell developed the Elementals with professors John Jennings of the University of California at Riverside and Stacey Robinson of the University of Illinois — artists collectively known as Black Kirby. This visual aesthetic ties the work into the rich, imaginative cultures and iconography of the African diaspora.</p> <p>Through these conversations with the Elementals they encounter, people can explore broad social issues surrounding hip hop, such as gender, fashion, and location. At the end of their journey, they can take home a personalized playlist of songs.&nbsp;</p> <p>“We designed the Breakbeat Narratives installation by integrating Microsoft conversational AI technologies, which made our user modeling more personable, with a music visualization platform from the TunesMap Educational Foundation,” Harrell says.</p> <p>The exploration of social issues is about as close to the heart of Harrell’s mission in the Center for Advanced Virtuality as one can get. In the center, Harrell designs virtual technologies to stimulate creative expression, cultural analysis, and positive social change.</p> <p>“We wanted to tell stories that pushed beyond stereotypical representations, digging into the complexities of both empowering and problematic representations that often coexist,” he says. “This work fits into our endeavor called the Narrative, Orality, and Improvisation Research (NOIR) Initiative that uses AI technologies to forward the art forms of diverse global cultures.”</p> <p>Through this art project enabled by contemporary technologies, Harrell hopes that he has helped museum leadership to achieve their goal of celebrating hip-hop’s heritage and legacy.</p> <p>“Now, people internationally can have a stake in this great art.”</p> Designed by an MIT team using artificial intelligence, “The [R]evolution of Hip Hop Breakbeat Narratives” is an immersive art installation designed for the forthcoming Universal Hip Hop Museum in New York City.Photo: MIT Center for Advanced VirtualityOffice of Open Learning, Machine learning, Artificial intelligence, History, Arts, Comparative Media Studies/Writing, Technology and society, Music, Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical Engineering & Computer Science (eecs), School of Humanities Arts and Social Sciences, School of Engineering Boston-area girls discover a passion for coding Weekend robotics workshops help middle and high school girls dispel “computing phobia.” Fri, 13 Dec 2019 12:20:01 -0500 Dorothy Ryan | Lincoln Laboratory <p>“My goal is to make computing 'normal' for girls,” says Sabina Chen from Lincoln Laboratory's Advanced Capabilities and Systems Group, who led a workshop that taught middle school girls how to program a robotic car to autonomously follow colored cones. The girls attended this enrichment class for eight consecutive Saturdays from September to November. “The class is about exposure [to computing] and interest-building,” she explains.</p> <p>While Chen was introducing the 20 middle school girls to coding and computer visualization software at the Beaver Works facility in MIT's Building 31, Eyassu Shimelis of the laboratory's Advanced Concepts and Technologies Group was conducting a similar series of classes for 21 high school girls in MIT's Building 33.</p> <p>The motivation behind holding the girls-only workshops is to foster a curiosity and familiarity in computing that may lead to a future increase in the number of women engaged in computer science. According to, in 2018 only 18 percent of bachelor's degrees in computer science were awarded to women; in electrical engineering, a major that often leads to professions involving computing, the percentage is even lower, at 13.7. The Bureau of Labor Statistics reports that women make up only about 21 percent of computer programmers, 19 percent of software developers, and 32 percent of website developers.</p> <p>While multiple theories exist as to why women are underrepresented in computer-related majors and jobs, one consensus is that young women do not have confidence in their ability to master computers. “The girls came in thinking they can't do it,” Chen says, adding that she finds the course worthwhile when “their eyes sort of sparkle when they realize they can do it.”</p> <p>Both workshops are based on a rigorous four-week course offered to high school seniors through the <a href="">Beaver Works Summer Institute</a> (BWSI). The summer course lets students explore the various technologies that can be used to enable an MIT-designed RACECAR (Rapid Autonomous Complex Environment Competing Ackermann steeRing) robotic vehicle to compete in fast, autonomous navigation around a mini “Grand Prix” racetrack. Although the Saturday sessions could not delve as deeply as the summer course into the software and various systems needed for the cars to tackle the obstacle-ridden Grand Prix, these weekend “crash courses” did cover the coding and computer-vison technology that allowed the girls to race their cars around a circular course set up in Building 31's high-bay space.</p> <p>Chen developed the curriculum for the middle school program that was offered to both boys and girls this past summer in conjunction with the BWSI. “It is designed so students learn a specific concept, apply it, and see an immediate result,” she says. The gratification of witnessing a hands-on application of a lesson is what keeps the students interested. Her curriculum will soon be online so that schools, robotics clubs, or even interested individuals can adapt it for themselves.&nbsp;</p> <p>Shimelis taught a similar version of a <a href="">RACECAR preliminary course</a> for Boston-area high schoolers. That course was developed in 2018 by Andrew Fishberg, who passed along his program when he moved on to tackle graduate studies at MIT. Shimelis is tweaking the course to address feedback from BWSI RACECAR students and teaching assistants, and to adapt it to his teaching style.</p> <p>Both Chen and Shimelis say they did not modify their courses for the girls-only sessions that were new this fall. They agree that the girls were eager to learn and capable of handling the classwork. “Many of the girls were faster at grasping the concepts than students in my summer course,” Shimelis notes. This is high praise, because to be accepted for the BWSI program, students must complete a prerequisite RACECAR online tutorial and submit teacher recommendations and stellar school transcripts.