This week, IBM announced new computer chips it’s referring to as “cognitive computing.” According to press materials, the chips are inspired by neurobiology, and work like a biological brain, with neurons and synapses:
While they contain no biological elements, IBM’s first cognitive computing prototype chips use digital silicon circuits inspired by neurobiology to make up what is referred to as a “neurosynaptic core” with integrated memory (replicated synapses), computation (replicated neurons) and communication (replicated axons).
IBM has two working prototype designs. Both cores were fabricated in 45 nm SOI-CMOS and contain 256 neurons. One core contains 262,144 programmable synapses and the other contains 65,536 learning synapses. The IBM team has successfully demonstrated simple applications like navigation, machine vision, pattern recognition, associative memory and classification.
IBM’s overarching cognitive computing architecture is an on-chip network of light-weight cores, creating a single integrated system of hardware and software. This architecture represents a critical shift away from traditional von Neumann computing to a potentially more power-efficient architecture that has no set programming, integrates memory with processor, and mimics the brain’s event-driven, distributed and parallel processing.
IBM’s long-term goal is to build a chip system with ten billion neurons and hundred trillion synapses, while consuming merely one kilowatt of power and occupying less than two liters of volume.
Now I’m no big city neuroscientist, but as it stands, with 256 “neurons” the chips would possess less cognitive power than a nematode worm and several orders of magnitude less than that of a fruit fly. Even a chip system with 10 billion neurons would have about ten times less cognitive ability than a human (give or take), although with 100 trillion synapses that system would be massively parallel, enabling it to process an incredible amount of data.
Still, and correct me if I’m wrong, here, once you get to 10 billion simulated neurons, it seems like the logical next step will be 100 billion, which is meeting or exceeding human cognitive power, and then a trillion, and so on, and at that point you’ve got what amounts to a ridiculously powerful superintelligence.
Given IBM’s pioneering work in supercomputing and AI, I’m willing to give them the benefit of the doubt here. A radical new chip architecture is big news by itself, and if it works as promised, by learning through experiences rather than being programmed, we’ll have something new and exciting on our hands, and a potential path forward for true AI. Call it cautious optimism.