The life of a LIDS student may not be glamorous, but their research often is—at least in terms of its broader implications for the world. Erik Sudderth is one of many talented young researchers at the lab whose work is groundbreaking in both scope and possibility. Erik hails from Cupertino, California, the heart of the Silicon Valley. “At my high school,” he says with a shrug and a smile, “everyone’s dad was an engineer.” After completing his undergraduate degree at UC San Diego, Erik pulled up his California roots, at least temporarily, for New England winters and a university community he calls simply “great.”

Erik’s dissertation research straddles the related fields of computer vision and machine learning. At its “most ambitious,” he explains, computer vision attempts to get computers to “see” like human beings, that is, recognize objects in an image the way you or I might. “The idea is that images are a really rich source of information. You put pictures up and you can tell me an amazing number of things: what the objects are, what their shape is, the material they’re made out of.” But there are many different levels of recognition, which is where machine learning, the more theoretical side of Erik’s work, comes in. It provides the mathematical tools with which he tackles computer vision problems. “Try to think about writing down from first principles what defines a chair. It’s kind of hard to capture all of that. But what I can do is get lots of pictures of all kinds of chairs, and give them to a machine learning algorithm which will then try to turn that into a recognition system for computer vision.”

Computer vision is a field perhaps best known to the public for recent advances in the surveillance world, including face recognition technology. “Surveillance is an obvious application [for computer vision], but it’s not one I’m personally wild about,” Erik admits. “It has all these Big Brother connotations.” He is quick to acknowledge that there are plenty of places, such as airports, where more efficient surveillance systems are necessary. “I don’t see any real difference between having a security guard looking at a bank of several dozen monitors versus the way a computer vision system would work in that case.” But take the idea further and a camera on every street corner may not be too farfetched, at which point the implications are troubling. As Erik concludes, “the potentially scary thing is that with advances, you won’t be anonymous as you’re moving about.”

There are many other applications for computer vision research, notably in the special effects world, where filmmakers currently have to go “to crazy lengths,” as Erik says, to manipulate images. Electronics companies are also interested in the development of “natural interfaces” with products like home televisions or computers. “This is the idea of being able to use images and video, and maybe speech, to let people interact with their computers in a more natural way, rather than having to use a keyboard and mouse.” Imagine being able to speak to your television, or have it recognize gestures you make. Of course, we’re not there yet, and Erik muses that “people don’t yet know exactly how audio and video should be used. There’s a sense that more natural interfaces would be useful, but the details haven’t yet been worked out.”

Erik didn’t immediately start doing this kind of research when he entered MIT. His first three years were funded by a fellowship that allowed him to explore a variety of possibilities, giving him freedom he describes as both good and bad. “It’s good because you get to pick something you’re interested in. But there are times when it’s a little daunting, because you’ve got to find something concrete that you can really make progress on.” He had worked in a computer vision lab during his undergrad years and knew he was interested in it, but his Master’s thesis “didn’t have anything to do with computer vision.” The diversity of study at the Institute and the flexibility and support of his advisers brought him around to his current PhD research, combining his previous interests with new problems. Erik’s most recent work, for which he presented findings at the LIDS Student Conference this January, is “hand tracking.” He starts with a three dimensional geometric model of the hand, takes a video of someone moving their hand, and from that can “automatically extract what the hand is doing in 3D.”

As he looks back over his LIDS experience, Erik is enthusiastic about precisely this kind of intellectual freedom and exposure to new ideas that an MIT education affords. Among his most memorable moments at LIDS didn’t happen in Cambridge at all, but in Italy, at a physics and machine learning conference that he attended with Professor David Forney. That experience fueled his ideas in the years that followed. Overall, he says, “[MIT] is a great environment in that you’re exposed to lots of these connections. Even if you don’t work on them directly, they float around in the back of your mind, where they may be useful some time in the future.”

That future brings Erik, like most of his fellow LIDS students, three main options upon graduation: postdoctoral research, teaching, or a position in a research lab. He isn’t sure which one is most appealing yet, though he affirms that the world of academia is where he wants to be in the long term. Because his wife will finish her graduate degree in ecology at Harvard, other considerations enter the picture: “We need to find a place where we can both find things we want to do. Because of those concerns, I’m flexible.” No matter what the future holds, however, Erik knows he’ll miss the community at LIDS and MIT. “There’s such an interesting group of people who are really excited about what they’re doing. If you ask them about some problem you’re having, they’ll probably have a bunch of clever ideas for you. And that’s really great.”

Erik Sudderth is a member of Professor Alan Willsky’s Stochastic Systems Group and also works with Professor William Freeman, of CSAIL. For more information, you can access his website at