Deep Blue Beats Kasparov in Chess

Date February 10, 1996

Deep Blue, an IBM computer designed to play chess at the highest competitive level, stunned the chess world by defeating World Chess Champion Garry Kasparov in the first game of their 1996 match. Kasparov swept the rest of the match without losing another game. However, in their 1997 rematch, the computer triumphed.

Locale Philadelphia, Pennsylvania; New York, New York

Key Figures

• Garry Kasparov (b. 1963), world chess champion, hailed by some as the greatest player in chess history
• Feng-hsiung Hsu (b. 1959), principal designer and team leader in the development of Deep Blue
• Murray Campbell (b. 1967), expert chess player and computer scientist active in the development of Deep Blue
• A. Joseph Hoane, Jr. (b. 1962), software engineer who joined the Deep Blue team when it moved to IBM
• Joel Benjamin (b. 1964), international chess grand master who helped expand Deep Blue’s chess knowledge

Summary of Event

Building a chess machine that could compete with the best human players had been a goal of computer scientists interested in artificial intelligence since the 1950’s. Mastering chess seemed an ideal task for computers, since the game involves a limited number of physical objects governed by simple, clearly defined rules that can be quantified. A properly programmed computer can methodically consider various continuations of a given position and assign numerical values to every possible move, identifying the best next move. By 1980, several chess machines could defeat all but the most expert human players. That year, computer chess proponents established the $100,000 Fredkin Prize, which would be awarded to the creator of the first computer to defeat the world chess champion under standard tournament conditions.

Deep Blue began as chess-playing chips designed by Feng-hsiung Hsu as his doctoral dissertation in the computer science program at Carnegie Mellon University. Once he started designing chess machines, Hsu set out to win the Fredkin Prize. When Hsu graduated in 1989, International Business Machines (IBM), aware of the publicity value to the company if he succeeded, hired Hsu and his associate, Murray Campbell.

In 1991, IBM changed the name of the machine to Deep Blue (previous versions had been called Deep Thought), echoing IBM’s “Big Blue” nickname. Hsu developed a more powerful move-generating computer chip. Assisted by Campbell and IBM engineer A. Joseph Hoane, Jr., he embedded more than two hundred chips in IBM’s newest supercomputer, the RS/6000 SP2. Harnessing the power of thirty-two computers working together through parallel processing, the machine could examine over 100 million moves per second.

When Deep Blue defeated World Chess Champion Garry Kasparov on February 10 in the first game of their 1996 match, chess aficionados reacted with shock. No chess machine had ever won against Kasparov before. However, the champion swept the Philadelphia match, tying two and winning three of the next five games. Deep Blue’s team went back to their lab to improve their machine for a planned rematch.

For IBM, Deep Blue’s first victory against Kasparov more than justified its sponsorship of the chess machine. News of the triumph was on the front page of major American newspapers. A flood of favorable worldwide newspaper, magazine, and TV coverage greeted the win and continued even as Kasparov dominated the rest of the match. IBM willingly agreed to a rematch, supported its computer chess team in developing an even more powerful version of Deep Blue, and offered a $1.1 million purse, $700,000 of which would go the victor.

Hsu worked hundred-hour weeks designing another new chess chip with improved position-evaluation functions, while Campbell and Hoane modified the software to take advantage of the new chip. Grand master Joel Benjamin played test games against the machine, uncovering gaps in its chess knowledge that the engineers corrected, thereby increasing the acuity of Deep Blue’s positional analysis. Commentators pointed out that IBM provided a new supercomputer that was twice as fast as its 1996 predecessor, able to consider 200 million positions per second. However, most commentators undervalued Hsu’s improved chip, as well as the increased chess knowledge added to Deep Blue’s evaluation procedures by its chess instructor and engineers.

More than two hundred reporters representing every major news organization, along with television crews from around the world, arrived in New York to cover the 1997 rematch. Expert chess commentators unanimously predicted an easy win for Kasparov; one even asserted that Kasparov would win every game, since his victories in the final two games the previous year proved that he had learned to take Deep Blue’s measure. When Deep Blue resigned at the forty-fifth move of their opening encounter on May 7, Kasparov had won three games in a row; expectations of an easy triumph for the world champion appeared validated.

The second game upset that assumption. This time, Kasparov resigned on the forty-fifth move, believing the computer had a clear path to victory. However, to Kasparov’s dismay, chess aficionados who had been following the match over the Internet discovered a move he had overlooked that would have forced Deep Blue to concede a draw. News of his slip shook Kasparov’s confidence and may have affected his play during the succeeding games.

Games three, four, and five ended in draws, leaving the match tied. The sixth and final game would be decisive. The audience at the press conference following the fifth game demonstrated their bias, applauding Kasparov and booing Deep Blue’s operators. The incident deeply upset Hsu, whose account published five years later showed that his hurt feelings still smarted.

Unlike the previous lengthy, heavily fought contests, game six on May 11 lasted less than one hour. Some commentators termed Kasparov’s surprising seventh move a blunder, but others considered it a bold attempt to present the computer with an unexpected situation it was not programmed to counter effectively. The move failed to confuse Deep Blue, which relentlessly pressed Kasparov into an untenable position, forcing him to resign after nineteen moves.

Significance

For IBM, the outcome of the chess matches was an unalloyed success; once again the company appeared to lead the way in computer technology and innovation. Sales of supercomputers boomed. Favorable publicity sent the price of IBM stock upward, adding billions to the value of the company and to the net worth of stockholders. Aware they could not better their position, IBM retired Deep Blue, shifted Hsu and his team members to other duties, and assigned the computer to research problems. On July 29, 1997, Hsu, Hoane, and Campbell received the Fredkin Prize, splitting the $100,000 award. IBM kept Deep Blue’s $700,000 winner’s share for its research funds.

Kasparov accused IBM of setting up a hostile environment that damaged his thinking, and he angered Deep Blue engineers by hinting that a hidden human, rather than the computer, selected the computer’s most damaging moves. Chess fans who had cheered Kasparov’s triumph in the first 1997 game turned negative, criticizing him for not exhibiting his normal level of skill. They condemned the quality of chess displayed by both champion and computer, saying the games lacked interest.

A few computer analysts claimed that Deep Blue had demonstrated true artificial intelligence in defeating Kasparov, but most commentators, including Hsu, disagreed. Hsu insisted that the only intelligence displayed at the chess table was Kasparov’s; the computer simply followed rules laid down by its designers. The engineer also rejected the idea that the matches were a contest of man versus machine. The real contest, Hsu asserted, was between one outstanding man and the men who programmed the machine. To his satisfaction, the programmers triumphed.

Bibliography

Hsu, Feng-hsiung. Behind Deep Blue: Building the Computer That Defeated the World Chess Champion. Princeton, N.J.: Princeton University Press, 2002. The primary designer of Deep Blue narrates its development from a doctoral dissertation topic to its triumph over the world chess champion.

Newborn, Monroe. Kasparov Versus Deep Blue: Computer Chess Comes of Age. New York: Springer-Verlag, 2003. Easily understandable narrative of the two matches by a computer-chess proponent.

Pandolfini, Bruce. Kasparov and Deep Blue: The Historic Chess Match Between Man and Machine. New York: Simon & Schuster, 1997. Chess instructor Pandolfini provides move-by-move analysis of all six games of the 1997 match.