Larry Hardesty | MIT News Office

System could help prevent robots from overwhelming human teammates with information.

Communication system could make it easier to design systems that enable humans and robots to work together in emergency-response teams. Image: Jose-Luis Olivares/MIT.

Autonomous robots performing a joint task send each other continual updates: “I’ve passed through a door and am turning 90 degrees right.” “After advancing 2 feet I’ve encountered a wall. I’m turning 90 degrees right.” “After advancing 4 feet I’ve encountered a wall.” And so on.

Computers, of course, have no trouble filing this information away until they need it. But such a barrage of data would drive a human being crazy.

At the annual meeting of the Association for the Advancement of Artificial Intelligence last weekend, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) presented a new way of modeling robot collaboration that reduces the need for communication by 60 percent. They believe that their model could make it easier to design systems that enable humans and robots to work together — in, for example, emergency-response teams.

“We haven’t implemented it yet in human-robot teams,” says Julie Shah, an associate professor of aeronautics and astronautics and one of the paper’s two authors. “But it’s very exciting, because you can imagine: You’ve just reduced the number of communications by 60 percent, and presumably those other communications weren’t really necessary toward the person achieving their part of the task in that team.”

The work could have also have implications for multirobot collaborations that don’t involve humans. Communication consumes some power, which is always a consideration in battery-powered devices, but in some circumstances, the cost of processing new information could be a much more severe resource drain.

In a multiagent system — the computer science term for any collaboration among autonomous agents, electronic or otherwise — each agent must maintain a model of the current state of the world, as well as a model of what each of the other agents takes to be the state of the world. These days, agents are also expected to factor in the probabilities that their models are accurate. On the basis of those probabilities, they have to decide whether or not to modify their behaviors.

Communication costs

In some scenarios, a robot’s decision to broadcast a new item of information could force its fellows to update their models and churn through all those probabilities again. If the information is inessential, broadcasting it could introduce serious delays, to no purpose.

And the MIT researchers’ work suggests that 60 percent of communications in multiagent systems may be inessential. The state-of-the-art method for modeling multiagent systems is called a decentralized partially observable Markov decision process, or Dec-POMDP. A Dec-POMDP factors in several types of uncertainty; not only does it consider whether an agent’s view of the world is correct and whether its estimate of its fellows’ worldviews is correct, it also considers whether any action it takes will be successful. The robot may plan, for instance, to move forward 20 feet but find that crosswinds blow it off course.

Dec-POMDPs generally assume some prior knowledge about the environment in which the agents will be operating. Because Shah and Vaibhav Unhelkar, a graduate student in aeronautics and astronautics and first author on the new paper, were designing a system with emergency-response applications in mind, they couldn’t make that assumption. Emergency-response teams will usually be entering unfamiliar environments, and the very nature of the emergency could render the best prior information obsolete.

Adding the requirement of mapping the environment on the fly, however, makes the problem of computing a multiagent plan prohibitively time consuming. So Shah and Unhelkar’s system ignores uncertainty about actions’ effectiveness and assumes that whatever an agent attempts to do, it will do.

Balancing act

When an agent acquires a new item of information — that, for instance, a given passage through a building is blocked — it has three choices: it can ignore the information; it can use it but not broadcast it; or it can use it and broadcast it.

Each of these choices has benefits but imposes costs. In Shah and Unhelkar’s model, communication is a cost. But if an agent incorporates new information into its own model of the world and doesn’t broadcast it, it also incurs a cost, as its worldview becomes more difficult for its fellows to estimate correctly. For every new item of information an agent acquires, Shah and Unhelkar’s system performs that cost-benefit analysis, based on the agent’s model of the world, its expectations of its fellows’ actions, and the likelihood of accomplishing the joint goal more efficiently.

The researchers tested their system on more than 300 computer simulations of rescue tasks in unfamiliar environments. A version of their system that permitted extensive communication completed the tasks at a rate between 2 and 10 percent higher than the version that reduced communication by 60 percent.

In the experiments, however, all the agents were electronic. “What I’d be willing to bet, although we have to wait until we do the human-subject experiments, is that the human-robot team will fail miserably if the system is just telling the person all sorts of spurious information all the time,” Shah says. “For human-robot teams, I think that this algorithm is going to make the difference between a team that can function effectively versus a team that just plain can’t.”

In a separate research project, members of Shah’s group have asked teams of human subjects to execute similar virtual rescue missions that computer systems did in the experiments reported in the new paper. Using machine-learning algorithms, the researchers have mined the results for statistics on human communication patterns, which can be incorporated into the new model to more explicitly accommodate human-robot teams.

“It is well-understood that in human teams, when one team member gains new information, broadcasting this new information to all team members is generally not a good solution, especially when the cost of communication is high,” says Tim Miller, an assistant professor of computing and information systems at the University of Melbourne in Australia. “This work has applications outside of multiagent systems, reaching into the critical area of human-agent collaboration, where communication can be costly, but more importantly, human team members are quickly overloaded if presented with too much information.”

Read this article on MIT News.

Larry Hardesty | MIT News Office

Algorithm lets planning systems generate backup plans efficiently.

“The problem with planning contingencies is that there are so many things that can go wrong, if you generated plans for all possible contingencies, you’d go nuts,” Brian Williams says. Image: MIT News.

Planning algorithms are widely used in logistics and control. They can help schedule flights and bus routes, guide autonomous robots, and determine control policies for the power grid, among other things.

In recent years, planning algorithms have begun to factor in uncertainty — variations in travel time, erratic communication between autonomous robots, imperfect sensor data, and the like. That causes the scale of the planning problem to grow exponentially, but researchers have found clever ways to solve it efficiently.

Now, researchers at MIT and the Australian National University (ANU) have made the problem even more complex, by developing a planning algorithm that also generates contingency plans, should the initial plan prove too risky. It also identifies the conditions — say, sensor readings or delays incurred — that should trigger a switch to a particular contingency plan.

Despite the extra labor imposed by generating contingency plans, the algorithm still provides mathematical guarantees that its plans’ risk of failure falls below some threshold, which the user sets.

“The problem with planning contingencies is that there are so many things that can go wrong, if you generated plans for all possible contingencies, you’d go nuts,” says Brian Williams, a professor of aeronautics and astronautics whose group developed the new system. “So then the question ends up being, ‘How many contingencies do you generate?’”

Pedro Santana, a graduate student in aeronautics and astronautics, is first author on a paper describing the system, which he presented at the annual meeting of the Association for the Advancement of Artificial Intelligence last weekend. He’s joined by Williams and Sylvie Thiébaux, a professor of computer science at the Australian National University and a researcher with Australia’s National Information Communications Technology Australia (NICTA) research program, which has a partnership with MIT.

Probabilistic pruning

As Williams explains, the range of possible decisions that a planner faces can be represented as a data structure called a graph. A graph consists of nodes, which are usually represented as circles, and edges, which are represented as line segments connecting the nodes. Network diagrams and flow charts are familiar examples of a graph.

In a planning system, each node of the graph represents a decision point, such as, “Should I take the bus or the subway?” A path through the graph can be evaluated according to the rewards it offers — you reach your destination safely — and the penalties it imposes — you’ll be five minutes late. The optimal plan is the one that maximizes reward.

Factoring in probabilities makes that type of reward calculation much more complex: The average bus trip might be 15 minutes, but there’s some chance that it will be 35; the average subway trip might be 18 minutes, but it’s almost never more than 24. In that context, for even a relatively simple planning task, canvassing contingency plans can be prohibitively time consuming.

To make the problem tractable, the MIT and ANU researchers borrowed a technique from some earlier work from Williams’ group. Before the planner begins constructing the graph, it asks the user to set risk thresholds. A researcher trying to develop a data-gathering plan for a multimillion-dollar underwater robot, for example, might be satisfied with a 90 percent probability that the robot will take all the sensor readings it’s supposed to — but they might want a 99.9 percent probability that the robot won’t collide with a rock face at high speed.