</p> <p>Chen says she was pleased by the change she saw in the girls from the beginning to the end of her workshop. “At the end, they were a lot more sure of themselves and more willing to explore their own ideas without fear.”</p> <p>According to Chen and Shimelis, the success of the two workshops can, in large part, be attributed to the dedicated help of a number of people. Sertac Karaman, an MIT engineering professor who developed the original RACECAR course for undergraduate students, provided guidance to both the instructors and students. A cadre of volunteers served as teaching assistants: Kourosh Arasteh, Olivia Brown, Juliana Furgala, Saivamsi Hanumanthu, Elisa Kircheim, Tzofi Klinghoffer, Marko Kocic, Aryk Ledet, Harrison Packer, Joyce Tam, and Jing Wang from Lincoln Laboratory's staff; and Andrew Schoer, a Boston University grad student who is a participant in the Laboratory's Lincoln Scholars tuition assistance program.</p> <p>The success of the workshops is captured in one student's answer to a course-evaluation question about what she gained: “I see myself coding in the future!”</p> Middle school and high school girls meet in MIT's Building 31 to showcase the capabilities of the autonomous cars their teams programmed to navigate a circuitous course. Photo: Joel GrimmLincoln Laboratory, Beaver works, STEM education, K-12 education, Education, teaching, academics, Robotics, Programming, Women in STEM, Classes and programs, Computer science and technology Differences between deep neural networks and human perception Stimuli that sound or look like gibberish to humans are indistinguishable from naturalistic stimuli to deep networks. Thu, 12 Dec 2019 13:05:01 -0500 Kenneth I. Blum | Center for Brains, Minds and Machines <p>When your mother calls your name, you know it’s her voice — no matter the volume, even over a poor cell phone connection. And when you see her face, you know it’s hers — if she is far away, if the lighting is poor, or if you are on a bad FaceTime call. This robustness to variation is a hallmark of human perception. On the other hand, we are susceptible to illusions: We might fail to distinguish between sounds or images that are, in fact, different. Scientists have explained many of these illusions, but we lack a full understanding of the invariances in our auditory and visual systems.
</p> <p>Deep neural networks also have performed speech recognition and image classification tasks with impressive robustness to variations in the auditory or visual stimuli. But are the invariances learned by these models similar to the invariances learned by human perceptual systems? A group of MIT researchers has discovered that they are different. They presented their findings yesterday at the 2019 <a href="">Conference on Neural Information Processing Systems</a>.
</p> <p>The researchers made a novel generalization of a classical concept: “metamers” — physically distinct stimuli that generate the same perceptual effect. The most famous examples of metamer stimuli arise because most people have three different types of cones in their retinae, which are responsible for color vision. The perceived color of any single wavelength of light can be matched exactly by a particular combination of three lights of different colors — for example, red, green, and blue lights. Nineteenth-century scientists inferred from this observation that humans have three different types of bright-light detectors in our eyes. This is the basis for electronic color displays on all of the screens we stare at every day. Another example in the visual system is that when we fix our gaze on an object, we may perceive surrounding visual scenes that differ at the periphery as identical. In the auditory domain, something analogous can be observed. For example, the “textural” sound of two swarms of insects might be indistinguishable, despite differing in the acoustic details that compose them, because they have similar aggregate statistical properties. In each case, the metamers provide insight into the mechanisms of perception, and constrain models of the human visual or auditory systems.
</p> <p>In the current work, the researchers randomly chose natural images and sound clips of spoken words from standard databases, and then synthesized sounds and images so that deep neural networks would sort them into the same classes as their natural counterparts. That is, they generated physically distinct stimuli that are classified identically by models, rather than by humans. This is a new way to think about metamers, generalizing the concept to swap the role of computer models for human perceivers. They therefore called these synthesized stimuli “model metamers” of the paired natural stimuli. The researchers then tested whether humans could identify the words and images.
</p> <p>“Participants heard a short segment of speech and had to identify from a list of words which word was in the middle of the clip. For the natural audio this task is easy, but for many of the model metamers humans had a hard time recognizing the sound,” explains first-author Jenelle Feather, a graduate student in the <a href="" target="_blank">MIT Department of Brain and Cognitive Sciences</a> (BCS) and a member of the <a href="" target="_blank">Center for Brains, Minds, and Machines</a> (CBMM). That is, humans would not put the synthetic stimuli in the same class as the spoken word “bird” or the image of a bird. In fact, model metamers generated to match the responses of the deepest layers of the model were generally unrecognizable as words or images by human subjects.
</p> <p><a href="">Josh McDermott</a>, associate professor in BCS and investigator in CBMM, makes the following case: “The basic logic is that if we have a good model of human perception, say of speech recognition, then if we pick two sounds that the model says are the same and present these two sounds to a human listener, that human should also say that the two sounds are the same. If the human listener instead perceives the stimuli to be different, this is a clear indication that the representations in our model do not match those of human perception.”
</p> <p>Examples of the model metamer stimuli can be found in the video below.</p> <div class="cms-placeholder-content-video"></div> <p>Joining Feather and McDermott on the paper are Alex Durango, a post-baccalaureate student, and Ray Gonzalez, a research assistant, both in BCS.
