IBM engineers have succeeded in creating a stamp-sized chip that performs similarly as the wonderful human brain. Dharmendra Modha, lead investigator in the SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics) project, maintains that this chip is small in basically a supercomputer that’s not bigger than a stamp, weighs as light as a feather, and consumes power as much as an hearing-aid would. The SyNAPSE project cost $53.5 million in sum, and was funded by DARPA (Defense Advanced Research Projects Agency).
Using designs from a monkey brain created in 2010, a prototype was developed in 2011 which fused a core to 256 neurons in a chip as big as a worm’s brain. At present, the chip (codenamed TrueNorth) accommodates 4000x more cores, is 15x smaller than the prototype, and consumes 100x lesser power. Modha said as he held the machine in his hands, that this was a product of the new era, and was successful only after a decade of research. While their were many naysayers an year ago who didn’t believe such a thing to be possible, Modha has proved them all wrong and states that soon this ‘machine’ shall be employed to real applications as well.
TrueNorth houses a million neurons, 4,096 neurosynaptic cores and 256 synapses all positioned in a 64 x 64 array. It can comprise 400 million bits of memory as well, while cores are connected to each other through a network. The chip can be tiled continuously and can be doubled in power and size. As of now, scientists have succeeded in tiling 16 chips together in a 4 x 4 array, which in basic rendering is four billion synapse cores and 16 million neurons together.
TrueNorth is much more powerful as compared to standard computer chips, but it is, in a way, different. Modha explains this difference taking the analogy of the left and right brains. While a standard computer chip interprets a more left-brained set of instructions, TrueNorth is a right-brain interpreting chip which ‘perceives’ more.
Of its many applications, one application that could prove out really beneficial is its ability to detect traffic and identification realtime. In a test run, the chip pulled out people and cars separately and highlighted positions, using not more than a few milliwatts of power in the process. Computers do it, but the chip does with better efficiency and reduced power consumption. So this implied that the chip could now be used in smaller devices that perform complex computing chores without juicing power.
Via : ExtremeTech