Here are some numbers that blew me away when I heard them earlier last week. You brain is using just a few watts of power right now as it sees and processes these words, hears and sorts through sounds around you and makes mental notes about grocery lists, or dry cleaning that needs picking up. By contrast, a computer uses about 40,000 times more power and runs about 9,000 times slower just to model a mouse brain and a human brain is about 1,000 times more complex. Given that, it’s no surprise several groups are hard at work trying to create a computer chip with brain-like efficiency.
Stanford Bioengineer Kwabena Boahen, PhD, and his graduate student Ben Varkey Benjamin have announced a milestone in this effort: they’ve modeled one million neurons in real time on a or a circuit-board called Neurogrid that contains sixteen chips called Neurocores. Their publication, in the Proceedings of the Institute of Electrical and Electronics Engineers, goes into more detail about exactly how they are using electronics parts to mirror our own intricate collection of cells, as does this story about the work.
What I found most interesting are the possible uses of such a chip. Obviously, it could make our personal electronics smaller, smarter and less power hungry. But the chip can also for the first time model how our brain works, and how it fails to work in some diseases. This is something that once required supercomputing capabilities, plus lots of time and power. Now anyone can do it.
The chip also makes possible the dream of interpreting signals from the brain and, in real time, using those signals to drive robotic limbs for paralyzed people. As things are now, a person would be tethered to a computer and a power supply to interpret brain signals, and the limb wouldn’t move in real time. A Neurocore-like chip could conceivably be implanted, interpreting signals and driving robots in real time with minimal power needs. Boahen is working with his Clark Center neighbor and fellow Bio-X affiliate Krishna Shenoy, PhD, who is professor of electrical engineering and neurobiology, on making that dream a reality.
This video by my colleague Kurt Hickman shows where the team is now in working with Neurogrid to drive robot movement.
Photo by Kurt Hickman