University of Alberta Edmonton, Canada May 30, 1997

Kasparov vs. Deep Blue -- the rematch

Rob Lake, U of A Webmaster, reports on his trip to New York to witness the most critical match between man and machine this century

By Rob Lake

The Brain's Last Stand.

This was the title of a recent edition of Newsweek magazine about the Kasparov-Deep Blue chess match. With the machine's recent victory, are we close to seeing a computer display the intelligence shown by HAL from 2001? I don't think so, but we should be impressed with Deep Blue.

For eight years, the IBM team (which includes a U of A alumnus Murray Campbell) prepared for this match against Kasparov. Two previous encounters, in 1989 and 1996, resulted in Kasparov victories. This time the Deep Blue team developed hardware capable of analysing 200 million positions a second, improved the program's position evaluation function and gave the program additional chess knowledge. Everybody knew the program was stronger-but the ultimate test lay ahead with Kasparov.

The 1997 match, played for a prize of $1.1 million, was held at the Equitable Center in New York City from May 3 to 11. All six games were played on the 35th floor in a room off-limits to the audience and most of the press. Spectators could watch the games on closed-circuit television in a 400-seat auditorium located in the basement. Although IBM initially provided a press centre on the 49th floor, this was relocated one floor above as media interest swelled.

Game commentary and analysis was provided by International Masters Maurice Ashley and Mike Valvo, and by Grandmaster Yasser Sierawan. Their analysis, anecdotes, and occasional special guests kept the games highly entertaining, especially for chess novices.

Game 1 was won by Kasparov. He played solid positional chess and capitalized on several time-wasting moves by Deep Blue. Many spectators and experts began talking about a Kasparov rout. My own prediction prior to the match was 4-2 for Kasparov.

Game 2 was probably the most spectacular chess game ever played by a computer. It was also the most controversial. Kasparov played passively and Deep Blue never gave him a chance for counter play. Kasparov finally resigned when the position looked hopeless, rather than play it to conclusion. Much to Kasparov's chagrin, subsequent analysis showed the game to be a draw.

This appeared to badly shake Kasparov's confidence for the remainder of the match. More importantly, Deep Blue had just played a near flawless game of chess.

Games 3, 4, and 5 were draws. In each game Kasparov had a strategic advantage and, in at least one game, missed a win. However, Deep Blue continued to find clever drawing resources, often to the amazement of everyone present.

Infamous game 6 ended after 75 minutes of play with Kasparov's resignation on move 19. In reality the game was over in 15 minutes and on move 7 when Kasparov accidentally transposed moves and fell into an opening trap. Everyone was stunned with this result.

In retrospect, I believe Kasparov lost the match for several reasons. First, game 2 bothered him and shook his confidence. He appeared increasingly nervous and demoralized as the week progressed. Second, he played too passively. In order to avoid complicated tactical positions against Deep Blue, he played very conservative and passive openings to get Deep Blue out of standard published play. I believe he should have played the openings he was most familiar with, and tried to win the middle and end game. Third, he underestimated the improvements made to Deep Blue. After game 2, he discovered he knew very little about the opponent he was facing due to all the changes made in the past year by the Deep Blue team.

Was this the brain's last stand?

No, it was a triumph in man's ingenuity to create a program to play chess at a level capable of defeating the strongest chess player ever. To be fair, most of the winning chances were with Kasparov and Deep Blue is far from mastering chess. And there are many intellectual areas where computers are far behind what we take for granted (such as reading and understanding this article). I doubt I will see a HAL-like computer in my lifetime.

As David Letterman says, humans still have the ultimate trump card. We can pull the plug at any time.

Folio front page

Office of Public Affairs

University of Alberta