A front-row seat at the main event

Jonathan Schaeffer had made up his mind, and that was final: he wasn't going to New York to work as a referee in one of the greatest chess matches of all time. Still, the organizers were relentless.

"I told them a few times that I couldn't go," said Schaeffer, one of about 100 chess grand masters in Canada, who holds a Canada Research Chair in artificial intelligence at the U of A.

It's understandable why the organizers were so interested in Schaeffer. During the 1980s Schaeffer developed what was then considered one of the world's best chess programs, called Phoenix. When IBM's Deep Blue computer appeared certain to topple his own program and perhaps defeat Kasparov (which it did, in a non-sanctioned 1997 match), Schaeffer changed his focus to checkers. The move paid off - his program won the world's first machine-vs.-human world title, winning the title from then-reigning checkers champ Marion Tinsley in 1994.

With that background, Schaeffer seemed a natural choice for the job. But he was busy. He had classes to teach. And, he admits, he was worried. What if he made an unpopular ruling? "I've adjudicated things at the local tournament level and stuff, but to be asked to go to a tournament with the greatest player in the world and maybe make a decision that would make a lot of people unhappy. but they twisted my arm, and twisted my arm, and twisted my arm. I'm glad I went."

The assignment gave Schaeffer a front-row seat at an historic event. Kasparov had been humiliated by his loss to IBM's Deep Blue, which was capable of analyzing 200 million moves per second. Deep Junior, developed by Israeli programmers Amir Baran and Shay Bushinsky, beat 18 other computers in the July, 2002, world championships of chess-playing supercomputers. It is capable of calculating three million moves per second.

Kasparov is capable of calculating a paltry one or two moves per second.

Schaeffer says Kasparov's talent is the perfect illustration of human supercomputing. Chess depends on two elements: search and knowledge. The computer's strength is search; the human's is knowledge.

"Search is simple to explain - it evaluates space," said Schaeffer. "Knowledge is an understanding of the game, and Kasparov has a lot more knowledge about chess. Humans are very good with knowledge."

Computers are better equipped to evaluate possible results of a move by determining an opponent's reply to a move, the computer's own best response, and the player's subsequent reply. The computer, Schaeffer says, is considering scenarios about "10 or 12 moves deep," making them efficient players.

"Kasparov knows the computer is looking far ahead and he knows that if he makes a mistake the computer is going to pounce."

In spite of its capabilities, the computer worried its programmers at one point. In an effort to confuse Deep Junior, Kasparov employed an unorthodox strategy. Once, he moved his king twice in three moves. Following one such measure Deep Junior paused to consider its options. Twenty-five minutes passed. Deep Junior appeared to have blacked out. Its programmers considered rebooting the computer before one of them shook the mouse to discover the problem was that Deep Junior's screen saver had been activated.

But the computer impressed onlookers with a style of play that Schaeffer described as "human-like," avoiding the "ugly" moves that often characterize computer chess.

"There were only a handful of times that you could tell which player was Kasparov and which player was Deep Junior," he said.

Kasparov, the world's No. 1-ranked chess player since 1984, won the world championship in 1985, losing it for the first time last year. He has a style of his own. When Kasparov is certain he has won a game and victory is imminent, he removes his watch and places it on the table next to the board. He shows his emotions clearly.

"He wouldn't be much of a poker player," said Schaeffer. "He's very emotional, and that's one of the reasons he is so popular with the media. I saw him at a time when he was very unhappy. In Game 3 he made a trivial oversight and lost - you could see he was angry with himself."

In the following game Kasparov ultimately offered Deep Junior a draw, which its programmers accepted.

"It was a great result for Kasparov and a great result for the programmers, who didn't have the horsepower of Deep Blue," said Schaeffer, who is obviously glad he accepted the offer to work at the match.

"For me, to be able to be behind the scenes, meet Garry Kasparov and talk to him, and to meet some of the elite of the chess world and to be a part of history - it was thrilling."