Brainstorming Engineer looks to gray matter for new technology
ESSEX, Vt. -- It's nearly as fast as the most advanced computer, but uses a fraction of the energy. It simultaneously zaps information to thousands of points and is equipped to correct itself.
It's not made of silicon, and it came long before the computer chip.
It is the human brain.
In searching for new ways to advance computers, engineers are looking to man's gray matter for inspiration. And while comparing the two, they often wonder why a computer can't act more like a brain.
For IBM senior technologist Kerry Bernstein, the technological possibilities are astounding.
His curiosity was piqued five years ago as he chatted with a neurosurgeon at his local gym.
Martin Bednar, then head of neurosurgery research at the University of Vermont, explained to him how neurons and the brain work. In exchange Bernstein talked to him about how transistors and chips works.
"And it hit us like a freight train simultaneously that in fact the underlying physics is the same," Bernstein said.
"So that begged the question: If the underlying physics is the same, have both systems been evolving toward the same solutions to compute problems?"
Yes, they have.
But the brain naturally had a better way.
It operates at roughly 12 kilohertz -- the equivalent of 12,000 cycles per second -- and burns a fraction of the power computers do, Bernstein says. That makes it exponentially more efficient than the fastest computer, he says.
"The reason is because of something that we can't do in electronics," Bernstein said. "It's this notion of massive parallelism." Meaning one bit of data can spread to 100,000 other neurons, he said.
As computers become faster, and chips become smaller, crammed with more information requiring more precision, they use more energy. That's the dilemma engineers are facing and there are real physical limitations, Bernstein says.
"The digital paradigm has served us well. ... As we do more and more of this it's becoming onerous," he said. "What's very clear is that we cannot continue doing what we're doing. The industry cannot."
Desperate for direction
Computer engineers, like Bernstein, are desperate for a new direction.
And they're confident that it's out there. Electronics is still at the steep part of its evolution, Bernstein says. It evolves much faster than humans.
Mammals add a cubic inch of brain matter every 100,000 years, Bernstein says, while processors are predicted to double in performance and capability once every 12 to 18 months.
"We cannot sustain this growth," he said.
Bernstein's first talk with Bednar led to an ongoing collaboration with neurosurgeons. Neurosurgeons came to speak at IBM and Bernstein gives a talk every year at a conference for neurosurgeons.
"It goes both ways. We've gotten a lot of inspiration from the brain already in microprocessing design," he said.
Learning how the brain works inspires more effective circuit design. It's led to quantum computing and the implantation of neurons on computer chips.