The researchers’ algorithm treats these thresholds as a “risk budget” that it spends as it explores paths through the graph. If the planner determines that a given branch of the graph will exceed the budget, it lops it off.

Staying optimistic

That determination has to be rapid, however. So the researchers use some simple rules of thumb — or, in computer parlance, heuristics — to evaluate branches. Every path through a given branch, for instance, might include a different car route between two points, each with its own probability distribution of possible travel times. But if traversing a straight line between the points, at the maximum allowed speed, will still incur intolerable delays, there’s no point in evaluating probabilities for every route. The branch can be eliminated.

As long as the heuristics are optimistic — possibly underestimating risk but never overestimating it — the planner can lop off branches without compromising the quality of the final plans. Sometimes those heuristics are application-specific, like the one that evaluates routes geometrically. But sometimes they’re not.

For instance, one of the reasons that probability distributions add so much complexity to planning calculations is that they’re nonlinear: Their graphs take on shapes that are more complicated than simple lines. In a paper being presented at the International Conference on Automated Planning and Scheduling in June, Santana — again first author — Williams, and colleagues at MIT, the University of São Paulo in Brazil, and Caltech describe a way to produce linear approximations of probability distributions that are much easier to work with mathematically.

Those approximations are optimistic: They never overestimate risk. But a computer can evaluate them thousands of times faster than it can a nonlinear distribution. Such heuristics offer hope that the researchers’ planning system could update plans on the fly, in light of new information, as well as generating contingency plans in advance.

"The authors have tackled a very important problem that is highly relevant to NASA missions, where we are increasingly reliant on autonomous behaviors programmed into our spacecraft,” says Michel Ingham, technical group supervisor for the System Architectures and Behaviors group at NASA’s Jet Propulsion Laboratory.

“Their work is especially relevant for in-situ missions on the surface of asteroids and comets, planetary bodies like Mars, or not-too-distant-future missions to outer-planet icy moons like Europa and Enceladus. Our robotic spacecraft exploring these locales will be challenged to execute their missions efficiently in the presence of significant uncertainty in both their own state and the state of the surrounding environment, while avoiding obstacles and hazards. They will need to appropriately balance accomplishing their scientific and exploration objectives with keeping themselves healthy and out of trouble,” he added.

“The authors have provided a method and capability for planning a robot’s activities, which optimizes the reward associated with accomplishing mission objectives while bounding the ‘execution risk’ taken by the mission. This means that our future mission operators could be able to specify how conservative or aggressive they want the spacecraft to be by choosing an appropriate risk bound and have some guarantee that the spacecraft will not exceed this level of risk."

Read this article on MIT News.

The Department of Electrical Engineering and Computer Science has appointed Michael Carbin as the Jamieson Career Development Professor.

Michael Carbin joined EECS in January 2016 as an assistant professor. His research interests include the design of programming systems that deliver improved performance and resilience by incorporating approximate computing and self-healing.

Carbin is also a principal investigator at the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL). Carbin received his BS from Stanford University and his PhD from MIT in 2015. His work on program analysis at Stanford University as an undergraduate received an award for Best Computer Science Undergraduate Honors Thesis. As a graduate student, he has received the MIT Lemelson Presidential and Microsoft Research Graduate Fellowships. His recent research on verifying the reliability of programs that execute on unreliable hardware has received best paper awards at OOPSLA 2013 and 2014.

The Department of Electrical Engineering and Computer Science has appointed Luqiao Liu as the Robert J. Shillman (1974) Career Development Professor.

Luqiao Liu joined EECS in July 2015 as an assistant professor. He is an expert in the properties of magnetic materials and their application to non-volatile memory and spin-based logics. His research aims to use spin-based electronics to solve the power, scaling, and speed issues associated with conventional semiconductor devices.

Liu is also a core and resident member of the MIT Microsystems Technology Laboratories (MTL). Liu received his BS in physics from Peking University, Beijing, in 2006 and his PhD in applied physics from Cornell Uniersity, Ithaca, NY, in 2012. Prior to joining MIT, he was a research staff member in the magnetic random access memory (MRAM) group of IBM T. J. Watson Research Center. Professor Liu's research focuses on exploring new materials and mechanisms that can be used as building blocks for spintronic devices, which operate at high speed with low power consumption.

MIT News Office

Aleksander Madry, the NBX Career Development Assistant Professor of Electrical Engineering and Computer Science, among 126 selected from across the U.S. and Canada.

Eleven MIT researchers from eight School of Science and School of Engineering departments are among the 126 American and Canadian researchers awarded 2016 Sloan Research Fellowships, the Alfred P. Sloan Foundation announced today.

New MIT-affiliated Sloan Research Fellows are: Otto X. Cordero, an assistant professor of civil and environmental engineering; Vadim Gorin, an assistant professor of mathematics; R. Scott Kemp, an assistant professor of nuclear science and engineering; Yen-Jie Lee, an assistant professor of physics; Gene-Wei Li, an assistant professor of biology; Aleksander Madry, the NBX Career Development Assistant Professor of Electrical Engineering and Computer Science; Ankur Moitra, an assistant professor of mathematics; Yogesh Surendranath, an assistant professor of chemistry; William A. Tisdale, an assistant professor of chemical engineering; Mark Vogelsberger, an assistant professor of physics; and Jing-Ke Weng, the Thomas D. and Virginia W. Cabot Assistant Professor of biology.

Awarded annually since 1955, the Sloan Research Fellowships are given to early-career scientists and scholars whose achievements and potential identify them as rising stars among the next generation of scientific leaders. This year’s recipients are drawn from 52 colleges and universities across the United States and Canada.

“Getting early-career support can be a make-or-break moment for a young scholar,” said Paul L. Joskow, president of the Alfred P. Sloan Foundation, in a press release. “In an increasingly competitive academic environment, it can be difficult to stand out, even when your work is first rate. The Sloan Research Fellowships have become an unmistakable marker of quality among researchers. Fellows represent the best-of-the-best among young scientists.”

Administered and funded by the foundation, the fellowships are awarded in eight scientific fields: chemistry, computer science, economics, mathematics, evolutionary and computational molecular biology, neuroscience, ocean sciences, and physics. To qualify, candidates must first be nominated by fellow scientists and subsequently selected by an independent panel of senior scholars. Fellows receive $50,000 to be used to further their research.

Since the beginning of the program, 43 Sloan Fellows have earned Nobel Prizes, 16 have won the Fields Medal in mathematics, 68 have received the National Medal of Science, and 15 have won the John Bates Clark Medal in economics.

For a complete list of this year’s winners, visit the Sloan Research Fellowships website.

Read this article on MIT News.

David L. Chandler |  MIT News Office

Ultrathin, flexible photovoltaic cells from MIT research could find many new uses.

The MIT team has achieved the thinnest and lightest complete solar cells ever made, they say. To demonstrate just how thin and lightweight the cells are, the researchers draped a working cell on top of a soap bubble, without popping the bubble. Photo: Joel Jean and Anna Osherov.

Imagine solar cells so thin, flexible, and lightweight that they could be placed on almost any material or surface, including your hat, shirt, or smartphone, or even on a sheet of paper or a helium balloon.

Researchers at MIT have now demonstrated just such a technology: the thinnest, lightest solar cells ever produced. Though it may take years to develop into a commercial product, the laboratory proof-of-concept shows a new approach to making solar cells that could help power the next generation of portable electronic devices.

The new process is described in a paper by MIT professor Vladimir Bulović, research scientist Annie Wang, and doctoral student Joel Jean, in the journal Organic Electronics.

Bulović, MIT’s associate dean for innovation and the Fariborz Maseeh (1990) Professor of Emerging Technology, says the key to the new approach is to make the solar cell, the substrate that supports it, and a protective overcoating to shield it from the environment, all in one process. The substrate is made in place and never needs to be handled, cleaned, or removed from the vacuum during fabrication, thus minimizing exposure to dust or other contaminants that could degrade the cell’s performance.

