MIT chip models human brain synapse

Forget artificial intelligence; researchers at MIT say they've figured out how to mimic the way a real brain works.

Forget artificial intelligence; researchers at MIT say they've figured out how to mimic the real deal.

The goal is to replicate how the brain learns new tasks. To do that, an MIT group has created a 400-transistor processor that's designed to model a single brain synapse.

The two-neuron MIT chip can't scale to model a whole working brain: Human brains contain 100 billion neurons, each of which interacts with many others. But unlike past systems that were designed to "think" like computers -- they were either on or off -- the MIT chip allows current to flow in a stream that ebbs and flows, just like ions do in an actual brain cell.

The researchers can "capture each and every ionic process that's going on in a neuron," said Chi-Sang Poon, the team's principal research scientist, in a story on the MIT website.
The ability to mimic different types of processes is important. Researchers hope to use the chip to build systems that model specific neural functions, such as the visual processing system -- a model of which could be used to build, say, an artificial retina. These systems might be faster than traditional computers -- faster, even, than real human brains. Uh-oh.

Read more about hardware in Computerworld's Hardware Topic Center.