When Computers Get Brains

So much happened last week, what with Wall Street in need of a sedative and Gerard Depardieu in need of a bathroom, you probably missed the news that a team led by IBM has created a computer chip that mimics how a brain works.

Big deal, right?  Hadn’t they already created the computer that delivered a smackdown of those two “Jeopardy” whizzes turned hapless humans?

Actually, this latest creation is something very different and potentially more momentous. Watson, the “Jeopardy” god, is a ridiculously powerful computer that, nonetheless, operated in a fairly conventional way—except it was retrieving info from a database of more than 200 million pages of content.

But the new invention, the “cognitive” computer chip, is a leap into uncharted territory. The chips, as they become more sophisticated, could eventually copy the brain’s ability to sense surroundings, recognize patterns, and—dare I say it—learn on their own.

IBM’s Dharmendra Modha headed up the project, which so far has involved researchers from four universities and more than $20 million from DARPA, the Defense Department’s high-end research arm best known for creating the predecessor of the Internet. Modha uses a right-brain, left-brain analogy to explain what the team has conceived.

Computers have the left-brain part down cold.  They’re sequential and analytical and make us humans seem immensely dull-witted when it comes to processing numbers and data. But they can’t make connections that aren’t programmed or pull in information from their  surroundings to re-evaluate the data. That’s where the right-brain computer would come in, says Modha. Without requiring much memory, it would be able recognize changes in the environment and consider those before taking action.

Modha, thankfully, has offered a few real-world examples—traffic lights that can take in sights, sounds and smells, and, by pulling them together, flag an unsafe intersection before an accident happens. Or a grocer’s glove with sensors that integrate temperature, smell and vision to determine if produce has gone bad. As Modha sees it, cognitive chips would work with existing computers to produce a total brain experience.

A breakthrough came two years ago, when scientists developed something they call BlueMatter, a software algorithm that simulates the pattern of connections within the brain. By 2020, they think they’ll have a computer that can go brain-to-brain with a human being.

That sounds a bit optimistic. The human brain has about 100 bilion neurons. IBM’s two cognitive chips have 256 neurons each. That’s about the brain power of an earthworm. But the chips have taught themselves how to play Pong. Which sets up the possibility that one of us could lose at Pong to the equivalent of an earthworm.

And so it begins.

Brain candy

Here are other tasty things going on in brain research:

• You’re getting sleepy: Researchers for a California firm called NeuroSky are testing car headrests that can pick up your brain signals and set off an alarm if it detects you’re dozing off.
• It’ll come to you: Turns out the brain doesn’t need external stimuli to remember something; sometimes it just needs a little time.
• Sweet sensations: When it comes to sugar cravings, wanting and liking are two different things
• Hit makers: Teenagers’ brain waves while listening to a song could help determine if it will be a hit.

Bonus: No one has made more out of the notion that the future will belong to right-brained people than Daniel Pink, the author of A Whole New Mind: Moving From the Information Age to the Conceptual Age. Here’s one of the better interviews with him.

Does the idea of a computer that “thinks” like a human creep you out?  Or do you think it’s time to get over our fear of computers like Hal in 2001: A Space Odyssey?

Randy Rieland | | READ MORE

Randy Rieland is a digital media strategist and contributing writer in innovations for Smithsonian.com.