“The innovative step is the realization that you can grow the substrate at the same time as you grow the device,” Bulović says.

In this initial proof-of-concept experiment, the team used a common flexible polymer called parylene as both the substrate and the overcoating, and an organic material called DBP as the primary light-absorbing layer. Parylene is a commercially available plastic coating used widely to protect implanted biomedical devices and printed circuit boards from environmental damage. The entire process takes place in a vacuum chamber at room temperature and without the use of any solvents, unlike conventional solar-cell manufacturing, which requires high temperatures and harsh chemicals. In this case, both the substrate and the solar cell are “grown” using established vapor deposition techniques.

One process, many materials

The team emphasizes that these particular choices of materials were just examples, and that it is the in-line substrate manufacturing process that is the key innovation. Different materials could be used for the substrate and encapsulation layers, and different types of thin-film solar cell materials, including quantum dots or perovskites, could be substituted for the organic layers used in initial tests.

But already, the team has achieved the thinnest and lightest complete solar cells ever made, they say. To demonstrate just how thin and lightweight the cells are, the researchers draped a working cell on top of a soap bubble, without popping the bubble. The researchers acknowledge that this cell may be too thin to be practical — “If you breathe too hard, you might blow it away,” says Jean — but parylene films of thicknesses of up to 80 microns can be deposited easily using commercial equipment, without losing the other benefits of in-line substrate formation.

A flexible parylene film, similar to kitchen cling-wrap but only one-tenth as thick, is first deposited on a sturdier carrier material – in this case, glass. Figuring out how to cleanly separate the thin material from the glass was a key challenge, explains Wang, who has spent many years working with parylene.

The researchers lift the entire parylene/solar cell/parylene stack off the carrier after the fabrication process is complete, using a frame made of flexible film. The final ultra-thin, flexible solar cells, including substrate and overcoating, are just one-fiftieth of the thickness of a human hair and one-thousandth of the thickness of equivalent cells on glass substrates — about two micrometers thick — yet they convert sunlight into electricity just as efficiently as their glass-based counterparts.

No miracles needed

“We put our carrier in a vacuum system, then we deposit everything else on top of it, and then peel the whole thing off,” explains Wang. Bulović says that like most new inventions, it all sounds very simple — once it’s been done. But actually developing the techniques to make the process work required years of effort.

While they used a glass carrier for their solar cells, Jean says “it could be something else. You could use almost any material,” since the processing takes place under such benign conditions. The substrate and solar cell could be deposited directly on fabric or paper, for example.

While the solar cell in this demonstration device is not especially efficient, because of its low weight, its power-to-weight ratio is among the highest ever achieved. That’s important for applications where weight is important, such as on spacecraft or on high-altitude helium balloons used for research. Whereas a typical silicon-based solar module, whose weight is dominated by a glass cover, may produce about 15 watts of power per kilogram of weight, the new cells have already demonstrated an output of 6 watts per gram — about 400 times higher.

“It could be so light that you don’t even know it’s there, on your shirt or on your notebook,” Bulović says. “These cells could simply be an add-on to existing structures.”

Still, this is early, laboratory-scale work, and developing it into a manufacturable product will take time, the team says. Yet while commercial success in the short term may be uncertain, this work could open up new applications for solar power in the long term. “We have a proof-of-concept that works,” Bulović says. The next question is, “How many miracles does it take to make it scalable? We think it’s a lot of hard work ahead, but likely no miracles needed.”

“This demonstration by the MIT team is almost an order of magnitude thinner and lighter” than the previous record holder, says Max Shtein, an associate professor of materials science and engineering, chemical engineering, and applied physics, at the University of Michigan, who was not involved in this work. As a result, he says, it “has tremendous implications for maximizing power-to-weight (important for aerospace applications, for example), and for the ability to simply laminate photovoltaic cells onto existing structures.”

“This is very high quality work,” Shtein adds, with a “creative concept, careful experimental set-up, very well written paper, and lots of good contextual information.” And, he says, “The overall recipe is simple enough that I could see scale-up as possible.”

The work was supported by Eni S.p.A. via the Eni-MIT Solar Frontiers Center, and by the National Science Foundation.

Read this article on MIT News.

Peter Dizikes |  MIT News Office

New book by Noam Chomsky and Robert Berwick explores how people acquired unique language skills.

“Why Only Us: Language and Evolution” (MIT Press), by Robert C. Berwick (top left) and Noam Chomsky (bottom left)

For many years, researchers tried to teach other kinds of animals some human language. Chimps, dolphins, gorillas — it didn’t seem to matter which animals they tried. Few experiments were regarded as success stories.

Small children, however, learn whichever language they are taught, and abundant evidence points toward the universality of human language. Platoons of linguists have detailed strong syntactical similarities among the world’s tongues. And biologists have begun to identify some of the genes involved in the development of speech and possibly language.

“Human language is a generative system that determines an infinite set of possible semantic objects,” says Noam Chomsky, Institute Professor and Professor of Linguistics Emeritus at MIT.

“People don’t realize how uniform the human population is,” adds Robert C. Berwick, a computer scientist at MIT. “We’re all very alike as humans, and this language capacity is incredibly uniform. If you take a baby from Southern Africa and put it in Beijing, they’ll speak Chinese.”

Now Berwick and Chomsky have collaborated on a new book on the topic, “Why Only Us?,” published on March 1 by the MIT Press, which explores the grand riddles of human language — what makes it unique, as well as where, when, why, and how humans acquired a distinctive, language capacity of nonpareil sophistication.

Out of Blombos?

The questions of when and where human language emerged are probably the simplest to grapple with. Like some other scholars, Berwick and Chomsky think the emergence of symbolic behavior is a guidepost indicating when human language developed. The Blombos cave artifacts in South Africa, comprising engravings and beads that are 80,000 years old, are a possible landmark. Modern humans arose about 200,000 years ago, so the development of our language capacity most likely falls in between those two points in time. Still, Chomsky notes how “thin the empirical record is” on this count.

Precisely what evolved, Berwick and Chomsky contend, is what Chomsky calls “Merge:” the human cognitive capacity to take any two things that we now recognize as sentence elements, and combine them into a new, more complex, hierarchically structured phrase.

“In its simplest terms, the Merge operation is just set formation,” the authors write. But if it sounds simple, this operation is precisely what allows human language to be infinite; there is absolutely no limit on the number of sentences we can form.

There are some other things that mark human language as distinct, so far as we know — for instance, our statements do not have to make reference to the external world. But the unbounded nature of language appears crucial at all times.

If so, how did such a powerful capability emerge in people? Berwick and Chomsky suggest it resulted from not a giant evolutionary leap but a modest evolutionary step that turned out to be very useful.

“What we’re arguing is that there was probably a very small change which had large effects,” Chomsky says. In the book as well, the authors suggest our language capacity was “the result of a minor mutation” in genomic terms that had far-reaching changes in our capacities.

“Evolution has [often] assembled lots of other parts that enable a whole host of other behaviors that you didn’t have before,” Berwick observes. “It [the language capacity] is standard in that kind of picture. It’s fully compatible with what a Darwinian might have thought.”

He adds: “We’re getting more and more of an understanding of the genomic basis for some of these traits, but it’s extremely challenging to work out.”

The real leap: intentional, conceptual thoughts

The hardest question to answer, it seems, is why humans should have a uniquely unbounded language. Or, to put it another way, what purpose did language play that made it a useful trait in evolutionary terms?

Berwick and Chomsky, following decades of work and theorizing by Chomsky, do not believe that language evolved primarily as a form of communication. Rather, it is an offshoot of the development of our cognitive capacities — an “inner mental tool,” as they write, at the interface of intentional thought and the ability to think conceptually.

In this sense, “Merge would be just like any other ‘internal’ trait that boosted selective advantage,” they write, something that would be helpful in planning, making inferences, and other basic capacities.

That said, Berwick and Chomsky readily acknowledge they do not possess a full hypothesis explaining how people developed the capacity for having those abstract conceptual thoughts in the first place.

