Cognitive Computing: Building a Machine that can Learn from Experience
University of Wisconsin-Madison (12/17/08) Smith, Susan Lampert
University of Wisconsin-Madison psychiatrist Giulio Tononi is working with scientists from Columbia University and IBM to develop software for a thinking computer, while nanotechnology and supercomputing experts from Cornell University, Stanford University, and the University of California-Merced are developing the hardware. The collaborative effort has been awarded a $4.9 million grant from the Defense Advanced Research Projects Agency for the first phase of DARPA's Systems of Neuromorphic Adaptive Plastic Scalable Electronics project. The goal is to create a computer capable of sorting multiple streams of changing data to find patterns and make logical decisions. The finished cognitive computer must also be no larger than the size of a small mammal's brain and use as little power as a 100-watt light bulb. Although the project draws inspiration from the brain's architecture and function, Tononi says that it not possible or desirable to recreate the entire structure of the brain down to the synapse level. "A lot of the work will be to determine what kinds of neurons are crucial and which ones we can do without," he says. Value and reward systems are important, and learning is crucial because a cognitive computer must be able to learn from experience. Tononi says the artificial brain will need to be able to change as it learns from experience, and the design will most likely convey information using electrical impulses modeled after the spiking neurons found in mammal brains.
http://www.news.wisc.edu/16085