Our old friend the Macaque, so often used for cognitive studies, has come through yet again. IBM has just announced the manufacture of a pair of “neurosynaptic computing chips”, their SyNAPSE system is to be funded by DARPA (that is the US tax payer of course) to the tune of $21M (1).
What’s DARPA got to do with Macaques I hear you ask, well, nothing, but IBM have been probing Macaques neural networks for some time. I rush to say that’s not equating their customers with Macaques, but due to the recognition that to take computing to a new level, building in the plasticity of the brain can make cognitive computing possible.
Last year, Modha and Singh of IBM have published their work on tracing a unique network of the Macaque brain using the public domain database CoCoMac (2). This enabled them to sort out the structure of the network. They demonstrated that the brain network was not uniform like a 3D fishing net, but has increasing complexity as it went to specific functional regions, more like a group of trees growing from common roots. This means that the complex networks feed information down to simpler sub-nets.
Why is this important? Because it like a funnel concentrating the information while processing it down to the nub of the matter. This is a good model for a cognitive system to work efficiently.
Brains are nicely plastic in that the network is modified by the information it is handling, that is it is making new connections. Ultimately this also provides a route for repair of the network. Making new connections on a chip is currently not possible on the fly, but the SyNAPSE system can apparently modulate the signal so it can effectively switch connections on and off. Well almost, it’s more like turning down the signal so it lost in the background and turning the one you want up so you’re not trying to process everything with everything, as I understand it. IBM could put it much better but they aren’t saying too much at present.
This is potentially a huge step forward towards cognitive computing. The next step beyond this is for our computers to start being intuitive. I hope they don’t start showing too much initiative though.