“There is no explanation of where those come from,” Berwick says.

“The nature of elementary human concepts, such as table or chair, is unknown, and what’s striking about them is that they’re radically different than anything in the animal world,” Chomsky says. “It’s very different from other animals.”

“Why Only Us” has received advance praise from other scholars. Martin Nowak, a professor of mathematics and biology at Harvard University, calls it “captivating and a must for everyone interested in evolution and humans.”

And Berwick and Chomsky note that they hope to inspire further research, potentially integrating neuroscience to a growing extent, in addition to proposing answers to these scientific mysteries.

Or, as they write, “a vast array of language phenomena remain unexplained and even barely examined, but the picture sketched here seems to us the most plausible one we have, and one that offers many opportunities for fruitful research and inquiry.”

Read this article on MIT News.

Jennifer Chu | MIT News Office

New quantum computer, based on five atoms, factors numbers in a scalable way.

“Researchers have designed and built a quantum computer from five atoms in an ion trap. The computer uses laser pulses to carry out Shor’s algorithm on each atom, to correctly factor the number 15. Image: Jose-Luis Olivares/MIT

What are the prime factors, or multipliers, for the number 15? Most grade school students know the answer — 3 and 5 — by memory. A larger number, such as 91, may take some pen and paper. An even larger number, say with 232 digits, can (and has) taken scientists two years to factor, using hundreds of classical computers operating in parallel.

Because factoring large numbers is so devilishly hard, this “factoring problem” is the basis for many encryption schemes for protecting credit cards, state secrets, and other confidential data. It’s thought that a single quantum computer may easily crack this problem, by using hundreds of atoms, essentially in parallel, to quickly factor huge numbers.

In 1994, Peter Shor, the Morss Professor of Applied Mathematics at MIT, came up with a quantum algorithm that calculates the prime factors of a large number, vastly more efficiently than a classical computer. However, the algorithm’s success depends on a computer with a large number of quantum bits. While others have attempted to implement Shor’s algorithm in various quantum systems, none have been able to do so with more than a few quantum bits, in a scalable way.

Now, in a paper published today in the journal Science, researchers from MIT and the University of Innsbruck in Austria report that they have designed and built a quantum computer from five atoms in an ion trap. The computer uses laser pulses to carry out Shor’s algorithm on each atom, to correctly factor the number 15. The system is designed in such a way that more atoms and lasers can be added to build a bigger and faster quantum computer, able to factor much larger numbers. The results, they say, represent the first scalable implementation of Shor’s algorithm.

“We show that Shor’s algorithm, the most complex quantum algorithm known to date, is realizable in a way where, yes, all you have to do is go in the lab, apply more technology, and you should be able to make a bigger quantum computer,” says Isaac Chuang, professor of physics and professor of electrical engineering and computer science at MIT. “It might still cost an enormous amount of money to build — you won’t be building a quantum computer and putting it on your desktop anytime soon — but now it’s much more an engineering effort, and not a basic physics question.”

Seeing through the quantum forest

In classical computing, numbers are represented by either 0s or 1s, and calculations are carried out according to an algorithm’s “instructions,” which manipulate these 0s and 1s to transform an input to an output. In contrast, quantum computing relies on atomic-scale units, or “qubits,” that can be simultaneously 0 and 1 — a state known as a superposition. In this state, a single qubit can essentially carry out two separate streams of calculations in parallel, making computations far more efficient than a classical computer.

In 2001, Chuang, a pioneer in the field of quantum computing, designed a quantum computer based on one molecule that could be held in superposition and manipulated with nuclear magnetic resonance to factor the number 15. The results, which were published in Nature, represented the first experimental realization of Shor’s algorithm. But the system wasn’t scalable; it became more difficult to control the system as more atoms were added.

“Once you had too many atoms, it was like a big forest — it was very hard to control one atom from the next one,” Chuang says. “The difficulty is to implement [the algorithm] in a system that’s sufficiently isolated that it can stay quantum mechanical for long enough that you can actually have a chance to do the whole algorithm.”

“Straightforwardly scalable”

Chuang and his colleagues have now come up with a new, scalable quantum system for factoring numbers efficiently. While it typically takes about 12 qubits to factor the number 15, they found a way to shave the system down to five qubits, each represented by a single atom. Each atom can be held in a superposition of two different energy states simultaneously. The researchers use laser pulses to perform “logic gates,” or components of Shor’s algorithm, on four of the five atoms. The results are then stored, forwarded, extracted, and recycled via the fifth atom, thereby carrying out Shor’s algorithm in parallel, with fewer qubits than is typically required.

The team was able to keep the quantum system stable by holding the atoms in an ion trap, where they removed an electron from each atom, thereby charging it. They then held each atom in place with an electric field.

“That way, we know exactly where that atom is in space,” Chuang explains. “Then we do that with another atom, a few microns away — [a distance] about 100th the width of a human hair. By having a number of these atoms together, they can still interact with each other, because they’re charged. That interaction lets us perform logic gates, which allow us to realize the primitives of the Shor factoring algorithm. The gates we perform can work on any of these kinds of atoms, no matter how large we make the system.”

Chuang’s team first worked out the quantum design in principle. His colleagues at the University of Innsbruck then built an experimental apparatus based on his methodology. They directed the quantum system to factor the number 15 — the smallest number that can meaningfully demonstrate Shor’s algorithm. Without any prior knowledge of the answers, the system returned the correct factors, with a confidence exceeding 99 percent.

“In future generations, we foresee it being straightforwardly scalable, once the apparatus can trap more atoms and more laser beams can control the pulses,” Chuang says. “We see no physical reason why that is not going to be in the cards.”

Mark Ritter, senior manager of physical sciences at IBM, says the group’s method of recycling qubits reduces the resources required in the system by a factor of 3 — a significant though small step towards scaling up quantum computing.

“Improving the state-of-the-art by a factor of 3 is good,” says Ritter. But truly scaling the system “requires orders of magnitude more qubits, and these qubits must be shuttled around advanced traps with many thousands of simultaneous laser control pulses.”

If the team can successfully add more quantum components to the system, Ritter says it will have accomplished a long-unrealized feat.

“Shor's algorithm was the first non-trivial quantum algorithm showing a potential of ‘exponential’ speed-up over classical algorithms,” Ritter says. “It captured the imagination of many researchers who took notice of quantum computing because of its promise of truly remarkable algorithmic acceleration. Therefore, to implement Shor's algorithm is comparable to the ‘Hello, World’ of classical computing.”

What will all this eventually mean for encryption schemes of the future?

“Well, one thing is that if you are a nation state, you probably don’t want to publicly store your secrets using encryption that relies on factoring as a hard-to-invert problem,” Chuang says. “Because when these quantum computers start coming out, you’ll be able to go back and unencrypt all those old secrets.”

This research was supported, in part, by the Intelligence Advanced Research Project Activity (IARPA), and the MIT-Harvard Center for Ultracold Atoms, a National Science Foundation Physics Frontier Center.

Read this article on MIT News.

MIT News Office

Devadas one of four faculty awarded MIT's highest undergraduate teaching award.

Each year, the MacVicar Faculty Fellows Program recognizes professors who exhibit exceptional undergraduate teaching, educational innovation, and mentoring. The awardees this year are Srinivas Devadas, the Edwin Sibley Webster Professor of Electrical Engineering and Computer Science; Jeffrey Grossman, professor of materials science and engineering; Michael Sipser, dean of the School of Science and professor of mathematics; and Patricia Tang, an associate professor of music and theater arts.

Founded in 1992, the program was created to honor the legacy of Margaret MacVicar, an MIT alumna and professor of physical science who served as the Institute’s first dean for undergraduate education, from 1985 to 1990. MacVicar is credited with numerous far-reaching educational initiatives, including the Undergraduate Research Opportunities Program (UROP). Established in 1969 — when MacVicar was just 26 and in her first year on the MIT faculty — UROP has since been emulated worldwide.

