IBM announced its creation of a programming architecture for a cognitive computer chip inspired by the human brain, so that developers can produce apps for them once the chips are eventually realized.
The hardware for the cognitive computers is constructed around neurosynaptic cores based on the brain, and they include 256 processors, 256 memory elements, and 64,000 processor-memory connections. IBM's long-term goal is to build a cognitive machine scaled to 100 trillion synapses, and the company has created a model of the system on its Blue Gene supercomputer, using it to develop the programming architecture.
The central component of the language is an object-oriented abstraction of each neurosynaptic core, or corelet. Each corelet has 256 outputs and 256 inputs, and project leader Dharmendra S. Modha says "our architecture is like a bunch of LEGO blocks with different features. Each corelet has a different function, then you compose them together."
Cognitive computers can be used for apps entailing pattern recognition and other big data-sifting challenges that are beyond the capabilities of traditional computers. Modha envisions future computers merging both traditional systems and the cognitive chips.
From Forbes
View Full Article
Abstracts Copyright © 2013 Information Inc., Bethesda, Maryland, USA
No entries found