Yes, you read correctly: IBM is prepping Watson to take on America’s brightest minds on the nation’s most famous knowledge showdown. They’re building superfast, superintelligent technology (it’s not called a supercomputer for nothing) that can think like a human, only better. And being that I think like a human, only worse, I’m feeling particularly vulnerable.
Is Watson ‘Jeopardy’ Ready?
Watson may not look like much — a bunch of chips, wires, and casings, really — but don’t let appearances deceive you. This thing boasts cutting-edge natural language processing technology that can rival the knowledge base and brain-power of your favorite Einstein-loving smartypants.
“Watson is a compelling example of how the planet … is becoming smarter,” boasts IBM CEO Samuel J. Palmisano.
So, what exactly does that mean? I’m no genius, so I turned to IBM researcher and Watson project lead Dave Ferrucci for some fast answers. He says Watson starts by “studying” — and we’re talking an entire college career’s worth of material.
“The system would have ‘read’ many, many natural language texts — books, reference materials, all kinds of information — and tried to analyze and organize that information in such a way that it can see the meaning of the question and try to figure out what are likely answers,” Ferrucci explains.
Watson doesn’t stop with knowledge, though: Like a person — particularly one competing on “Jeopardy” — it has to be able to assess how confident it is with its answer. If it buzzes in with some iffy information, it could suffer some serious point losses.
“Determining the confidence is a really big part of the challenge,” Ferrucci says. “Humans can know what they know very rapidly. This is something the Watson computer is working very hard to do.”
A computer with confidence? Sheesh…that’s a real blow to the ol’ self-esteem.
Next: Under Watson’s hood
Under Watson’s Hood
Watson’s been under development for about two years now. Officially considered a question answering (QA) system, it follows in the footsteps of IBM’s previous human-demoralizing machine, Deep Blue. (Deep Blue, in case your nonsupercomputing mind has forgotten, took on a world chess champion in 1997 and won.) It goes without saying, though, that with a decade of extra technology evolution under its belt, Watson will be even stronger than its predecessors.
Watson will likely be built on an architecture comparable to IBM’s Blue Gene, Ferrucci says. Translation: It’ll be fast. Really fast. Blue Gene runs on speeds described with terms such as petaFLOPS and teraFLOPS (and no, those have nothing to do with excessive forehead sweat — they measure FLoating point Operations Per Second). While Watson’s specifics have yet to be finalized, the fastest possible processing is what engineers are after.
“Because of the high levels of precision and confidence required, and the fact that you have do this very quickly, you require massive parallel computation,” Ferrucci says.
Not feeling mush-brained yet? Consider this: Watson will also run with no network connection of any kind. Instead, all of its data will be internally stored — just like a real “Jeopardy” contestant.
Watson’s “Jeopardy” run is being targeted for late 2010. After that, the world will become its playground, and its boundaries are anyone’s guess.
“I think about call centers,” Ferrucci predicts. “I think about getting staff precise medical information, getting people access to information quicker and more effectively to make more rapid decisions in business.”
Ultimately, the technology driving Watson could one day help you do almost any computing task more naturally and easily. That’s the broad goal Ferrucci and his team really want to achieve — after, of course, dominating the “Jeopardy” dojo.
“We really want computers to operate more effectively in human terms, and interact with humans in ways that they feel more comfortable interacting,” he says.
Operate in human terms? Interact comfortably with humans? Geez. I just can’t compete with this thing.