The nomination process for the MacVicar awards is rigorous, requiring supporting letters and extensive documentation from several sources, including department heads, faculty, current students, and course evaluations. Provost Martin A. Schmidt selected the fellows, with input from an advisory committee of faculty and students chaired by Dean for Undergraduate Education Dennis M. Freeman. Fellows receive $10,000 annually for 10 years to support their undergraduate teaching. With the addition of the 2016 fellows, the program now sponsors 43 professors.

The Institute will honor the fellows and celebrate excellence in undergraduate education on MacVicar Day, Friday, March 11, with a symposium titled “From Hand to Mind: Advances in Evidence-based Teaching.” Freeman will introduce the 2016 fellows and moderate the panel. Speakers include Martin Culpepper, professor of mechanical engineering; Michael Cuthbert, associate professor of music; David Darmofal, professor of aeronautics and astronautics; Catherine Drennan, professor of biology; Robert Miller, professor of electrical engineering and computer science; and Janet Rankin, interim director of the Teaching and Learning Laboratory.

The symposium will take place from 2 to 4 p.m. in Bartos Theater (Room E15-070), followed by a reception honoring the new MacVicar Fellows from 4 to 5 p.m. in Bartos Lobby. The symposium and reception are open to the entire MIT community.

Srinivas Devedas

Devadas completed a BTech degree in electronics at the Indian Institute of Technology in Madras, India, and earned his MS and PhD in electrical engineering at the University of California at Berkeley. Devadas joined the MIT faculty in 1988, received tenure in 1995, and was promoted to full professor in 1999. He has also served in several leadership roles in the Department of Electrical Engineering and Computer Science (EECS), including associate head and interim head. In 2012, Devadas was named the Edwin Sibley Webster Professor of Electrical Engineering and Computer Science.

“I’m deeply honored to be selected as a MacVicar Faculty Fellow,” Devadas says. “I thank the EECS leadership over my 28 years at MIT for giving me great freedom in choosing my teaching duties and providing me opportunities to teach with, and learn from, literally dozens of my talented colleagues. I am deeply grateful to all my colleagues who I have partnered with in teaching our wonderful students.”

Devadas’ colleagues appreciate partnering with him, as well. “Srini not only was amazing in class, he also was a great mentor to us,” according to one nomination. “His enthusiasm for teaching was inspiring and contagious. He instilled in us and indeed, in all the teaching staff, the idea that one should tirelessly work to improve the material.”

Another colleague cited the many ways Devadas demonstrated “extreme” dedication in a new subject, 6.S04 (Fundamentals of Programming): personally proctoring make-up exams; meeting after hours with undergraduate lab assistants; helping students debug their code in laboratory sessions — “even though we have lab assistants for that!” — and changing his sabbatical plans so he could teach the new subject next year. “And I couldn’t omit the custom 6.S04 frisbees that he had printed, with guinea pigs on them to symbolize the pilot status of the subject, thrown to students who answer questions in lecture!”

Students perceive Devadas as a caring instructor with an inspiring sense of optimism. One noted that Devadas gave him some sage advice before he left MIT for graduate school, “advice that continues to guide me to this day: the best ideas come from a willingness to approach each problem with an enthusiastic outlook. Simply stated, Prof. Devadas is an incredibly positive person and his attitude resonates throughout his teaching every day.”

Another student was struck by the respect Devadas has for students’ needs, such as granting extensions for unforeseen circumstances like illnesses. Devadas’s response, the student wrote, would be “assuaging the student’s concerns and assuring them that they could finish the assignment at their convenience. At a school that is as high-pressure as MIT, such sensitivity goes a long way in ensuring that students don’t get overwhelmed by classwork.”

Read this article on MIT News

Mijal Tenenbaum | Department of Electrical Engineering and Computer Science

Senior Neerja Aggarwal reflects on her love for theater, ultimate Frisbee, and her journey to electrical engineering.

Since she was a little girl launching air-pressure rockets in her back yard in Houston, Texas, Neerja Aggarwal knew that she loved math and science. “I was always up to something,” the electrical engineering major recalls.

Aggarwal, now a senior at MIT, is still constantly up to something. She is the founder and leader of Voltage, an undergraduate electrical engineering club, part of the first group of students to graduate with a major in theater arts, and a member of sMITe, MIT’s Women’s Ultimate Frisbee team. This summer, she will start working on her MEng at MIT.

Engineering mindset

Aggarwal’s path to electrical engineering included a few stops along the way, as she discovered new disciplines and ways of approaching problems.

“I think the biggest thing I’ve learned at MIT is that you really don’t know what you don’t know,” she says. “I had no idea what else was out there.”

In high school, Aggarwal enrolled in a magnet program focused on medicine because she thought she wanted to be a doctor. Through the program she did rotations at Houston’s Anderson Cancer Center, and realized that she didn’t want to be a doctor — she wanted to build devices that doctors could use.

Going into the fall of her sophomore year, Aggarwal was ready to declare her major in materials science and engineering. She had done research in a chemistry lab at Rice University during high school, and loved the solid state chemistry class she took during her freshman year.

However, she kept thinking about an introductory electrical engineering and computer science class she had taken as a freshman. “I got to program a robot to navigate though a maze, and it was the first time that I thought about systems thinking, and how you can control a system,” she remembers. “And that’s what engineering is: it’s a way of thinking, it’s a mindset.”

Founding Voltage

The experience prompted her to declare electrical engineering as her major. Aggarwal is one of a growing number of students opting to study electrical engineering and computer science (EECS) — enrollment in the department has roughly doubled since 2010–2011 from 637 students to 1204 students in the current academic year.

Aggarwal was happy with her decision to change majors, but missed some of the benefits of a small academic department she had experienced in the Department of Materials Science.

“I really liked the small major feeling of materials science, and I wanted to bring that to electrical engineering,” she says. Aggarwal decided to start Voltage, a club for undergraduate electrical engineers.

Voltage aims to bring students, alumni, and faculty together for interactions around research and coursework. A subcommittee of the MIT IEEE/ACM Club, Voltage started with study breaks where students could meet, find out who was in their classes, and learn about courses. Since then they have planned bigger events, including two research expos where faculty showcased their work to help students find research opportunities.

Aggarwal and Voltage have also been working to inspire more students to study electrical engineering. In Fall 2015 they hosted the Electrical Engineering Expo with EECS, which connected students with electrical engineering internships and research opportunities.

Practical knowledge

The summer before her senior year, Aggarwal worked at FormLabs, an MIT spinout in Somerville specializing in 3-D printing. “The definition of my job was ‘fix things that are broken and make things work better’,” she explains.

In September 2015, FormLabs launched Form 2, the printer that Aggarwal worked on. Much of her internship focused on troubleshooting. “My biggest contribution was actually that I used what I learned in a lot of my signals and systems courses at MIT to program the heater for the printer,” she says.

During her junior year, Aggarwal worked on a research project through EECS’s SuperUROP program with Rajeev Ram, professor of electrical engineering, in his Physical Optics and Electronics (POE) Group. The project she worked on was aimed at building a wearable optical health monitor.

Aggarwal says that that her work at FormLabs and her research with Ram’s group helped her grow into a more confident engineer. She remembers she went into her SuperUROP without knowing anything about how to build a high power wearable laser, the focus of her research project. However, with Ram’s encouragement, she quickly realized that learning how to solve new problems was the point of SuperUROP and research itself.

Both of these experiences showed her the importance of being ready to learn and figure out new ways to think. “Throughout my SuperUROP experience and during my time at FormLabs, I had to learn everything on the job. And that’s what I realized engineering is. It’s being able to learn what you need to, to get the job done.”

Rediscovering theater

Another thing that Aggarwal learned from her SuperUROP is the importance of diverse interests. “I do theater, I do electrical engineering, I do design, I do sports. I really appreciated that Professor Ram always encouraged me to keep pursuing all of my interests, and never told me to limit myself to one thing.”