</p> <p>There is another type of failure of deep networks that has received a lot of attention in the media: adversarial examples (see, for example, "<a href="">Why did my classifier just mistake a turtle for a rifle?</a>"). These are stimuli that appear similar to humans but are misclassified by a model network (by design — they are constructed to be misclassified). They are complementary to the stimuli generated by Feather's group, which sound or appear different to humans but are designed to be co-classified by the model network. The vulnerabilities of model networks exposed to adversarial attacks are well-known — face-recognition software might mistake identities; automated vehicles might not recognize pedestrians.
</p> <p>The importance of this work lies in improving models of perception beyond deep networks. Although the standard adversarial examples indicate differences between deep networks and human perceptual systems, the new stimuli generated by the McDermott group arguably represent a more fundamental model failure — they show that generic examples of stimuli classified as the same by a deep network produce wildly different percepts for humans.
</p> <p>The team also figured out ways to modify the model networks to yield metamers that were more plausible sounds and images to humans. As McDermott says, “This gives us hope that we may be able to eventually develop models that pass the metamer test and better capture human invariances.”
</p> <p>“Model metamers demonstrate a significant failure of present-day neural networks to match the invariances in the human visual and auditory systems,” says Feather, “We hope that this work will provide a useful behavioral measuring stick to improve model representations and create better models of human sensory systems.”
</p> Associate Professor Josh McDermott (left) and graduate student Jenelle Feather generated physically distinct stimuli that are classified identically by models, rather than by humans.Photos: Justin Knight and Kris BrewerBrain and cognitive sciences, Center for Brains Minds and Machines, McGovern Institute, Research, Machine learning, Artificial intelligence, School of Science, Neuroscience This object-recognition dataset stumped the world’s best computer vision models Objects are posed in varied positions and shot at odd angles to spur new AI techniques. Tue, 10 Dec 2019 11:00:01 -0500 Kim Martineau | MIT Quest for Intelligence <p>Computer vision models have learned to identify objects in photos so accurately that some can outperform humans on some datasets. But when those same object detectors are turned loose in the real world, their performance noticeably drops, creating reliability concerns for self-driving cars and other safety-critical systems that use machine vision.</p> <p>In an effort to close this performance gap, a team of MIT and IBM researchers set out to create a very different kind of object-recognition dataset. It’s called <a href="">ObjectNet,</a> a play on ImageNet, the crowdsourced database of photos responsible for launching much of the modern boom in artificial intelligence.&nbsp;</p> <p>Unlike ImageNet, which features photos taken from Flickr and other social media sites, ObjectNet features photos taken by paid freelancers. Objects are shown tipped on their side, shot at odd angles, and displayed in clutter-strewn rooms. When leading object-detection models were tested on ObjectNet<strong>,</strong> their accuracy rates fell from a high of 97 percent on ImageNet to just 50-55 percent.</p> <p>“We created this dataset to tell people the object-recognition problem continues to be a hard problem,” says <a href="">Boris Katz</a>, a research scientist at MIT’s <a href="">Computer Science and Artificial Intelligence Laboratory</a> (CSAIL) and <a href="">Center for Brains, Minds and Machines</a> (CBMM).&nbsp; “We need better, smarter algorithms.” Katz and his colleagues will present ObjectNet and their results at the <a href="">Conference on Neural Information Processing Systems (NeurIPS)</a>.</p> <p>Deep learning, the technique driving much of the recent progress in AI, uses layers of artificial "neurons" to find patterns in vast amounts of raw data. It learns to pick out, say, the chair in a photo after training on hundreds to thousands of examples. But even datasets with millions of images can’t show each object in all of its possible orientations and settings, creating problems when the models encounter these objects in real life.</p> <p>ObjectNet is different from conventional image datasets in another important way: it contains no training images. Most datasets are divided into data for training the models and testing their performance. But the training set often shares subtle similarities with the test set, in effect giving the models a sneak peak at the test.&nbsp;</p> <p>At first glance, <a href="">ImageNet</a>, at 14 million images, seems enormous. But when its training set is excluded, it’s comparable in size to ObjectNet, at 50,000 photos.&nbsp;</p> <p>“If we want to know how well algorithms will perform in the real world, we should test them on images that are unbiased and that they’ve never seen before,” says study co-author <a href="">Andrei Barbu</a>, a research scientist at CSAIL and CBMM.<em>&nbsp;</em></p> <p><strong>A dataset that tries to capture the complexity of real-world objects&nbsp;</strong></p> <p>Few people would think to share the photos from ObjectNet with their friends, and that’s the point. The researchers hired freelancers from Amazon Mechanical Turk to take photographs of hundreds of randomly posed household objects. Workers received photo assignments on an app, with animated instructions telling them how to orient the assigned object, what angle to shoot from, and whether to pose the object in the kitchen, bathroom, bedroom, or living room.&nbsp;</p> <p>They wanted to eliminate three common biases: objects shown head-on, in iconic positions, and in highly correlated settings — for example, plates stacked in the kitchen.&nbsp;</p> <p>It took three years to conceive of the dataset and design an app that would standardize the data-gathering process. “Discovering how to gather data in a way that controls for various biases was incredibly tricky,” says study co-author <a href="">David Mayo</a>, a graduate student at MIT’s <a href="">Department of Electrical Engineering and Computer Science.