In a discussion last evening, a friend and I were debating over what is true A.I. I stated the simplistic autonomous robots used today is the best accomplished. We both agreed that with the limited number of rules these robots follow, it is more like an algorithm they are following. Our conclusion is modeling the human mind, the most complex machine known to man. On the heels of the talk, I decided to look into specifications of a brain.
The brain transfers data at 10 hertz. The 4+ gigahertz processors seen today can not approach the processing power of the brain, why? Parallel processing and data volume transfer speeds. With 100 billion neurons (the processors), 100 trillion synapses (memory), and axons in every neuron (data patheways). Why not just model the brain? With advancements in nanotechnology, supercomputing, and neuroscience this is a possibility.
Columbia University’s Stefano Fusi describes the brain vs. the computer.
IBM will attempt to do just that. In a departure from the von Neumann computer architecture, that transfers data on a bus between a single processor and memory storage, to a brain like parallel handling of data called SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics).
IBM researchers John Arthur and Paul Merolla describe the inspiration for the project.
IBM's first SyNAPSE based system will be a scaled down mini-brain.
? 256 neurons
? An array of 256 x 256 (65,536) synapses
? 256 axons
This "core" will be constructed with only a few million transistors, much less than the billions found in many CPUs today.
The final step for IBM is to combine these cores into a 100 billion neuron, 100 trillion synapse system rivaling human brain power. It will be have a volume of 2-litres and consume about 1 kW. IBM's Blue Gene super computer has 147,456 processors, 144 terabytes of memory, and is in large air conditioned cabinet consuming 2 megawatts of power.
Steven Esser of IBM research describes the software
The SyNAPSE comes from a $21 million dollar grant from DARPA. Six IBM labs and four universities (Cornell, the University of Wisconsin, University of California at Merced, and Columbia) along with some government researchers.
Will this lead to true, real time, human like artificial intelligence? IBM thinks so. They are already "waving the banner" of its use in monitoring for natural disasters and sending out warnings. Saving people's lives in the cornerstone for research grants these days.