IBM on Thursday announced it has created a chip designed to imitate the human brain’s ability to understand its surroundings, act on things that happen around it and make sense of complex data.
Instead of requiring the type of programming that computers have needed for the past half-century, the experimental chip will let a new generation of computers, called “cognitive computers,” learn through their experiences and form their own theories about what those experiences mean.
The chips revealed Thursday are a step in a project called SyNAPSE (Systems of Neuromorphic Adaptive Plastic Scalable Electronics). The two chip prototypes are a step toward letting computers “reason” instead of reacting solely based on data that has been pre-programmed, IBM says.
“Imagine traffic lights that can integrate sights, sounds and smells and flag unsafe intersections before disaster happens,” said Dharmendra Modha, the project leader for IBM Research. “Or imagine cognitive co-processors that turn servers, laptops, tablets and phones into machines that can interact better with their environments.”
The chips’ processing power is not unlike that of IBM’s Watson supercomputer that beat two human champions on “Jeopardy!” this year.
Other scenarios the researchers envision: A computing system that could monitor the world’s water supply — measuring things like temperature, pressure, wave height and acoustics — then give a warning when it thinks a tsunami is likely.
Or imagine a sensor that a grocery store owner could use to read sights, smells and temperatures and give an alert that produce may have gone bad.
“The computers we have today are more like calculators,” Modha told tech blog VentureBeat. “We want to make something like the brain. It is a sharp departure from the past.”
One of the prototype cores contains what amounts to 262,144 programmable synapses, and the other contains 65,536 learning synapses.
Using the chips, IBM researchers have built a “brain wall” at a San Jose, California, lab. The long-term goal? A one-square-centimetre chip with the equivalent of 1 million neurons and 10 billion synapses.
Modha and other researchers say that using current programming techniques, any computer that approached what they’re trying with the SyNAPSE project would have to be larger and would suck up more energy.
For the next phase of SyNAPSE, IBM has assembled teams from Columbia University, Cornell University, the University of California, Merced and the University of Wisconsin, Madison.
Phase 2 of the project has been rewarded $21 million in funding from DARPA, the U.S. military’s research branch.