IBM unveils the world’s most powerful neurosynaptic chip, marking another milestone in the development of “thinking machines”.
IBM recently announced a major milestone in the development of computer chips that mimic the human brain, unveiling a new postage stamp-sized device.
Called the Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) chip, the new processor was created in conjunction with Cornell Tech and iniLabs and was funded by DARPA.
Built using neural network architecture, SyNAPSE has one million programmable neurons, 256 million programmable synapses and can churn through 46 billion synaptic operations per second, per watt.
Though its size and processing capability aren’t close to reaching some of IBM’s advanced, traditional chips, the manner in which it can process operations is more akin to a human brain than a conglomeration of silicon and metals.
Unlike traditional chip architectures, neural networks attempt to replicate the cognitive processes that occur in the human brain. Through the use of complex algorithms and intricately pattered [DA1] circuitry IBM’s designers are hoping to build an inorganic system that can hypothesize, learn from past experiences and come up with creative solutions to complex problems.
While IBM’s breakthrough marks an important advance in the field of “cognitive computing” Dharmendra Modha, chief scientist and a vocal proponent of neurosynaptic chips, says cognitive computing is still in the development stage: “We have not built a brain … but we have come the closest to creating learning function and capturing it in silicon in a scalable way to provide new computing capability that was not possible before.”
With further work Modha and his team imagine SyNAPSE-like chips could power a new generation of adaptable sensors, fuel evolving learning systems and transform our mobile devices from simple gadgets to extensions of our senses.
In the end, Modha and his team have visions of building what they’ve called “a brain in a box”. Outfitted with a complex and energy efficient architecture the artificial brain would have the equivalent of 100 billion synapses and consume a single kilowatt of power. When a device of that capability would be available is anyone’s guess. However, Modha and his team have only been at work since 2008 and I imagine that in the coming decades they’ll make many more significant advances.
Image Courtesy of IBM