After Chess Romp, Deep Blue's Makers Mull Their Next Move
One designer seeks to create a personlike robot but admits he's decades away from it.
NEW YORK — What next?
For 40 years, one of the grand challenges of artificial-intelligence research has been to beat the reigning chess champion in a full match. On Sunday, computer scientists notched that victory, beating Garry Kasparov here in New York with a stunning victory in the sixth and final game. But it's not clear what their next challenge will be.
Should they tackle more difficult games? Or, having built an artificial chess player, should they try to build an artificial person? Computer scientists doubt the first will capture the public imagination or draw the necessary resources. And they blush at the second idea.
Human intelligence is so far beyond a super-calculating chess machine that it's like asking the conqueror of Mt. Everest to travel to the next galaxy. Even then, scientists admit there is more to man than mere intelligence.
"If you took all of the computers in the world and put them in parallel, you could not approach the capabilities of the human being," says Joseph DeBlasi, executive director of the Association for Computing, based in New York.
"My holy grail is a robot who would work like a person," adds Paul Rosenbloom, deputy director of the intelligence systems division at the Information Sciences Institute at the University of Southern California.
But he and other scientists admit that they are decades and perhaps even centuries from building such a machine. It would have to combine such varied abilities as holding an interesting and open-ended conversation, fully vacuuming a house without bumping into things, learning anything from physics to music, and having the ability to compose a piece of music and then improve it after someone criticized it.
Ironically, it was the unexpected loss by Mr. Kasparov, the chess champion, that showed how far computers have to go before becoming peoplelike.
It will take months and perhaps even years to discover whether IBM's Deep Blue computer is really the world's best chess player. In his games, Kasparov played unusual openings in an attempt to confuse the machine. Except in Game 1, where he won, the gambit didn't work. Hours after the match, Susan Polgar, the women's reigning chess champion, challenged Deep Blue to a match, suggesting a woman's intuition might lead to a more interesting match against the computer.
But the IBM team that built Deep Blue seems not to have decided whether to take up the gauntlet of competitive chess. "So far we've been doing science in the laboratory, constructing computers," says Chung-Jen Tan, leader of the Deep Blue team. "If you want to take us away from that and become a professional chess player, that's a very interesting thought. We'll have to think about that."
IBM's main goal in the match, he adds, was to demonstrate the power of parallel processing - how coordinating many computers to work on a single problem can solve very complicated problems, such as chess. The same process can be applied to various real-world problems, such as financial modeling, which will eventually affect consumers, Dr. Tan says.
But is building a Wall Street wizard the next grand challenge for artificial intelligence? It might well attract corporate dollars, but computer scientists aren't sure it would inspire a new generation of researchers. As an epic battle of wits, the chess confrontations had a romantic appeal.
The end to Sunday's battle came quickly. Kasparov appeared to blunder early in the opening of the sixth game and resigned suddenly after only 19 moves, stunning the audience and chess commentators who were following the game. At his press conference afterward, he was alternately humble and combative.
He said he was inadequately prepared. He claimed he could beat the machine under regular championship conditions. At one point, he appeared to imply IBM had given its Deep Blue computer human help during the games, but then backed off, saying only that the machine's internal calculations needed to be made public and studied.
"I proved to be vulnerable," he admitted, especially after his loss in the second game, when he resigned even though the computer calculated that he could have gotten a draw. "When I see something that is well beyond my understanding, I'm scared. And that was something well beyond my understanding."
The chess world was also taken aback that its champion had crumbled so easily.
"Garry Kasparov seems to me to have been really afraid after Game 5 and he gave himself an excuse to lose this final game," says Patrick Wolff, chess grandmaster and former US champion. "He could say that this final game wasn't a real competition, didn't really count, something or other - the same kind of psychological nonsense that I've come to expect from myself and from other chess players, but never Garry Kasparov before."