Aggarwal grew up acting in school plays, but had stopped during high school to focus on science. At MIT, she discovered the student theater community and immediately decided to get involved.

She directed Oscar Wilde’s The Importance of Being Earnest with the Dramashop during her junior fall. A year later Aggarwal directed Now Then Again, her first full-length production, with The Experimental Theater Company, MIT’s newest student-run theater group.

She describes this experience as “the hardest thing I’ve ever done at MIT.” She still loves acting, but she loves directing even more. “I have a passion for it, just like I have a passion for science,” she says.

Aggarwal enjoyed her SuperUROP research so much that she has decided to stay at MIT next year to pursue her MEng with Ram’s group. She will simultaneously be working on a theater arts thesis.

Aggarwal says, with no hesitation, that her favorite part of MIT is the people. “You don’t get to find such a group of people anywhere else in the world. Everybody here is so passionate about something, and it’s very obvious on their faces, and in how they talk about things. That really is the best part.”

An all-star panel of women shared their experiences as innovators and entrepreneurs at EECS's StartMIT class in January.

Moderated by Jesse Draper, creator and host of the Emmy-nominated “Valley Girl Show,” the January 19 panel included Susan Hockfield, president emerita of MIT; Helen Greiner ’89, SM ’90 CEO and founder of CyPhy Works and co-founder of iRobot; Payal Kadakia ‘05, CEO and co-founder of ClassPass; and Dina Katabi, the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science, with opening remarks from Anantha Chandrakasan, EECS department head and the Vannevar Bush Professor of Electrical Engineering and Computer Science, and closing remarks from Joi Ito, director of the MIT Media Lab.

Learn more at http://www.eecs.mit.edu/news-events/media/empowering-innovation

Department of Electrical Engineering and Computer Science

Vivienne Sze recognized as outstanding junior faculty.

Vivienne Sze, the Emanuel E. Landsman (1958) Career Development Assistant Professor, has received the 2016 3M Non-Tenured Faculty Award. Presented by 3M’s Research and Development Community in partnership with 3Mgives, the award recognizes outstanding new faculty who excel in research, experience, and academic leadership. Sze received the award in support of her work on next-generation embedded vision systems.

Her research focuses on joint design of algorithms, architectures and circuits to build energy efficient and high performance systems. Her work on implementation-friendly video compression algorithms was used in the development of the latest video coding standard HEVC/H.265, enabling it to deliver better compression than previous standards, while still achieving high processing speeds and low hardware cost. She aims to develop energy-aware algorithms and efficient architectures for various energy-constrained applications including portable multimedia, health monitoring and distributed sensing.

Sze joined MIT in August 2013. She is a core member of the Microsystems Technology Laboratories (MTL) and a principal investigator at the Research Laboratory of Electronics (RLE). Sze received the B.A.Sc. degree in Electrical Engineering from the University of Toronto in 2004, and the S.M. and Ph.D. degree in Electrical Engineering and Computer Science from MIT in 2006 and 2010, respectively. From September 2010 to July 2013, she was a Member of the Technical Staff in the Systems and Applications R&D Center at Texas Instruments.

She has received various awards including the Jin-Au Kong Outstanding Doctoral Thesis Prize in 2011, the 2007 DAC/ISSCC Student Design Contest Award, the 2008 A-SSCC Outstanding Design Award, the Natural Sciences and Engineering Research Council of Canada (NSERC) Julie Payette fellowship in 2004, the NSERC Postgraduate Scholarships in 2005 and 2007, and the Texas Instruments Graduate Woman’s Fellowship for Leadership in Microelectronics in 2008. In 2012, she was selected by IEEE-USA as one of the “New Faces of Engineering”.

Patsy Sampson | Department of Electrical Engineering and Computer Science

Department of Electrical Engineering and Computer Science pilots new ways to build leadership and teamwork skills for its postdocs.

Clockwise from top left: Ziawasch Abedjan, Sara Cleto, Zhizhou Zhang, and Shahin Kamali. Photos: Patsy Sampson

“What I liked the most about the Postdoc6 Workshop was the playback theater!” says Zizhuo Zhang, postdoctoral associate in the Computational Biology Group with Manolis Kellis, professor of computer science.

Improv theater is just one of the tools the Department of Electrical Engineering and Computer Science (EECS) is using to help its postdocs beef up their leadership and teamwork skills as part of an initiative called Postdoc6. The program aims to knit together the often disperse postdoc community, while helping postdoctoral associates and fellows train for their future careers.

With over 200 postdoctoral associates in its labs, EECS under the direction of Department Head Anantha Chandrakasan, created Postdoc6 in late 2012. “Postdocs are a key part of EECS,” notes Chandrakasan, the Vannevar Bush Professor of Electrical Engineering and Computer Science. “We are committed to ensuring that they have access to the resources they need to succeed in their research and achieve their goals. With Postdoc6, we hope to build a vibrant postdoc community — opening the door to the many opportunities at MIT.”

Increasing collaboration

In mid-January, Zhang and fifteen other EECS postdocs participated in a two-day leadership workshop held at MIT’s Endicott House. The sixteen EECS postdocs received training based on transactional analysis to build communication and self-management skills.

Transactional analysis is a system of popular psychology based on the study of relationships as individuals shift among the roles of parent, child, and adult. This method can be mastered in two- to three-day sessions with the goal of building trust and openness.

The workshop also featured training for follow-up peer groups to provide supportive networks that last long after the workshop ends.

“Since WWII, science has become increasingly collaborative,” says Nir Shavit, professor of electrical engineering and computer science and Postdoc6 faculty coordinator. “Now, authorship on major papers includes multiple authors and sometimes research groups. There is more and more need for collaborative and interdisciplinary work,” he adds. “The goal is to build a new generation of scientists that are more aware of human interactions, allowing for real communication in the labs.”

Learning to Lead

Shahin Kamali, postdoctoral researcher in computer science under Charles Leiserson, the Edwin Sibley Webster Professor of Computer Science, was upbeat following the experience. “Now I feel that MIT cares about me,” he says. With plans to join a peer group, Kamali says that the workshop also revealed his interest in helping people understand each other.

Judith Birkenfeld, postdoctoral associate in the Madrid M+Vision Consortium at the Research Laboratory of Electronics and Research Fellow at Brigham and Women’s Hospital and Harvard Medical School, says she was impressed with the takeaway. “The Monday after the workshop, I could go to my office and apply what I learned right away,” she says. She is also impressed by how the facilitators at the workshop were able to engage all participants. “They recognized your strengths and shortcomings quickly, and gave concrete advice on how to improve. Through subtle suggestions, I was amazed to see how people could appear completely different by applying minimal changes.”

“You are not alone in struggling through a project,” says Dan Congreve, who earned his PhD under Mark Baldo, professor of electrical engineering. Congreve will be completing his postdoctoral position in the Research Laboratory of Electronics in several months to lead a research group as a Fellow in the Rowland Institute at Harvard. He joined the workshop knowing the issues of communication to lead a group were universal. “Finding a way to effectively say ‘This is why I am frustrated’,” he says, “was a major takeaway for any member of a team.”

“We didn't focus on playing a ‘boss’,” says Ziawasch Abedjan, postdoc with Michael Stonebraker, adjunct professor in computer science and 2014 Turing Award winner. “Instead,” Abedjan adds, “it was implied to inherently lead by acting as an adult who treats the other person as an adult. We wanted to avoid or break the inner-attitudes that resemble the roles of parent and child.”

Nir Shavit led an additional part of the workshop known as playback theater — using improv theater techniques to train groups of scientists to enhance collaborative and communication skills. “The playback process,” says Shavit, “is actually a microcosm of what’s happening in the lab.”

“Performing spontaneously is not a native skill for most serious researchers including postdocs,” notes Sara Cleto, postdoc in the Synthetic Biology Group with Tim Lu, associate professor of electrical engineering. “But, we could be as goofy as we wanted, since we felt no one would sit and judge us. We all participated.”