</a> “We also had to run experiments to make sure our instructions were clear and that the workers knew exactly what was being asked of them.”&nbsp;</p> <p>It took another year to gather the actual data, and in the end, half of all the photos freelancers submitted had to be discarded for failing to meet the researchers’ specifications. In an attempt to be helpful, some workers added labels to their objects, staged them on white backgrounds, or otherwise tried to improve on the aesthetics of the photos they were assigned to shoot.</p> <p>Many of the photos were taken outside of the United States, and thus, some objects may look unfamiliar. Ripe oranges are green, bananas come in different sizes, and clothing appears in a variety of shapes and textures.</p> <p><strong>Object Net vs. ImageNet: how leading object-recognition models compare</strong></p> <p>When the researchers tested state-of-the-art computer vision models on ObjectNet, they found a performance drop of 40-45 percentage points from ImageNet. The results show that object detectors still struggle to understand that objects are three-dimensional and can be rotated and moved into new contexts, the researchers say. “These notions are not built into the architecture of modern object detectors,” says study co-author <a href="">Dan Gutfreund</a>, a researcher at IBM.</p> <p>To show that ObjectNet is difficult precisely because of how objects are viewed and positioned, the researchers allowed the models to train on half of the ObjectNet data before testing them on the remaining half. Training and testing on the same dataset typically improves performance, but here the models improved only slightly, suggesting that object detectors have yet to fully comprehend how objects exist in the real world.</p> <p>Computer vision models have progressively improved since 2012, when an object detector called AlexNet crushed the competition at the annual ImageNet contest. As datasets have gotten bigger, performance has also improved.</p> <p>But designing bigger versions of ObjectNet, with its added viewing angles and orientations, won’t necessarily lead to better results, the researchers warn. The goal of ObjectNet is to motivate researchers to come up with the next wave of revolutionary techniques, much as the initial launch of the ImageNet challenge did.</p> <p>“People feed these detectors huge amounts of data, but there are diminishing returns,” says Katz. “You can’t view an object from every angle and in every context. Our hope is that this new dataset will result in robust computer vision without surprising failures in the real world.”</p> <p>The study’s other authors are Julian Alvero, William Luo, Chris Wang, and Joshua Tenenbaum of MIT. The research was funded by the National Science Foundation, MIT’s Center for Brains, Minds, and Machines, the MIT-IBM Watson AI Lab, Toyota Research Institute, and the SystemsThatLearn@CSAIL initiative.</p> ObjectNet, a dataset of photos created by MIT and IBM researchers, shows objects from odd angles, in multiple orientations, and against varied backgrounds to better represent the complexity of 3D objects. The researchers hope the dataset will lead to new computer vision techniques that perform better in real life. Photo collage courtesy of the researchers.Quest for Intelligence, Center for Brains Minds and Machines, Electrical engineering and computer science (EECS), Computer Science and Artificial Intelligence Laboratory (CSAIL), School of Engineering, Algorithms, Artifical intelligence, Computer science and technology, Computer vision, Data, Machine learning, Software Intelligent Towing Tank propels human-robot-computer research A novel experimental facility integrates automation and active learning, illuminating a path to accelerated scientific discovery. Mon, 09 Dec 2019 16:05:01 -0500 MIT Sea Grant <p>In its first year of operation, the Intelligent Towing Tank (ITT) conducted about 100,000 total experiments, essentially completing the equivalent of a PhD student’s five years’ worth of experiments in a matter of weeks.</p> <p>The automated experimental facility, developed in the MIT Sea Grant Hydrodynamics Laboratory, automatically and adaptively performs, analyzes, and designs experiments exploring vortex-induced vibrations (VIVs). Important for engineering offshore ocean structures like marine drilling risers that connect underwater oil wells to the surface, VIVs remain somewhat of a phenomenon to researchers due to the high number of parameters involved.</p> <p>Guided by active learning, the ITT conducts series of experiments wherein the parameters of each next experiment are selected by a computer. Using an “explore-and-exploit” methodology, the system dramatically reduces the number of experiments required to explore and map the complex forces governing VIVs.</p> <p>What began as then-PhD candidate Dixia Fan’s quest to cut back on conducting a thousand or so laborious experiments — by hand — led to the design of the innovative system and a <a href="">paper</a> recently published in the journal <em>Science Robotics</em>.</p> <p>Fan, now a postdoc, and a team of researchers from the MIT Sea Grant College Program and MIT’s Department of Mechanical Engineering, École Normale Supérieure de Rennes, and Brown University, reveal a potential paradigm shift in experimental research, where humans, computers, and robots can collaborate more effectively to accelerate scientific discovery.</p> <p>The 33-foot whale of a tank comes alive, working without interruption or supervision on the venture at hand — in this case, exploring a canonical problem in the field of fluid-structure interactions. But the researchers envision applications of the active learning and automation approach to experimental research across disciplines, potentially leading to new insights and models in multi-input/multi-output nonlinear systems.</p> <p>VIVs are inherently-nonlinear motions induced on a structure&nbsp;in an oncoming irregular cross-stream, which prove vexing to study. The researchers report that the number of experiments completed by the ITT is already comparable to the total number of experiments done to date worldwide on the subject of VIVs.