Researchers from IBM introduced a chip that the size of a postage stamp that operates like a supercomputer and mimics the functioning of the human brain. For example, in a standard video test could recognize the presence of people, cyclists, trees and other objects with greater accuracy, speed and economy than any conventional computer resources.
The chip so-called “neurosynaptic” opens up a whole range of possibilities in computing, from cars that run themselves to artificial intelligence systems that can be installed on smartphones, explained their creators.
Researchers at IBM, Tech and Cornell collaborators around the world said they required a whole new design concept compared to previous computing architectures, moving towards a system called “cognitive computing”.
“We are inspired by the cerebral cortex to design this chip”, said Dharmendra Mohda, chief scientist at IBM for computation inspired by the brain.
Mohda explained that the lineage of the current computer goes back to machines created in the 1940s (the so-called Von Neumann architecture), which are essentially “calculators of sequential numbers”, acting mathematically or carrying out tasks on the left side of the brain, but little else.
But the new chip, named “TrueNorth” operates imitating the right side of the brain, where they are functions that process the information perceived by the senses, so we can respond to images, aromas and ambiguous information about their environment to “learn” to respond in different situations.
The system manages to do this using a large network of “neurons” and “Synapse”, similar to those that the human brain uses to use information gathered from the senses, encoding the information as patterns of pulses that circulate inside, instead of mathematical calculations.
The researchers designed TrueNorth with a million programmable neurons and 256 million programmable synapses inn a chip with 4096 cores neurosynaptic and 5400 million transistors; the version presented is the second generation of these processors with a performance that seeks to be closer to that of an animal brain.
Another key factor of this chip is its low power consumption, since it can work with a small battery like those used in hearing aids, so it could be installed in cars or smart phones. Those 5.4 billion transistors consume 70 milliwatts of power, while a conventional PC chip has less than half of those transistors but consumes 35 to 140 watts.
Its inventors believe it will still take years before the chip is available in commercial applications, but note that has the potential to “transform society”, especially when in a future “hybrid computers” combine the capabilities of the left and right side of our brain.