Researchers Assist IBM in Cognitive Computer Chip Design
Innovative new design is modeled on nervous systems, creating computers that can learn through experience, find correlations, create hypotheses and remember outcomes
A team of researchers led by IBM, including a pair of professors from the University of California, Merced, unveiled today a new generation of experimental computer chips designed to emulate the brain's abilities for perception, action and cognition. The result could be processors that use much less power and far less space than those found in today's computers.
UC Merced professors Chris Kello and Stefano Carpin have been and will continue heading up one aspect of the project — designing and implementing virtual environments to test these revolutionary new systems. The building blocks of cognitive computers, these cores are expected to learn through experiences, find correlations, create hypotheses and remember and learn from the outcomes, mimicking the brain's structural and synaptic plasticity.
The goal of the project is to create a system that not only analyzes complex information from multiple sensory inputs at once, but also automatically rewires itself as it interacts with its environment — all while approaching the remarkable power and size efficiency of the human brain. To get there will require research that incorporates principles from nanoscience, neuroscience, computer science and cognitive science.
"This project represents interdisciplinary research at its finest," said Kello, a cognitive scientist in UC Merced's School of Social Sciences, Humanities and Arts. "For decades, scientists and engineers have worked on theories of cognition and intelligent algorithms without taking seriously the basic fact that human intelligence is supported by brains that weigh about 3 pounds and consume about 20 watts of power. By contrast, today's supercomputers weigh tons and consume megawatts of power."
UC Merced recently received a $300,000 grant for Phase 2 of the project — known as Systems of Neuromorphic Adaptive Plastic Scalable Electronics, or SyNAPSE — as part of $21 million in new funding from the Defense Advanced Research Projects Agency (DARPA) to the IBM team. Phases 0 and 1 have been successfully completed, and the first two prototype chips have already been fabricated and are undergoing testing.
Dharmendra Modha, project leader for IBM Research, said future applications of the technology could include traffic lights that can integrate sights, sounds and smells and flag unsafe intersections before disaster happens, or cognitive co-processors that would allow servers, laptops, tablets and phones to better interact with their environments.
The UC Merced work involves creating virtual environments in which to test this technology without the costs or complications of testing them in the real world.
"We are developing a high-fidelity simulation environment to test this new technology," said Carpin, a computer scientist in the School of Engineering. "This effort builds upon our numerous years of experience in this area, and we are proud that UC Merced is playing an important role in this project."