</p> <p>The reason for this is the large number of independent parameters, from flow velocity to pressure, involved in studying the complex forces at play. According to Fan, a systematic brute-force approach — blindly conducting 10 measurements per parameter in an eight-dimensional parametric space — would require 100 million experiments.</p> <p>With the ITT, Fan and his collaborators have taken the problem into a wider parametric space than previously practicable to explore. “If we performed traditional techniques on the problem we studied,” he explains, “it would take 950 years to finish the experiment.” Clearly infeasible, so Fan and the team integrated a Gaussian process regression learning algorithm into the ITT. In doing so, the researchers reduced the experimental burden by several orders of magnitude, requiring only a few thousand experiments.</p> <p>The robotic system automatically conducts an initial sequence of experiments, periodically towing a submerged structure along the length of the tank at a constant velocity. Then, the ITT takes partial control over the parameters of each next experiment by minimizing suitable acquisition functions of quantified uncertainties and adapting to achieve a range of objectives, like reduced drag.</p> <p>Earlier this year, Fan was awarded an&nbsp;MIT Mechanical Engineering de Florez Award for "Outstanding Ingenuity and Creative Judgment"&nbsp;in the development of the ITT. “Dixia’s design of the Intelligent Towing Tank is an outstanding example of using novel methods to reinvigorate mature fields,” says Michael Triantafyllou, Henry L. and Grace Doherty Professor in Ocean Science and Engineering, who acted as Fan’s doctoral advisor.</p> <p>Triantafyllou, a co-author on this paper and the director of the MIT Sea Grant College Program, says, “MIT Sea Grant has committed resources and funded projects using deep-learning methods in ocean-related problems for several years that are already paying off.” Funded by the National Oceanic and Atmospheric Administration and administered by the National Sea Grant Program, MIT Sea Grant is a federal-Institute partnership that brings the research and engineering core of MIT&nbsp;to bear on ocean-related challenges.</p> <p>Fan’s research points to a number of others utilizing automation and artificial intelligence in science: At Caltech, a robot scientist named “Adam” generates and tests hypotheses; at the Defense Advanced Research Projects Agency, the Big Mechanism program reads tens of thousands of research papers to generate new models.</p> <p>Similarly, the ITT applies human-computer-robot collaboration to accelerate experimental efforts. The system demonstrates a potential paradigm shift in conducting research, where automation and uncertainty quantification can considerably accelerate scientific discovery. The researchers assert that the machine learning methodology described in this paper can be adapted and applied in and beyond fluid mechanics, to other experimental fields.</p> <p>Other contributors to the paper include George Karniadakis from Brown University, who is also affiliated with MIT Sea Grant; Gurvan Jodin from ENS Rennes; MIT PhD candidate in mechanical engineering Yu Ma; and Thomas Consi, Luca Bonfiglio, and Lily Keyes from MIT Sea Grant.</p> <p>This work was supported by DARPA, Fariba Fahroo, and Jan Vandenbrande through an EQUiPS (Enabling Quantification of Uncertainty in Physical Systems) grant, as well as Shell, Subsea 7, and the MIT Sea Grant College Program.</p> Dixia Fan stands in front of the ITT holding some of the structures towed in the tank to study VIVs, which are important for engineering offshore ocean structures like marine drilling risers that connect underwater oil wells to the surface.Photo: Lily Keyes/MIT Sea GrantMIT Sea Grant, Mechanical engineering, Oceanography and ocean engineering, School of Engineering, Fluid dynamics, Robots, Robotics, automation, Artificial intelligence, Machine learning, Algorithms Using computers to view the unseen A new computational imaging method could change how we view hidden information in scenes. Fri, 06 Dec 2019 11:05:01 -0500 Rachel Gordon | CSAIL <p>Cameras and computers together can conquer some seriously stunning feats. Giving computers vision has helped us <a href="" target="_blank">fight wildfires in California</a>, understand complex and treacherous roads — and even see around corners.&nbsp;</p> <p>Specifically, seven years ago a group of MIT researchers created a <a href="" target="_self">new imaging system</a> that used floors, doors, and walls as “mirrors” to understand information about scenes outside a normal line of sight. Using special lasers to produce recognizable 3D images, the work opened up a realm of possibilities in letting us better understand what we can’t see.&nbsp;</p> <p>Recently, a different group of scientists from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has built off of this work, but this time with no special equipment needed: They developed a method that can reconstruct hidden video from just the subtle shadows and reflections on an observed pile of clutter. This means that, with a video camera turned on in a room, they can reconstruct a video of an unseen corner of the room, even if it falls outside the camera's field of view.&nbsp;</p> <div class="cms-placeholder-content-video"></div> <p>By observing the interplay of shadow and geometry in video, the team’s algorithm predicts the way that light travels in a scene, which is known as “light transport.” The system then uses that to estimate the hidden video from the observed shadows — and it can even construct the silhouette of a live-action performance.&nbsp;</p> <p>This type of image reconstruction could one day benefit many facets of society: Self-driving cars could better understand what’s emerging from behind corners, elder-care centers could enhance safety for their residents, and search-and-rescue teams could even improve their ability to navigate dangerous or obstructed areas.&nbsp;</p> <p>The technique, which is “passive,” meaning there are no lasers or other interventions to the scene, still currently takes about two hours to process, but the researchers say it could eventually be helpful in reconstructing scenes not in the traditional line of sight for the aforementioned applications.