facebook rss twitter

IBM unveils chips that emulate the human brain

by Scott Bicheno on 18 August 2011, 17:46


Quick Link: HEXUS.net/qa6xc

Add to My Vault: x

Self-conscious silicon

Tech giant IBM has unveiled a new type of computing chip that is designed around the way the human brain works, claiming it has the potential dramatically reduce the power consumption and size of today's chips.

The technology is being referred to as neurosynaptic computing, and the potential products called cognitive computers. The micro architecture of the chips is designed to recreate the kind of synaptic activity you get in biological systems such as the brain, and two prototypes are currently being tested.

"This is a major initiative to move beyond the von Neumann paradigm that has been ruling computer architecture for more than half a century," said Dharmendra Modha, project leader for IBM Research.

"Future applications of computing will increasingly demand functionality that is not efficiently delivered by the traditional architecture. These chips are another significant step in the evolution of computers from calculators to learning systems, signalling the beginning of a new generation of computers and their applications in business, science and government."

Here's a passage from the announcement that summarizes the thinking behind the technology and is best presented without our re-interpretation.

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.  

Here's an overview vid from IBM, and a bit of fun underneath, chosen because embedding is disabled on this.




HEXUS Forums :: 5 Comments

Login with Forum Account

Don't have an account? Register today!
Moose with actual intelligence!! :surprised:
I don't know many details on the IBM chip yet, but another interesting project is Spinnaker, headed up by Manchester Uni's Prof. Steve Furber.


It's effectively a whole lot of chips, each with a bunch of ARM cores, and a network topology and messaging protocol that mimics neurological behaviour.
I played with the same sorts of things at York:

I used an artificial neural network in my MEng project, but this stuff is a bit different - rather than simulating neural networks in software, it's architecting hardware that behaves in the same or a similar way.

This is interesting, because if we can truly exploit neural network-based architectures, then we can make things supermassively parrallel and pretty much completely asynchronous… which would be nice.