Big Blue's Hand Of God

BEFORE THE CLASSIC REMATCH of Kasparov versus Deep Blue - the man-against-machine conflict that pitted the world's best chess player against an electronic foe - one thing was agreed upon by all parties and observers. The computer was not a ""thinking machine.'' Whatever occurred in the nether world of its silicon circuitry, the process bore no relation to human cognition. The six-game battle, waged in a room rigged as a television studio on the 35th floor of a midtown-Manhattan skyscraper, shaped up as strictly a test of one man's stratagems against an impersonal opponent.

Yet as the match unfolded, the psychological component became as prominent as in any human-to-human contest. And for the flesh-and-blood competitor charged with beating back this technological challenge to human superiority, an unsettling issue came to the fore: how do you respond to a computer that refuses to act like a computer?

The question was forced by Deep Blue's play on Sunday, May 4. Until then, everything had been going pretty much according to Garry Kasparov's plan, evolved after weeks of practice against Deep Blue's weaker cousins. The idea was to tempt the computer into situations where it had little to do, and would make wasted, or even foolish, moves. In Game One, Deep Blue had indeed blundered, and on the way to his win, the 34-year-old champion had even laughed at a machine misstep.

Game Two was something else. Deep Blue exquisitely played a classic opening, and never relinquished the pressure. ""You had the appearance that from the very beginning, Deep Blue saw the end,'' says IBM's hired-gun grandmaster Joel Benjamin. This of course was impossible - though IBM's chess prodigy can consider 200 million moves a second, there are so many possible variations that such omniscience is as unlikely as someone guessing where a particular atom might be on the moon. Situations that may occur 20 moves along are past the machine's ""horizon.'' But so fluid was its play that the grandmasters in attendance all understood Benjamin's contention that Blue ""played real chess'' that day.

What really shook Garry Kasparov, though, was a move that the computer didn't make. On Move 36, Blue had an opportunity to shift its queen to a devastating position - clearly the smart choice. Instead it took a subtler but superior tack that wound up to be near decisive in defeating Kasparov. After the champ resigned, he rushed back to his Plaza suite, cranked up his own computers and tried in vain to understand how a hunk of sand and metal could have understood chess so deeply. Apparently this puzzlement crossed into the realm of suspicion: the Kasparov camp was soon demanding to see Deep Blue's printouts. ""It was a masterpiece by computer,'' said Kasparov second Michael Khodarkovsky. ""We would like to understand why it was possible.'' (Eventually, a neutral arbiter examined the printout.) Later Kasparov blurted out his real complaint. ""Suddenly [Deep Blue] played like a god for one moment,'' he said.

His unhappiness was itself a victory for the Deep Blue scientists. They had their own psychological agenda for this match: to lure Kasparov into thinking that their contender was at heart just a faster, slicker version of every other chess-playing computer. It was no surprise that Kasparov would abandon his usual slashing attacks in order to exploit the computer's typical befuddlement at more subtle incursions. But Joel Benjamin carefully tutored the machine to avoid those traps. (In another sly move, IBM secretly recruited a second grandmaster, Miguel Illescas, to further plug Deep Blue's vulnerabilities.) Meanwhile, by not playing his normal game, Kasparov would require more time to plan his moves, and run into clock trouble.

Sure enough, after Deep Blue's ""Hand of God'' move, Kasparov began showing the effects of facing an opponent whose limitations were no longer certain. When the machine would make an unexpected, possibly foolish move, he'd look stunned. You could almost see his inner dialogue: ""Is this idiocy? Or something so brilliant that it only looks stupid?''

Deep Blue's failings are well documented, but human competitors have their own disadvantages. For one thing, they get tired, and Kasparov cited fatigue as a factor in managing only a draw in Game Four. But more important, only people can feel pressure. True, the match was no vacation for the hardworking Deep Blue programmers. But no one clamored to know what they'd had for breakfast, or whether they were depressed that they had made this move instead of that move. It was Garry Kasparov who was staring into the abyss that would come with being the champion who blew it for humanity. So what if Deep Blue is an inanimate object that knows how to do only one thing? It's Kasparov's thing!

Ultimately, the outcome of this rematch will determine its place in history. (After Game Five, an agitated Kasparov, forced into a draw to tie the match going into the final game, was still fixating on the issue of Deep Blue's printouts.) But this battle will also be remembered as a dramatic illustration of how spectacularly a mechanical opponent can surpass its assumed limitations - even if for a single God-like moment. As computers become more powerful, and as programmers become more successful at transforming those calculations into complex behaviors, we will see other examples of computers that don't ""act like computers.'' Consider carefully Kasparov's frustration. One day - very, very far into future, one hopes - it could be ours.