Bobby Travis

IBM Creates Brain-like Computer Chips That Learn and Remember  | 40Tech

Remember IBM? They may not be the premier computer manufacturer they once were, but the world’s oldest computer company is still skating on the cutting edge of technology. Their latest achievement? Two DARPA SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics) funded prototype chips that can learn and remember in a way that “begins to rival the brain’s function, power, and speed.”

Can you say Skynet?

Ok, so we’re nowhere near AI-driven world takeover, but thinking computer chips that require very little in the way of size or power to operate is definitely a step in the bright direction. What these chips really represent is a shift in the way computers process information. They move away from the traditional Von Neumann architecture that relies on processing and memory to work separate to each other. Both chips have 256 neurons, with one chip containing the programmable synapses, and the other, learning synapses that can “remember and learn from their own actions.”

IBM’s end goal is to create a shoebox-sized chip/brain that has some 10 billion neurons and 100 trillion synapses that runs on 1 kilowatt-hour of power (the human brain has 150 trillion synapses and needs about 20 watts). Eventually, they are looking to create cognitive computers that can take detailed input from multiple sources at the same time, process it and make a decision based on its own experiences as well as its programming. IBM has just finished phases 0 and 1 of the project and have already been experimenting with machine vision, associative memory, patterns recognition, navigation, and more.

Cognitive Computer Chip Demonstrates Handwriting Recognition

Like I said, don’t take a hammer to your computer and networked devices just yet — the fun’s only just beginning. These brain chips could herald a whole new age of computing.


New Computer Chip Modeled on a Living Brain Can Learn and Remember [Popsci]