The brain is a marvelous thing and scientists at IBM have managed to create a computer chip that works like it.
In the words of Dharmendra Modha, lead investigator for the SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics) project, the chip is essentially a supercomputer about as big as a postage stamp with the weight of a feather and power consumption of a hearing aid. A prototype was built in 2011 based off designs from a monkey brain created in 2010, fitting one core and 256 neurons in a chip about the size of a worm's brain. Today, the chip (codenamed "TrueNorth") houses over 4,000 times more cores although it is 15 times smaller which makes it consume 100 times less power.
"I'm holding in my hand a new machine for a new era. It's the culmination of over a decade of our research. Ten years ago, many believed this was impossible. The impossible has become possible, and the possible will very soon be real applications," mused Modha.
TrueNorth has a million neurons, 4,096 neurosynaptic cores and 256 synapses all arranged in a 64 x 64 array. It can accommodate 400 million bits of memory too, while cores are interconnected via a network. The chip can be tiled seamlessly and it can double in size and power. So far, scientists have managed to tile 16 chips together in a 4 x 4 array, translating to four billion synapse cores and 16 million neurons together.
So it's a very, very powerful computer chip basically?
Yes and no. While TrueNorth is indeed more powerful than the standard computer chips around now, it is also different. Modha explains it using the analogy of the left and right brains. A standard computer chip follows a more left-brained set of instructions while TrueNorth is more of a right-brained computer chip, a chip that "perceives" more.
One of the applications that the computer chip will be useful at is real-time identification and traffic. In a test, the chip picked out people and cars separately while highlighting positions using just a few milliwatts of power. Computers can already do this but using the chip vastly increases efficiency while dramatically reducing power consumption as well. This also opens up the possibility of using the chip on smaller devices to allow for more complex computing tasks without draining power.
The SyNAPSE project cost $53.5 million and was funded by the Defense Advanced Research Projects Agency.
