The wait is over. After months of hype, it will be man versus machine tonight on Jeopardy, as IBM's Watson supercomputer takes on the greatest champions in the game show's history, Ken Jennings and Brad Rutter. The three-day match provides perhaps the ultimate test of a supercomputer's ability to master natural language. In simulation games, Watson has mopped the floor with many winning Jeopardy contestants and a number of journalists.
Speed to the Buzzer: A key to success in Jeopardy is ringing in first, and securing the first opportunity to score. This may be an area where Watson has a decided advantage. "The machine is lightning-fast at buzzing in answers," says journalist Bruce Upbin from Forbes, who played several practice games against Watson. "The best Jeopardy! champions get to the buzzer first more than 60% of the time. Watson can sometimes push that buzz-in rate above 70%." The pressure to ring in first may also result in more incorrect answers, which may be a larger issue for Ken and Brad.
One Computer, Two Humans: Jeopardy is designed as a three-player game. But it's possible that a divide-and-conquer strategy will work to Watson's advantage. On the occasions when the supercomputer is slow to the buzzer, Jennings and Rutter may well split the points that are available for human consumption, increasing the likelihood that Watson could prevail.
Puns: The Jeopardy quizmasters love to build categories around puns and plays on words. This is the potential Achilles Heel for Watson. Punny questions and categories will represent an opportunity for the humans.
Daily Doubles Could Be Pivotal: If given the opportunity, will Watson bet the ranch? The Daily Doubles allow players to choose their wager - up to all of their cash - on whether or not they can provide the correct response on a single question. My guess is that the IBM team sees Daily Doubles in terms of risk rather than reward, and Watson is programmed to make modest wagers on Daily Doubles, but that may not always be the case (read more on Watson's strategy on Daily Doubles). The strategy may cut differently for Jennings and Rutter, especially if they are behind, which would raise the likelihood of a bold, game-changing gambit on a Daily Double. (Note: I never saw Rutter play, but Jennings was never particularly aggressive on Daily Doubles during his winning streak).
So who will win? IBM is clearly very confident about its prospects. The Watson team has a wealth of information about Rutter and Jennings and how they play the game, and can fine-tune the supercomputer to best take advantage of their tendencies.
I'll be surprised if Watson doesn't win. But I'll be pulling for Ken.