Nicolas Pinto hopes his research will spur the successful deployment of a prototype robotic car—a car that will avoid obstacles and potential crash situations, communicate with other cars, and essentially do most everything a human driver can do, better. And, of course, if a robotic car is theoretically in the cards, a humanoid robot that can also “see” can’t be far behind.

Sporting dual Master of Science degrees in computer science, the native of France is one of the driving forces behind what can best be described as a “computers with vision” movement at the Massachusetts Institute of Technology’s McGovern Institute for Brain Research. In MIT’s DiCarlo Lab, which partners with Cox Lab at Harvard’s Rowland Institute, the stated goal is to gain a better “understanding of the neuronal computations that support the brain’s remarkable ability to recognize visual objects.” The labs' scientists bust their humps to decipher how the brain—which, in many ways, is the most efficient, most powerful computing device on the planet— handles images. With this knowledge, a little luck, and a ton of work, they seek to successfully reverse engineer those processes synthetically.

None of this is easy. In the natural world, visual recognition of even the simplest object is a concept that remains mysterious and incredibly difficult to mimic. The folks at DiCarlo’s and Cox’s labs believe the brain takes in pixel-based images of everything the eyes see and then converts them into patterns of neuronal activity that emphasize object identity and dispense with visual clutter. They also feel that by understanding this natural recognition process, they’ll not only gain tremendous insight into it but also open a pathway into the inner workings of memory and thought. “Half of our brain is visually driven,” says Pinto, “When you think about something, you think visually. Our research and interest goes far beyond what our eyes see atany given moment.” The brain is crammed with tens of billions of neurons, all of which are linked through an intricate web with other neurons to do different things. Clearly, deciphering—and, in a sense—copying all that transpires within such a hyper-sophisticated, hyper-versatile network is like finding the proverbial needle in the haystack. In the wet lab, with human or animal subjects, it would literally take forever. But with the aid of very fast computers and biologically inspired algorithms, the Cox-DiCarlo team can more efficiently discover and simulate the processes.

The Cox-DiCarlo team prefers its computational equipment to “think” in the same crazily fast way the brain does—via parallel processing. And to that end, both labs feature gear that leans heavily on donated NVIDIA GPUs.

Indeed, the labs are strewn with NVIDIA GT200s, GT280s, and 9800GX2s. High-powered NVIDIA Teslas are there, too, as is one particularly noteworthy machine (dubbed the “Monster”) that sports two quad-core CPUs and no less than 16 GPUs (in the form of NVIDIA 9800GX2s).

“For us, GPUs are completely game-changing. They give us a drastic edge on the field.” Such remarks would seem to support Pinto’s claim that the team has experienced a “one hundred- fold speed-up relative to conventional systems” following the introduction of the NVIDIA GPUs.

In short, this speed boost enables the team to discover new vision models that would otherwise have been missed—and that’s good news for all of us in the long run.