New Computer Chip Inspired By Real Brains
Computers that can learn won't just reside in the heads of imagined "Terminator" robots anymore. Technology giant IBM has unveiled experimental computer chips designed for a new generation of brain-inspired computers capable of learning from what's going on around them.
Such cognitive computing aims to use information from many different inputs similar to how the human brain can harness its experiences of hearing, sight, smell and touch. The new chips could allow a sensor-studded glove to monitor the sights, smells, texture and temperature of grocery store products and flag bad or contaminated foods. Or they could enable computers to issue tsunami warnings based on second-by-second tracking of temperature, wave height and ocean tide.
"Imagine traffic lights that can integrate sights, sounds and smells and flag unsafe intersections before disaster happens, or imagine cognitive coprocessors that turn servers, laptops, tablets and phones into machines that can interact better with their environments," said Dharmendra Modha, project leader for IBM Research.
Modha's effort takes the human brain as a benchmark, given the brain's compact size and power-sipping efficiency that uses just 20 watts less than most household light bulbs. The brain has 100 billion neurons and 100 trillion synapses; the synapses are tiny gaps between neighboring neurons that act as communication pathways for chemical messengers.
IBM's latest chips have the silicon equivalent of 256 neurons and come in two flavors. One holds 262,144 programmable synapses, and the other has 65,536 learning synapses.
But IBM eventually wants to build a chip system equivalent to 10 billion neurons and 100 trillion synapses. Such a system would consume just a kilowatt of power about 10 hundred-watt light bulbs and take up less space than a two-liter bottle.
Cognitive computing still faces the huge challenge of using its new architecture in as flexible and coordinated manner as the ancient human brain . Still, the IBM team has already shown that their new chips can do simple tasks such as navigation, pattern recognition and recognizing visual objects.
The new chips form part of IBM's Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE) project. That effort recently received a new round of $21 million in funding from DARPA , a Pentagon agency dedicated to funding disruptive science and innovation for the U.S. military.
"These chips are another significant step in the evolution of computers from calculators to learning systems, signaling the beginning of a new generation of computers and their applications in business, science and government," Modha said.