Deep Blue beat Gary Kasparov at chess, and soon computers may be winning money from Alex Trebek.
On Thursday IBM demonstrated Watson, a computer consisting of a roomful of the company's Power7 processors. But the real genius in the system is that it can understand language well enough to figure out the wordplay that makes up a Jeopardy! question. On Feb. 14, 15 and 16 Watson will take on two Jeopardy! champions -- - Brad Rutter and Ken Jennings.
The grand prize for the competition will be $1 million, with second place earning $300,000, and third place earning $200,000. Rutter and Jennings will donate 50 percent of their winnings to charity, and IBM will donate 100 percent of its winnings to charity.
A chess computer needs only to be able to do math. But a statement like This president said he was a jelly donut, would stump most computers, who have a lot of trouble with the implied meaning of words. (The answer is who is John F. Kennedy?).
Watson works by running many algorithms at once, that govern how it processes the language. The computer then learns, by checking its algorithms against the answers it gets. In that sense it studies in a similar way to humans.
For people watching, a graphic showed the answer Watson thought was most likely to be correct. For example, asked A classic by Crockett Johnson, Harold and the blank crayon Watson thought the answer purple was 98 percent likely to be correct.
The demonstration, in which Watson played Jennings and Rutter, was a half-game of Jeopardy! (15 questions) and the computer did well, beating the two human contestants. Watson won $4,400 versus Jennings' $3,400 and Brad's $1,200 in simulated cash.
IBM spokesman Michael Loughran said unlike Deep Blue, IBM isn't just doing a technology demonstration, as it did with Deep Blue. The system will be sold to companies in fields such as health care, in which natural language processing is important.