&nbsp;</p> <p>“You can achieve quite a bit with non-line-of-sight imaging equipment like lasers, but in our approach you only have access to the light that's naturally reaching the camera, and you try to make the most out of the scarce information in it,” says Miika Aittala, former CSAIL postdoc and current research scientist at NVIDIA<strong>, </strong>and the lead researcher on the new technique. “Given the recent advances in neural networks, this seemed like a great time to visit some challenges that, in this space, were considered largely unapproachable before.”&nbsp;</p> <p>To capture this unseen information, the team uses subtle, indirect lighting cues, such as shadows and highlights from the clutter in the observed area.</p> <p>In a way, a pile of clutter behaves somewhat like a pinhole camera, similar to something you might build in an elementary school science class: It blocks some light rays, but allows others to pass through, and these paint an image of the surroundings wherever they hit. But where a pinhole camera is designed to let through just the amount of right rays to form a readable picture, a general pile of clutter produces an image that is scrambled (by the light transport) beyond recognition, into a complex play of shadows and shading.&nbsp;</p> <p>You can think of the clutter, then, as a mirror that gives you a scrambled view into the surroundings around it — for example, behind a corner where you can’t see directly.&nbsp;</p> <p>The challenge addressed by the team's algorithm was to unscramble and make sense of these lighting cues. Specifically, the goal was to recover a human-readable video of the activity in the hidden scene, which is a multiplication of the light transport and the hidden video.&nbsp;</p> <p>However, unscrambling proved to be a classic “chicken-or-egg” problem. To figure out the scrambling pattern, a user would need to know the hidden video already, and vice versa.&nbsp;</p> <p>“Mathematically, it’s like if I told you that I’m thinking of two secret numbers, and their product is 80. Can you guess what they are? Maybe 40 and 2? Or perhaps 371.8 and 0.2152? In our problem, we face a similar situation at every pixel,” says Aittala. “Almost any hidden video can be explained by a corresponding scramble, and vice versa. If we let the computer choose, it’ll just do the easy thing and give us a big pile of essentially random images that don’t look like anything.”&nbsp;</p> <p>With that in mind, the team focused on breaking the ambiguity by specifying algorithmically that they wanted a “scrambling” pattern that corresponds to plausible real-world shadowing and shading, to uncover the hidden video that looks like it has edges and objects that move coherently.&nbsp;</p> <p>The team also used the surprising fact that neural networks naturally prefer to express “image-like” content, even when they’ve never been trained to do so, which helped break the ambiguity. The algorithm trains two neural networks simultaneously, where they’re specialized for the one target video only, using ideas from a machine learning concept called <a href="">Deep Image Prior</a>. One network produces the scrambling pattern, and the other estimates the hidden video. The networks are rewarded when the combination of these two factors reproduce the video recorded from the clutter, driving them to explain the observations with plausible hidden data.</p> <p>To test the system, the team first piled up objects on one wall, and either projected a video or physically moved themselves on the opposite wall. From this, they were able to reconstruct videos where you could get a general sense of what motion was taking place in the hidden area of the room.</p> <p>In the future, the team hopes to improve the overall resolution of the system, and eventually test the technique in an uncontrolled environment.&nbsp;</p> <p>Aittala wrote a new paper on the technique alongside CSAIL PhD students Prafull Sharma, Lukas Murmann, and Adam Yedidia, with MIT professors Fredo Durand, Bill Freeman, and Gregory Wornell. They will present it next week at the Conference on Neural Information Processing Systems in Vancouver, British Columbia.&nbsp;</p> Based on shadows that an out-of-view video casts on nearby objects, MIT researchers can estimate the contents of the unseen video. In the top row, researchers used this method to recreate visual elements in an out-of-view video; the original elements are shown in the bottom row. Images courtesy of the researchers.Computer Science and Artificial Intelligence Laboratory (CSAIL), Electrical engineering and computer science (EECS), School of Engineering, Research, Algorithms, Computer science and technology, Computer vision, Video games, Video, Disaster response, Film and Television Restructuring the MIT Department of Electrical Engineering and Computer Science The Institute&#039;s largest academic department reorganizes with new leadership as part of the formation of the MIT Schwarzman College of Computing. Thu, 05 Dec 2019 15:00:01 -0500 Terri Park | Lori LoTurco | MIT Schwarzman College of Computing | School of Engineering <p>As part of the founding of the MIT Stephen A. Schwarzman College of Computing, the Department of Electrical Engineering and Computer Science (EECS), the largest academic unit at MIT, has been restructured to provide a stronger base for enhancing existing programs, creating new opportunities, and increasing connections to other parts of the Institute.</p> <p>Jointly part of the School of Engineering and Schwarzman College of Computing, EECS is now composed of three overlapping sub-units in electrical engineering (EE), computer science (CS), and artificial intelligence and decision-making (AI+D), which brings together computer science-heritage AI and machine learning with electrical engineering-heritage information and decision systems to exploit their significant synergies. The department will remain responsible for Course 6.</p> <p>The <a href="" target="_blank">organizational plan for EECS</a> was developed over the summer based on the <a href="" target="_blank">final report</a> of the Organizational Structure Working Group of the Computing Task Force.