Cat Brain Inspires Computer Design

Feedloader (Clickability)

A University of Michigan computer engineer, Wei Lu, has set out to develop a supercomputer the size of a 2-liter soda bottle that can mimic a cat brain. (Why a cat brain? It's a more realistic goal than a human brain, he says.)

Mimicking the function of a cat brain is possible with current technology—it just takes a huge supercomputer with its own dedicated power supply. Even then, it's 83 times slower than a real cat. Why is a real brain so much better as processing information? In part because, unlike a computer that executes code linearly, one piece at a time, a mammalian brain can do many things at once.

The connections in a biological brain also work more efficiently. The synapses that connect neurons can be reconfigured (with interesting consequences for memory), form many different pathways and strengthen them based on the intensity and timing of the electrical signals produced by the various neurons. A traditional transistor in a computer, though, only connects to a few neighbors and doesn't have any memory of past signals.

Lu's first step into creating a cat brain computer was to develop a "memristor" that acts more like a biological synapse and has a memory of past voltages to pass through it. He then connected two circuits with his memristor and showed that they system was capable of a type of memory and learning process called "spike timing dependent plasticity." Lu is still years away from his soda-bottle-sized cat brain dream, however. So if you want a cat brain to figure something out for you, you'll still need a furry friend.

Get the latest Science stories in your inbox.