Friday, August 2, 2013

K-NEST

Recent simulation of NEST on a K Computer in Japan demonstrated a neural network close to 1% of a human brain. Although this is encouraging and point to the possibility of simulating an entire brain using exa-scale computers in the future, it also raises many questions as to the utility of programmatic simulation of the human brain. It has been a fascination for many ordinary men and women, for ages. A powerful quantum computer, housed in such an efficient space, standing ready to learn and forget, tantalized those who attempted to replicate it in-silica – but is has all been in vein.

Simulation of the human brain by traditional means, using raw computing power, is unlikely to produce insights into its workings. The brain is an efficient quantum computer, not something that possesses very high computing power and memory as measured by traditional computing metrics. Its specs are mediocre at best but it does have a highly sophisticated operating system, that is able to take advantage of its limited capabilities. In the absence of such powerful software, able to adapt to modular and learning apps, some able to reconfigure on demand, it will be less interesting. Failure of the operating system or some critical apps, render the brain incompetent quickly, regardless of its supposed hardware depth.

Sheer scaling of computing power and memory, although a straightforward experiment, is unlikely to yield the mysteries of this remarkable organ.