Recognizing the challenge to express and manage emotion as a tool critical for focused research, Zhizhuo Zhang found the workshop including the improve theater was eye-opening. “In order to accomplish high impact research, it is necessary for a researcher to understand self motivation — to open your heart,” he says. “How to express yourself, to perform,” he adds, “was not only important in communicating your work, but it was the fun part of the workshop.”

Building Community

“Postdocs explore new areas as they launch their professional careers and create vital connections among groups within MIT and in other institutions,” says Muriel Médard, the Cecil H. Green Professor in EECS at MIT. As head of the Network Coding and Reliable Communication Group, Médard also notes that the contribution of postdocs in research groups in her field is growing in importance and complexity. “I see with pleasure that postdocs are increasingly collaborating with each other, creating a vibrant community of the next generation of researchers.”

As the 16 workshop participants resumed their daily research paths with new tools to apply to the challenges ahead, Postdoc6’s steering committee members launched a new social activity for all EECS postdocs. Tea Time, a weekly tea and cookie social gathering, will provide EECS postdocs an opportunity for community in a relaxed environment.

“The focus of Postdoc6 planning is all about forming community and making postdocs’ time at MIT better through workshops and community meetings such as Tea Time,” says Shavit. He adds that no matter what the mechanism to build community, EECS is seeking to build a positive experience for its postdocs.

Catherine Curro Caruso |  MIT News correspondent

Senior YQ Lu finds new ways to combine math and paper art, shares his passion for both.

“I enjoy coding, programming, making pretty things,” YQ Lu says. “But then I also enjoy making things that I can hold with my hand and examine.” Photo: M. Scott Brauer

Yongquan “YQ” Lu is perfectly at home surrounded by sheets of paper. For him, these rectangles represent endless possibilities for math-based art, ranging from exquisite, three-dimensional origami forms created without a single cut, to intricate, multicolored geometric designs assembled with laser-cut paper strips. Lu, a double major in mathematics and electrical engineering and in computer science, is an independent thinker who enjoys following his own path. During his time at MIT, this has meant combining his two loves, math and paper art, in increasingly interesting ways, while staying committed to community outreach.

Lu, who grew up in Singapore, was interested in math and paper art from an early age. During high school, he took six weeks during the school year to attend Canada/USA Mathcamp, where he was surrounded by students, counselors, and teachers who shared his excitement for math.

The experience motivated Lu to apply to MIT, but before college he was required to serve for two years in the Singapore military. Lu, who ended up working in an administrative office, approached his service with a positive mindset and a determination to learn as much as he could.

“Some of my friends hate doing national service because it feels like you have this bright future in front of you, but then everything is on hold,” he says. “But it was a really good experience for me. I think of that as almost a two-year-long internship.”

Art, paper, and computer science

Lu came to MIT expecting to study theoretical math, but after two years of coursework in math, he found himself increasingly drawn to computer science and eventually decided to add on a second major.

“I love all the math that I'm doing,” he explains. “But I also want to do something a lot more constructive, and be able to build things and apply my technical knowledge in a hands-on manner.”

Lu has been interested in origami since middle school, but during his military service he started creating paper sliceform art, which involves cutting out thin strips of brightly colored paper with perpendicular slits in them, and interlocking the strips to assemble complex geometric patterns.

Lu’s early attempts were done entirely by hand, but as he acquired more computer science knowledge, he realized that he might be able to write a program to automate the process. Lu teamed up with Erik Demaine, who is a professor studying algorithms at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) and an accomplished paper artist himself.

With Demaine’s guidance, Lu developed a Web program that allows sliceform artists to choose from a number of geometric designs and adjust their design based on different parameters. The paper strips are then laser cut and assembled by the artist. Lu created an open-source website that makes his program available to the public, and he has gotten feedback that some people are even using it as a template for other types of art. Lu also presented a paper on his work at Bridges, an annual conference that focuses on math, art, and culture, an experience that he really enjoyed.

“I went there and I immediately thought, oh I found my people,” he explains. “There were college professors but also a full-time math artist and people who were doing other kinds of paper art.”

Spreading math and art

Lu is currently the president of OrigaMIT, MIT’s origami group. The organization has 10 undergraduate officers, and while they enjoy creating complex origami pieces, their bigger mission is origami-based community outreach. They hold weekly folding workshops that are open to the public and organize programs for MIT students during finals period. Their biggest outreach event every year is an annual convention, a full day of origami lectures and workshops that draws people from all over the country. More than 200 people attended the last convention.

“It's the one thing I look forward to the most every year,” Lu says. “We wake up at six in the morning to set up everything, and then we have all these people from all over descending on MIT and folding.”

For the past three years, Lu has also been heavily involved in Amphibious Achievement, an athletic and academic mentorship program for high school students in the Boston area. Every Sunday in the fall and spring, the students, or Achievers, come to MIT and receive instruction in different sports and academic areas. Lu, who was the co-director of academics his sophomore and junior years, ran the classroom part of the program, where he tried to find innovative ways to teach the kids math that didn’t require drilling them on sums or having them memorize formulas.

“We did a lot of things by discovery that weren't necessarily high school math, but were logic puzzles and puzzles with games,” says Lu. “A lot of the Achievers were initially dubious about this and were very used to being handed an answer and a formula to apply, so it was very gratifying to see them actually enjoy the process and think and struggle through a puzzle which they had never done before.”

Beyond MIT

Lu, who graduates this May, has accepted a job at MarkForged, a 3-D printing company that was founded by an MIT graduate and has developed techniques for 3-D printing with strong, lightweight materials such as Kevlar, carbon fiber, and fiberglass. Lu likes the interdisciplinary nature of the company, where he will be designing software and working closely with the hardware specialists that are using his programs. For him, the company is a happy medium between computer science and hands-on design

“I enjoy coding, programming, making pretty things,” he says, “But then I also enjoy making things that I can hold with my hand and examine.”

Lu, who to his knowledge is currently one of only a handful of paper sliceform artists, also plans to continue creating and exhibiting his artwork in his spare time. He doesn’t envision selling his work, but someday he would like to sell kits of instructions and precut paper strips so that people can assemble their own paper sliceform creations.

“There are people that would spend 20 hours working on a jigsaw puzzle, and this is very much in the same vein of a puzzle,” he says. “I think there are people that would be interested.”

But for right now, Lu is intent on staying in the moment and taking things one step at a time.

“If you asked me four years ago, I wouldn't have been able to predict where I am now,” he says. “So instead of looking forward, my main priority is to be around some really smart people and learn from them.”

Read this article on MIT News.

MIT News Office

Collaboration has created a pioneering new graduate university in Russia, says project director.

“Now, just over four years since its inception, Skoltech is an active graduate university with outstanding faculty, students, research and education programs, and a fledgling campus,” says Bruce Tidor (pictured). Courtesy of Bruce Tidor

In 2011 MIT launched a multiyear collaboration to assist Russian partners in building a new university on the outskirts of Moscow, the Skolkovo Institute of Science and Technology (Skoltech).

MIT faculty have participated in developing Skoltech’s educational and research programs, which have a strong emphasis on innovation and entrepreneurship. The MIT Skoltech Initiative, the operational arm of the collaboration, has thus far involved more than 150 MIT faculty and 60 staff. The first phase of the collaboration has been completed, and the second phase launched on March 1.

Bruce Tidor, a professor of biological engineering and computer science and the director of the MIT Skoltech Initiative, spoke with MIT News about the collaboration’s progress to date, the impact of the program, and what the future holds.

Q: What has been achieved at Skolkovo through MIT’s involvement?

A: When MIT began work on this project there was no university, there were no buildings, there was no faculty, and there were no students. There was only a handful of highly motivated people guided by a very ambitious plan. Skoltech was intended as the academic engine for a 400-hectare hub of innovation and economic development — the Skolkovo Innovation Center — launched by Russia’s president at the time, Dmitry Medvedev.

