Ken Jennings, who won a record 74 consecutive games on "Jeopardy!,"… (Seth Wenig, Associated…)
Reporting from New York — — Will supercomputers one day crush our puny mortal brains? Some scientists at IBM believe it's possible. And if they have their way, it will start Monday night.
Beginning on Valentine's Day, an IBM-engineered machine named Watson will compete in a three-night, two-game stretch on "Jeopardy!," battling all-star champions Ken Jennings and Brad Rutter. Only one will emerge as interspecies overlord, but with 10 refrigerator-sized racks of IBM Power 7 Systems firing its mega-mind, Watson's got a pretty good shot. When IBM staged a practice round with Jennings and Rutter before a group of journalists, Watson won. Asked if he was nervous about the next big match, Rutter evoked "The Terminator," joking, "I will be [nervous] when Watson's progeny comes back from the future to kill me."
Watson's not that threatening yet — but Rutter has good reason to worry about his opponent. Named after IBM founder Thomas J. Watson, this industrial behemoth is the world's most sophisticated question-answering system, capable of analyzing puns, riddles and other linguistic complexities that traditionally have been the province of flesh-and-blood nerds. Much more than a glorified search engine, it doesn't just match key-word searches with links — like the calm-voiced computer on "Star Trek," it actually answers questions itself. Armed with vast databases of encyclopedias, plays, textbooks, dictionaries and other reference materials (for accuracy's sake, it doesn't use the Internet), Watson can respond to clues with the same speed, ear for irony and knowledge of obscure Lady Gaga trivia as an old-fashioned Homo sapien.
Now it's ready for the biggest man versus machine showdown since IBM's Deep Blue beat chess champ Garry Kasparov at his own game. And Jennings is a bit anxious about the whole thing. "It's a new experience for me to feel like an underdog, playing against this unstoppable supercomputer," he says. "At IBM's research lab, the center of the stage had a big Watson logo, like you're playing a basketball team on its own court. I knew this was gonna be an away game for humanity."
No matter who (or what) comes out on top, the idea that a humble collection of ones and zeroes can even compete on the same level as the complex proteins in the human skull may feel a little unsettling. Over the last decade, question-answering systems have become exponentially more sophisticated: They already help legal firms sift through case law. IBM hopes Watson's technology soon will enable hospitals to better diagnose illnesses and call centers to find quicker solutions for customers' problems, all of which raises bigger issues: Will Watson soon replace healthcare professionals and help-desk workers? Will its mastery of language allow it to hold conversations with friends and neighbors? Just how close toward world domination can Watson get?
Granted, it still has a long way to go. Watson still gets questions wrong, often in a way that's unintentionally funny. Asked what a grasshopper eats, it once responded "kosher." Probed about a three-word phrase that describes what frustrated drivers in the snow do ("spin their wheels"), it offered "lock and load." When determining which brand of paper towel had a lumberjack as its logo, its top choices were "Brawny" and "Jesus Christ."
Somehow, these mistakes make Watson surprisingly charming, lending him a certain vulnerability. "We're conditioned by Hollywood to view these super-powerful computers as possible world-ending threats and dictators, whether it's 'The Matrix' or '2001,' and we associate Watson's implacable voice with the voice of doom that's going to put us in meat farms some day," explains Jennings. "So it's a relief to see that AI is not that far along, that naïve errors are still possible."
Rutter points out that Watson's lack of self-awareness can also be a major advantage. "There's something about when it gives a shockingly incorrect answer with the same chirpy HAL 9000 voice that feels so hopeful," he says. "Watson won't get discouraged, it will just try again."
With its avatar-like "face" — a globe that glows green or orange, depending on its confidence level — Watson already seems almost human. In fact, Watson works in the same highly parallel way a human brain does: Just as we use billions of neurons at the same time to instantly process ideas, Watson analyzes millions of documents at once, delivering a list of possible responses, and selecting the one that's backed up with the most evidence, all within two or three seconds.