The human brain is the fastest and most efficient computer we've ever developed, but there’s still a lot about it we still don’t understand. However, with what we do know, can we use that knowledge to create faster and more energy-efficient computers? Scientists at Stanford University think so, and have developed a circuit board called Neurogrid that's stupendously faster and hugely more energy efficient than any standard PC.
Compared to the human brain, computers based on conventional silicon structures are slow and inefficient, needing 40,000 times more energy to run. It’s no surprise, then, that scientists are looking at simulating the brain for computing. The Stanford team created 16 special chips, or neurocores, for Neurogrid. Each of these neurocores simulates the brain’s synaptic connections between millions of neurons. Some of these connections share circuits on the board, and they're 100,000 times more energy-efficient than a standard PC. When put together, the entire circuit board is about the size of an iPad. But most impressively, the Neurogrid runs 9,000 times faster than modern computers, for some specific applications.
The Stanford team is researching using Neurogrid for future prosthetic technologies. The concept is to put a chip that would interpret movement inside a paralyzed person’s brain, giving them better functionality over their prosthetic limbs. Unfortunately, the Neurogrid isn’t currently cost-efficient. Each board’s creation costs about $40,000, but the Stanford team believes that new fabrication techniques could bring that price down to about $400 per board. They also plan on creating software that makes it easily configurable, as most programmers aren’t also neuroscientists. With these ideas in mind, the research team hopes that Neurogrid provides real and workable solutions for robotics and cybernetics.