IBM has a reputation for building PCs to take on humans; first it was Deep Blue versus Garry Kasparov at chess and now 'Watson' is set to take on competitors on the US game show Jeopardy! The announcement, by Jeopardy! host Alex Trebek, doesn't mention when the challenge will take place, but it does feature several confident-sounding IBM representatives.
For those of us not familiar with the programme, Jeopardy! is a quiz show on which contestants must answer trivia questions on topics such as history, literature, politics and pop culture. Think of it as a televised game of Trivial Pursuit.
The twist, though, is that these questions aren't answered in a straightforward manner, but rather employ riddle, irony, ambiguity and other such devices to increase their complexity. Importantly, while humans are naturally very good at coping with such complexities, computers, by design, are not.
Competing with a professional chess player may have been a challenge, but it was one that could ultimately be beaten by brute force as Deep Blue was calculating the best answer to a fixed problem. There are, after all, a finite number of moves possible in a chess game exist and calculating the right one to make is a matter of calculating statistics; a task computers excel at.
Interpreting natural language questions, conversely, is not a task computers are designed to cope with. Watson will have to interpret the queries presented to it to determine what question is actually being asked, then determine what answer is required to that question and deliver it faster than a human opponent can. As project leader David Ferrucci points out: "The challenge is to build a system that, unlike systems before it, can rival the human mind's ability to determine precise answers to natural language questions and to compute accurate confidences in the answers."
Or, in other words, Watson will - rather like a human does - draw up a shortlist of possible solutions to the problem presented and decide which is the right one. It sounds simple, but it isn't.
It's important to point out that Watson will be working on its own and won't have a connection to the internet. Presumably IBM is going to be kitting the computer out with a hefty data array to work with, though. Something along the lines of Encarta, the Encyclopaedia Britannica and Wikipedia rolled into one, I'd expect.
It's a hugely interesting challenge and I can hardly wait to see how IBM fares!