IBM supercomputer to challenge humans on 'Jeopardy'
Chess was easy compared to interpreting questions with puns and other word plays
- By Kevin McCaney
- Sep 03, 2010
Supercomputers have beaten world champions at chess, but could they keep pace with the likes of Cliff Clavin on “Jeopardy”?
IBM wants to find out this fall, when its DeepQA supercomputer competes on the game show in a contest of human vs. machine.
DeepQA is a product of IBM’s Blue Gene Research Project, whose Deep Blue supercomputer — though only 259th on the Top 500 list of supercomputers at the time — played two celebrated matches in 1996 and 1997 with then-World Chess Champion Garry Kasparov, with the computer winning the second match.
At first glance, it might seem that competing on a quiz show would be a step down from matching wits with a chess champ. But chess has strict rules and a standardized way of describing moves. It’s a logical game, right in a supercomputer's wheelhouse. “Jeopardy” is more complex, at least in terms of preparing a computer.
Giving DeepQA — nicknamed Watson after IBM founder Thomas Watson — a broad base of knowledge to draw on shouldn’t be difficult. The challenge will be interpreting the questions to find the right answers. And then, of course, buzzing in on time.
The clues on "Jeopardy" involve "irony, riddles, analyzing subtle meaning and other complexities at which humans excel and computers traditionally do not,” IBM said in a release. The computer’s performance will be the product of years of research in natural-language processing and demonstrate how quickly it can scan its database for the most likely answer.
The date for Watson’s appearance hasn’t been announced, but we’ll be looking for it. Let’s just hope Watson knows better than to bet it all during Final Jeopardy.
Kevin McCaney is a former editor of Defense Systems and GCN.