New and improved When Kasparov whipped an earlier version of Deep Blue in February, 1996, he won by exploiting weaknesses he detected in the first game. In a sense, he "psyched out" the computer, recognizing that if he played unconventionally, the machine would be forced outside its programming. After the defeat, IBM's team of Deep Blue wizards added some powerful features to their 1.5 ton chess player. They

		doubled its processing power, giving Deep Blue the astonishing ability to evaluate a billion positions every five seconds;
		enlarged its database of openings and end games with a vast reference library of championship games;
		brought in a series of human grandmasters to challenge the computer and find weaknesses; and
		found new ways to make the computer's 512 processors collaborate -- to "tell" each other when a position had already been examined, for example.

Kasparov's mistaken seventh move: pawn to h6, in front of black's right-hand rook. A better move is bishop to d6, shown by the incredible whirling cleric. .

It "played like God" The result of this tinkering was a machine that overwhelmed Kasparov, who at one point said it had "played like God." Kasparov was so flummoxed that he did something entirely uncharacteristic in the last game -- made the kind of mistake that top players don't make. Deep Blue's triumph -- the first victory for a computer over a world chess champion in a regulation match -- was clearly a milestone in the 50-year-old struggle for dominance between computers and chess players. But what did the victory say about the state of AI? After all, the computer's strength was not what most people would consider intelligence, but its "brute force" calculating ability. And brute force is not how we humans play chess. Remember that Kasparov can evaluate two or three positions per second, while Deep Blue can handle 200 million. So if their chess-playing strategies were equally efficient, Kasparov would need 66 million seconds -- more than two years -- to examine those 200 million moves. And since chess players get an average of three minutes to make a move, Kasparov would have to stare at the board for 360 years to examine the moves Deep Blue evaluated in three minutes. Yet since Kasparov won one game and tied three more, he was nearly as good as the computer, and thus used a far more efficient strategy. Questions like these make observers a trifle uneasy with Deep Blue's victory. "It was partly brute force, and partly some very clever programming," says David Waltz, president-elect of the American Association for Artificial Intelligence. "It's very difficult to get a lot of different computers working together." Still, he says, the core of Deep Blue remains brute force. "Are we really closer to a genuinely intelligent computer? That is more debatable." So let the debate begin.

There are 1 2 3 4 5 6 7 8 documents.
Glossary | Bibliography | Credits | Search