Computers work best with humans, not against them

A captivating real-world application to advanced computer science was the motivation when IBM assembled and programmed the chess-playing Deep Blue supercomputer to challenge Garry Kasparov, the world's chess champion, in 1996.

Chess was a good vehicle for complex computations. The game's rules are easy; playing at a competitive level is very difficult. The exercise would demonstrate how computers can partition large problems into small pieces."But there are only so many times you can tell that story and maintain credibility," said Gabriel M. Silberman of the IBM Centre for Advanced Studies at the IBM Toronto Laboratory.

The bottom line for IBM's promulgation of the Deep Blue chess project comes from the "B" in the company's name - business.

IBM stock reached a five-year high during Deep Blue's February 1996 match with Kasparov, and the event saw record-setting Internet traffic. Kasparov emerged victorious over the machine, captivating world headlines, and he promised a rematch. It was only fair, he said.

The rematch came in February 1997. A smarter, reprogrammed Deep Blue won this time in the sixth and final game. Kasparov hit the roof and IBM stock hit an all-time high.

Kasparov promised to crush the machine in the next rematch. IBM said, "No, thanks." It was time to move on. Today IBM is not among the vendors of commercial chess-playing software; Deep Blue's six-human team has been disbanded; and Deep Blue and its IBM peers are making computations to cut pharmaceutical development time in half and simulating nuclear weapons tests, as IBM announced last week.

Postmortems to the famous matchup of the world's sharpest chess brain and a supercomputer crammed with what Silberman called "chess steroids" reconfirm that computers are most productive when they work with humans instead of against them. The University of Utah computer science department hosted a colloquium last week where Silberman offered highlights of the first match that were once Deep Blue secrets.

The computer beat a human in world-class play for the first time in the first game of the first match. But as the second game opened, Deep Blue befuddled its programmers right off the bat. All of the right files were in place, but programmers hadn't made an important file accessible to the computer, so Deep Blue played the entire game based on other capabilities and lost.

The third game ended in a draw. So did the fourth. Kasparov unexpectedly offered a draw early in the fifth game. Deep Blue could analyze two million chess moves a second but it couldn't negotiate the offer, Silberman said.