</p> <p>“It is hard to imagine a School of Engineering without electrical engineering and a College of Computing without computer science. We expect that the creation of this new configuration will lead to a highly collaborative approach not only within EECS, but across campus and across disciplines,” says Dan Huttenlocher, dean of the MIT Schwarzman College of Computing.</p> <p>The plan calls for each of the units, termed a “faculty” to signify its differentiation from a traditional academic structure, each managed by a head of the faculty to lead the respective area and to contribute to the overall leadership of EECS, under the direction of the department head who will continue to oversee cross-cutting matters. The three faculty heads and the EECS head will each report jointly to Huttenlocher and to Anantha Chandrakasan, dean of the MIT School of Engineering.</p> <p>“This restructure will provide more autonomy to each unit,” says Chandrakasan. “The faculties will focus on faculty recruiting, mentoring, promotion, academic programs, and community building.”</p> <p>Asu Ozdaglar, Distinguished Professor of Engineering, a principal investigator at the Laboratory for Information and Decision Systems, and the newly appointed deputy dean of academics for the College of Computing, will remain the head of EECS, a position she has held since 2018. During her tenure, Ozdaglar has championed initiatives such as curriculum innovations to keep up with the ever-growing interest in computing, creation of new joint majors such as 6-14 (Computer Science, Economics, and Data Science), and Rising Stars in EECS, a workshop for female graduate students and postdocs interested in pursuing academic careers in computer engineering and electrical engineering, among many others.</p> <p>Joel Voldman, a professor of electrical engineering and computer science and an associate department head at EECS, will be the head of the faculty of electrical engineering. A principal investigator in the Research Laboratory of Electronics and the Microsystems Technology Laboratories, Voldman’s research focus is on developing microfluidic technology for biology and medicine, with an emphasis on cell sorting and immunology. In addition, he co-developed two introductory EECS courses: 6.03&nbsp;(Introduction to&nbsp;EECS via Medical Technology) and 6.S08/6.08&nbsp;(Interconnected Embedded Systems), and recently helped revise 6.002 (Circuits and Electronics).</p> <p>Arvind, the Charles W. and Jennifer C. Johnson Professor in Computer Science and Engineering, will step into the role of head of the faculty of computer science. Arvind’s research focuses on the specification and synthesis of complex digital systems, including microprocessors and accelerators, using a formalism known as guarded atomic actions. He leads the Computation Structures Group in the Computer Science and Artificial Intelligence Laboratory (CSAIL), and is a fellow of the&nbsp;Institute of Electrical and Electronics Engineers (IEEE) and the Association for Computing Machinery. He was elected into the&nbsp;National Academy of Engineering&nbsp;in 2008 and the Academy for Arts and Sciences in 2012.</p> <p>Antonio Torralba, the Thomas and Gerd Perkins Professor of Electrical Engineering and Computer Science, has been named the head of the faculty of artificial intelligence and decision-making. A principal investigator at CSAIL, a member of the Center for Brains, Minds, and Machines, and director of the MIT Quest for Intelligence and MIT-IBM Watson AI Lab, Torralba is the recipient of the 2008 National Science Foundation Career award, the best student paper award at the IEEE Conference on Computer Vision and Pattern Recognition in 2009, and the 2010 J.K. Aggarwal Prize from the International Association for Pattern Recognition. In 2017, he received the Frank Quick Faculty Research Innovation Fellowship and the Louis D. Smullin (’39) Award for Teaching Excellence.</p> <p>An advisory search committee made up of EECS faculty members — Ozdaglar (chair), Hari Balakrishnan, Marc Baldo, Duane Boning, Tommi Jaakkola, Patrick Jaillet, Dina Katabi, Jing Kong, Tomas Lozano-Perez, Alan Oppenheim, Daniela Rus, Armando Solar-Lezama, Collin Stultz, Ryan Williams, and Daniel Hastings — was formed to identify candidates to lead all three units to help guide the two deans in selecting the heads.</p> <p>Voldman, Arvind, and Torralba will begin their respective appointments on Jan. 1, 2020. Current Associate Department Head Saman Amarasinghe, an EECS professor and lead of the Commit compiler research group in CSAIL, will continue in his role until the new heads start their positions.</p> <p>“Thank you to everyone who served on the search committee and to Professer Amarasinghe for his tremendous leadership and contributions to EECS as an associate head. And please join us in congratulating Asu, Antonio, Arvind, and Joel for taking on these important new roles,” says Chandrakasan.</p> <p>“We look forward to working with the new leadership and all of the faculty in the department to help make EECS even stronger for our students and the MIT community, and more broadly, in leading this rapidly changing area,” adds Huttenlocher.&nbsp;</p> Clockwise from upper left: Asu Ozdaglar, Joel Voldman, Arvind, and Antonio TorralbaElectrical Engineering & Computer Science (eecs), Faculty, Administration, electronics, Artificial intelligence, Computer Science and Artificial Intelligence Laboratory (CSAIL), Machine learning, School of Engineering, MIT Schwarzman College of Computing Paul McEuen delivers inaugural Dresselhaus Lecture on cell-sized robots Cornell University professor and physicist uses nanoscale parts to create smart, active microbots. Wed, 04 Dec 2019 15:30:01 -0500 Amanda Stoll | MIT.nano <p>Functional, intelligent robots the size of a single cell are within reach, said Cornell University Professor Paul McEuen at the inaugural Mildred S. Dresselhaus Lecture at MIT on Nov. 13.</p> <p>“To build a robot that is on the scale of 100 microns in size, and have it work, that’s a big dream,” said McEuen, the John A. Newman Professor of Physical Science at Cornell University and director of Kavli Institute at Cornell for Nanoscale Science. “One hundred microns is a very special size. It is the border between the visible and the invisible, or microscopic, world.”</p> <p>In a talk entitled “Cell-sized Sensors and Robots” in front of a large audience in MIT’s 10-250 lecture hall, McEuen introduced his concept for a new generation of machines that work at the microscale by combining microelectronics, solar cells, and light.