The triumphant teamwork of humans and computers.

Science, technology, and life.
May 11 2007 6:05 PM

Chess Bump

The triumphant teamwork of humans and computers.

Illustration by Rob Donnelly. Click image to expand.

Ten years ago today, a computer beat the world chess champion in a six-game match. Since then, human champs have played three more matches against machines, scoring two draws and a loss. Grandmasters are being crushed. The era of human dominance is over.

William Saletan William Saletan

Will Saletan writes about politics, science, technology, and other stuff for Slate. He’s the author of Bearing Right. Follow him on Twitter.

It's not just chess. Everywhere you look—switchboards, ATMs, post offices—machines are replacing us. First they took farming and manufacturing jobs. Then they took service and office jobs. Chess was supposed to be a bastion of human ingenuity, an art they'd never conquer. Now they're conquering it. The smarter they get, the more we feel threatened.

Advertisement

Don't be afraid. We, too, are getting smarter, and computers are a big reason why. They're not our enemies. They're our offspring—our creations, helpers, and challengers.

We certainly needed the challenge. Chess computers, in particular, have exposed our complacency. Grandmasters used to dismiss computers as calculators, unfit for elite competition. Our vanity was so blinding that in 1997, when world champion Garry Kasparov lost to a machine called Deep Blue, he implied that the computer had received human coaching during the match.

Computers kept winning, and we kept whining. Commentators sniffed that the machines were making oafish moves and being generally outplayed. In postgame press conferences, players swore they'd been winning right up until the moment when, for unclear reasons, they lost. Five months ago, the current champ, Vladimir Kramnik, overlooked an instant checkmate by his artificial opponent, Deep Fritz. "I rechecked this variation many times and analyzed quite far ahead," Kramnik protested. "It seemed to me I was winning."

Kramnik's blunder was no accident. It happened because of flaws in the human brain. We thought we were smarter than computers for two reasons. First, we could choose a goal and figure out how to get there, whereas computers had to start with the available moves and see where they led. Second, computers had to think through every possible move, whereas we could recognize crucial patterns and focus on the moves that mattered. But that's why Kramnik missed the checkmate: It looked different from the usual threat pattern, and he was thinking too far ahead. Even the best brain sometimes needs computer assistance.

The remarkable thing about us isn't our supremacy over computers. It's our interaction with them. Yes, chess programs have been getting smarter. But they didn't do that on their own. Humans design the hardware and write the code. Grandmasters test and refine it. The machines get smarter because the code gets subtler because the programmers get wiser. That's how Deep Junior, the machine that played Kasparov four years ago, eclipsed Deep Blue's skill with just a fraction of Deep Blue's computing power.

In the old days, chess programs went around killing enemy pieces at every opportunity. Their human opponents understood that in chess, like war, other factors often matter more: territorial control, mobility, initiative, reach, coordination, supply lines, impregnability, and safety from decapitation. By trading material for these advantages, the humans won. So, programmers taught the machines to recognize and consider the same factors.

Unable to win with their old tricks, human players learned new ones. They played quirky openings to throw computers off-script. They plotted attacks a dozen moves ahead, beyond the machines' range of calculation. They hunkered down in defenses that to a computer looked impregnable. They cluttered the battlefield with obstructions, making it harder for computers to see threats or payoffs. They left irrelevant pieces on the board to absorb the machines' attention. It was a whole new game layered on top of the old one. The humans named it "anti-computer chess."

Now programmers are adding a third layer: anti-anti-computer chess. They're teaching machines to break old habits, see through clutter, and force the wide-open bloodbaths at which computers excel. In 2003, Deep Junior flummoxed Kasparov with a kamikaze attack unprecedented in computer annals. Last year, when Kramnik forced Deep Fritz off its opening script, the program invented a new variation and went on to win the game.