Now, just over four years since its inception, Skoltech is an active graduate university with outstanding faculty, students, research and education programs, and a fledgling campus. The inaugural class of approximately 50 MS students graduated in June 2015. Multidisciplinary research centers have been established in frontier areas related to information technology, energy, space, and biomedicine, and an innovation ecosystem nurtures capabilities designed to promote entrepreneurship. The major administrative offices and functions are now in place, with staff hired and trained, and policies and procedures in effect.

Much of this has been accomplished with the active assistance and engagement of MIT faculty and staff. An especially important role has been played by MIT Professor Edward Crawley, Skoltech’s founding president. (Professor Crawley recently stepped down as president, and the appointment of his successor, Alexander Kuleshov, was announced in February.)

Q: How has MIT participated in the Skoltech project?

A: Collaborative research projects at MIT funded by Skolkovo have connected MIT researchers to important problems and to faculty and other researchers in Russia, and these relationships will likely persist well beyond the terms of the projects. Over 100 talented students from Skoltech have participated in classwork and laboratory research at MIT, sharing their insights and perspectives with our students, staff, and faculty.

Course development for 33 subjects undertaken for Skoltech has also enriched MIT’s curriculum. The MISTI-Russia program, expanded student internships in Russia, and the resumption of Russian language teaching at MIT after a long hiatus are all direct outcomes of the collaboration. Learning in areas such as faculty recruitment and development, administrator capacity enhancement, and innovation and entrepreneurship programming can be applied both at MIT and in future major institutional partnerships. Crucial support has also been provided for MIT core activities (including named professorships and endowed chairs). In many ways, large and small, the collaboration to build Skoltech has had substantial beneficial impact at MIT.

Q: What is planned for the future relationship between MIT and Skoltech?

A: MIT’s four-year Skoltech collaboration agreement concluded at the end of February 2016. Skoltech now has sufficient academic and administrative personnel and structure to manage its own operations. MIT, the Skolkovo Foundation, and Skoltech have therefore agreed on a second phase of the relationship with reduced scale and a narrower focus on faculty-driven, collaborative projects.

The new agreement, which entered into effect on March 1 and will continue for another three years, features a small, focused set of core activities linked to a flexible capability to set up additional research and educational activities in the future. Three existing research center grants are ongoing, one in functional genomics led at MIT by Professor Daniel Anderson, a second in electrochemical energy storage led by MIT Professor Carl Thompson, and a third in energy systems and power distribution led by MIT Professor Konstantin Turitsyn.

Skoltech will continue to reach out to faculty at MIT to initiate new research activities, each of which will be governed by separate contracts managed within the relevant MIT department, laboratory, or center. To manage the new phase of the relationship, MIT will provide a small central service unit and will establish a faculty coordinating committee under the leadership of Associate Provost for International Activities Richard Lester. To reflect its reduced size and narrower focus, the relationship is being renamed the MIT Skoltech Program.

Read this article on MIT News.

Larry Hardesty | MIT News Office

System lets Web users share aspects of their browsing history with friends, researchers.

“Data has traditionally been used by anyone from corporations to the government,” Mor Naaman says. “But the goal of this system is to make the data more useful for the individuals themselves, to give them more control, and to make it more useful to communities.” Image: Jose-Luis Olivares/MIT

Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have developed a new system that allows Web users to share self-selected aspects of their online activity with their friends and the general public. The hope is to give users themselves, as well as academics and other scientists conducting research in the public interest, access to the same type of browsing data that big Web companies currently collect and mine to better target products to individual consumers.

The researchers also hope that systems like theirs could encourage changes in the regulatory environment that would give Web users more control over which of their data are collected and how they’re used.

They describe the new system, which they’ve dubbed Eyebrowse, in a paper presented last week at the Association for Computing Machinery’s conference on Computer-Supported Cooperative Work and Social Computing, in San Francisco. The paper also describes the results of a survey of potential end users, which helped guide the system’s design, and of a field trial of the system. The findings suggest that Web users could, indeed, find it worth their while to share data about their online activities.

“We thought of a pretty long list of specific applications of this data that were useful to the end user,” says David Karger, an MIT professor of electrical engineering whose group began developing the system in 2010. “Awareness of where your friends are, the ability to run into them, the ability to go somewhere and discover that they were there before, and you may want to talk to them about this thing that you both saw. There’s the ability to discover what’s popular, in a very broad way. There’s collaborative filtering.”

“Then there’s global analytics,” Karger continues. “Google has this interesting 50,000-foot view of the Internet, because they know all the clicks. Most people don’t. There are lots of interesting questions about social dynamics. What are Democrats reading? You can’t answer that question right now. There are things that the population as a whole would be interested in knowing, and also things that scholars would be interested in knowing.”

“The trackers don’t give us a choice about what gets tracked,” he adds. “And I’d really like to demonstrate that giving people a choice has positive benefits. And maybe someday that will turn into legislation that says that people have the right to decide whether they get tracked or not, in certain circumstances. If people do buy into voluntary tracking, then maybe we don’t need involuntary tracking, and that would be pretty wonderful.”

“But of course, it only works if people want to do it. So a lot of this paper was about understanding whether and how people want to do this.”

Managed browsing

For the past year and half, the lead researcher on the Eyebrowse project has been Amy Zhang, an MIT graduate student in electrical engineering and computer science and first author on the new paper. She’s joined by Karger, who is her thesis advisor, and by Joshua Blum, who received his master’s in electrical engineering and computer science from MIT this year.

Eyebrowse currently consists of two components: a website and an extension to Google’s Chrome Web browser. Installing the extension involves two mouse clicks and takes seconds. Thereafter, anytime the user visits a Web page, clicking the Eyebrowse icon on the browser task bar will pull down a window offering an array of features.

One is the opportunity to add the site’s domain name to the user’s “whitelist.” As long as the Eyebrowse extension is turned on, the system will record the user’s visits to pages on whitelisted sites. But the pull-down window also features a switch for turning Eyebrowse off, for private browsing. (The Eyebrowse icon, an open eye, “closes” when the system is off.) Similarly, it offers a button for reporting visits to sites not on the whitelist.

The pull-down window also lists which members of the Eyebrowse community have visited the page and when, any annotations that they have made to the page, a field that allows the user to make his or her own annotations, and a chat window for Eyebrowse users.

Viewing data

The Eyebrowse website looks much like Facebook’s “news feed,” with a list of pages recently visited by members of the Eyebrowse community. The user can toggle between two versions of the list, one that includes all Eyebrowse members and one that includes only those actively “followed” by the user.

By default, the lists are sorted according to a simple ranking algorithm, which factors in the number of people that have visited each page, the amount of time they spent there, and the time of the last visit. But the lists can also be sorted according to each of those factors independently.

The website also provides visualization tools that allow users to view both their own browsing histories and those of the Eyebrowse community at large, as graphs, pie charts, and “word clouds” that represent the frequency with which particular words turn up in the sites visited by Eyebrowse users.

The field trial involved 24 users, who used the system for varying lengths of time, from a week to almost three months. Most shared between 10 and 25 links a day, but participants whose friends were also using the system tended to share more, as many as 60 or even 80 links a day.

“What we have built in terms of potential applications only scratches the surface of what is possible with this data,” Zhang says. “That’s why the data that people have contributed to Eyebrowse is available in an API [application program interface] on the website for anyone to build on top of or analyze.”

“Data has traditionally been used by anyone from corporations to the government,” says Mor Naaman, an associate professor of information science at Cornell University’s new Cornell Tech campus in New York City. “But the goal of this system is to make the data more useful for the individuals themselves, to give them more control, and to make it more useful to communities.”

“In previous research, we found that it’s difficult for users to cognitively manage all these different privacy settings, remembering what they mean and when and how to apply them,” Naaman cautions. “I think that will be something to address going forward. But there is potential, and I hope to eventually see it in a commercial product.”

Read this article on MIT News.

Fourth-year EECS student Julian Straub studies how robots can better understand their complex surroundings.