&nbsp;The microbots, as he calls them, operate&nbsp;using optical wireless integrated circuits and&nbsp;surface electrochemical actuators.</p> <div class="cms-placeholder-content-video"></div> <p><strong>Kicking off the Dresselhaus Lectures</strong></p> <p>Inaugurated this year to honor MIT professor and physicist Mildred "Millie" Dresselhaus, the Dresselhaus Lecture recognizes a significant figure in science and engineering whose&nbsp;leadership and impact echo the late Institute Professor's life, accomplishments, and values. The lecture will be presented annually in November, the month of her birth.</p> <p>Dresselhaus spent over 50 years at MIT, where she was a professor in the Department of Electrical Engineering and Computer Science (originally the Department of Electrical Engineering) as well as in the Department of Physics. She was MIT’s first female Institute Professor, co-organizer of the first MIT Women’s Forum, the first solo recipient of a Kavli Prize, and the first woman to win the National Medal of Science in the engineering category.</p> <p>Her research into the fundamental properties of carbon earned her the nickname the “Queen of Carbon Science.” She was also nationally known for her work to develop wider opportunities for women in science and engineering.</p> <p>“Millie was a physicist, a materials scientist, and an electrical engineer; an MIT professor, researcher, and doctoral supervisor; a prolific author; and a longtime leader in the scientific community,” said&nbsp;Asu Ozdaglar, current EECS department head, in her opening remarks. “Even in her final years, she was active in her field at MIT and in the department, attending EECS faculty meetings and playing an important role in developing the MIT.nano facility.”</p> <p><strong>Pushing the boundaries of physics</strong></p> <p>McEuen,&nbsp;who first met Dresselhaus when he attended graduate school at Yale University with her son, expressed what a privilege it was to celebrate Millie as the inaugural speaker. “When I think of my scientific heroes, it’s a very, very short list. And I think at the top of it would be Millie Dresselhaus.&nbsp;To be able to give this lecture in her honor means the world to me.”</p> <p>After earning his bachelor’s degree in engineering physics from the University of Oklahoma, McEuen continued his research at Yale University, where he completed his PhD in 1990 in applied physics. McEuen spent two years at MIT as a postdoc studying condensed matter physics, and then became a principal investigator at the Lawrence Berkeley National Laboratory. He spent eight years teaching at the University of California at Berkeley before joining the faculty at Cornell as a professor in the physics department in 2001.</p> <p>“Paul is a pioneer for our generation, exploring the domain of atoms and molecules to push the frontier even further. It is no exaggeration to say that his discoveries and innovations will help define the Nano Age,” said Vladimir Bulović, the founding faculty director of MIT.nano and the&nbsp;Fariborz Maseeh (1990) Professor in Emerging Technology.</p> <p><strong>“</strong><strong>The world is our oyster”</strong></p> <p>McEuen joked at the beginning of his talk that speaking of technology measured in microns sounds “so 1950s” in today’s world, in which researchers can manipulate at the scale of nanometers. One micron — an abbreviation for micrometer — is one millionth of a meter; a nanometer is one billionth of a meter.</p> <p>“[But] if you want a micro robot, you need nanoscale parts. Just as the birth of the transistor gave rise to all the computational systems we have now,” he said, “the birth of simple, nanoscale mechanical and electronic elements is going to give birth to a robotics technology at the microscopic scale of less than 100 microns.”</p> <p>The motto of McEuen and his research group at Cornell is “anything, as long as it’s small.” This focus includes fundamentals of nanostructures, atomically-thin origami for metamaterials and micromachines, and microscale smart phones and optobots. McEuen emphasized the importance of borrowing from other fields, such as microelectronics technology, to build something new. Cornell researchers have used this technology to build an optical wireless integrated circuit (OWIC) — essentially a microscopic cellphone made of solar cells that power it and receive external information, a simple transistor circuit to serve as its brain, and a light-emitting diode to blink out data.</p> <p>Why make something so small? The first reason is cost; the second is its wide array of applications. Such tiny devices could measure voltage or temperature, making them useful for microfluidic experiments. In the future, they could be deployed&nbsp;as&nbsp;smart, secure tags for counterfeiting, invisible sensors for the internet of things, or used for neural interfacing to measure electrical activity in the brain.</p> <p>Adding a&nbsp;surface electrochemical actuator to these OWICs brings mechanical movement to McEuen’s microbots. By capping a very thin piece of platinum on one side and applying a voltage to the other, “we could make all kinds of cool things.”</p> <p>At the end of his talk, McEuen answered audience questions moderated by&nbsp;Bulović, such as how do the microbots communicate with one another and what is their functional lifespan. He closed&nbsp;with a final quote from Millie Dresselhaus: “Follow your interests, get the best available education and training, set your sights high, be persistent, be flexible, keep your options open, accept help when offered, and be prepared to help others.”</p> <p>Nominations for the 2020 Dresselhaus lecture can be submitted <a href="" target="_blank">on MIT.nano’s website</a>. Any significant figure in science and engineering from anywhere in the world may be considered.</p> Cornell University’s Paul McEuen gives the inaugural Mildred S. Dresselhaus Lecture on cell-sized sensors and robots.Photo: Justin KnightMIT.nano, Electrical engineering and computer science (EECS), Physics, School of Engineering, School of Science, Nanoscience and nanotechnology, Special events and guest speakers, Faculty, Women in STEM, Carbon, Robots, Robotics, History of MIT