Dan's Papers - August 10, 2007 - Chess Champions Cannot Beat Chess Machines, No Matter What

Professional Chess players are still trying to prove that a man can beat a machine. They held another of these Man-versus-Machine chess matches last week, this time in the Dutch city of Maastricht, where one of the best human chess players in the world, Jaap Van den Herik, met the most powerful chess playing computer in the world, called Rybka.

It was no contest. Over the last forty years, as scholars debated whether a machine could ever win such a match, the machines slowly crept up in complexities, portability and intelligence, while men just made do with the old, grey-matter status quo. If man was waiting for evolution to up his game, it was going to be a long wait. The tipping point came in a famous contest in 1997. Before that, when the best chess players in the world played a machine, the man always won. After that, it was the machine the winner, ever widening the gap between the two. Many can clearly recall that fateful day on February 6, 1997 when in New York City, the world champion at that time, Boris Kasparov, stood up at the end and angrily declared that he simply could not beat Big Blue. And that was that. Then he went off and took a nap.

But humans live in hope, which is something that cannot be said for machines. So when Van den Herik sat down on one side of the table to face Rybka, the fix was on. Or so they thought.

The match was played over a two day period. There are a lot of draws in chess, where both sides get themselves in a place where neither can win. And as always, each player would get half of a point for a draw and one full point for a win. The first to get to five would be the winner.

The fix consisted of an arrangement whereby the beginning of the game would allow van den Herik, whose brain has the fantastic ability to think ahead to every possible outcome for up to the first eleven moves, to strut his stuff.

This would not be the case for Rybka. Rybka has been programmed to think ahead to every possible for up to the first twenty-five moves. Not fair. They'd go in -- they did go in -- and in ten minutes they had performed the mechanical equivalent of a partial lobotomy, wiping out twenty-three of those twenty five possibilities. So Rybka was essentially blinded to the beginning of the game, except for a little spot where he could peek out.

Of course, Rybka fell behind at the start, but then he rallied into the middle game and ultimately was able to force a draw. He (it -- I guess it's an it) then did this four times, so the score was two to two and then, it seems that it experienced a learning curve. In the fifth game, it fell behind, caught up and then won. And it did so again in the sixth by a wider margain. Now the score was four to two.

Time out, said van den Herik's handlers. They talked to their fading champion, they wiped his brow, they massaged his wrists and his neck, they talked to him again. And then they talked to the referees.

The problem was, the handlers said, that van den Herik was playing this chess match with one of the participants under a handicap. That the handicap was not imposed on van den Herik was not the point. The point was that van den Herik was not used to playing in a handicap situation. It made him feel stressful and, feeling stressful, it put him off his game.

As far as Rybka was concerned, of course, there were no feelings involved one way or the other. He was a goddamned machine. He felt no effects of the fact that there was a handicap.

And so it was decided that there would not be anymore handicap. Van den Herik would play with what he knew. And Rybka could play with what it knew (which at that point, since there was no time to program what had been removed, meant only the first two moves.) Furthermore, van den Herik, because he had just suffered through a total of six games where this handicap business had been bothering him, would now have it all made up to him by having Rybka forced to open every game playing the black pieces.

In case you didn't know, in chess, the white always moves first. It's a terrific advantage, similar to the advantage one gets in tennis when one has the serve, with the opponent back on his or her heels.

On the other side of the table, Rybka's handlers said they'd like to discuss this new proposal. It was not lost on anybody that all of Rybka's handlers were human. Indeed, everybody in the room was human, although that didn't bother Rybka.

In another setting, those that went along with this sort of discrimination were called "Uncle Toms." But I guess we can just leave that there. Okay, said Rybka's handlers.

And so van den Herik moved first and fought the game to a draw. 4 1/2 to 2 1/2. And then van den Herik opened the next game with white and Rybka beat him silly. Game over. Match over. 5 1/2 to 2 1/2 in favor of the machine.

Are there lessons here? I don't know. Maybe it's that humans, no matter how often beaten, still think that being atop the food chain on this planet means you're the best. So if they can't win one way, they'll win another. If they can. Winners cheat. Anyway, Rybka is only a machine so he doesn't know there is cheating going on. Also, isn't it true that we humans CREATED Rybka?

The secret to beating Rybka, it seems to me, is as soon as it becomes obvious that Rybka is going to win, pick him up and hurl him against a concrete wall as hard as you can. That'll do it.

By the way, I recently read an article about a new computer that has been programmed to play checkers so well that it cannot lose. People have tried to beat it, of course. But as checkers is a fairly simple game with a fairly limited number of possibilities, it is perfectly possible to build one that, no matter what happens, will always win because there are no possible losing moves it could make.

In this article, I read about the greatest checkers player of all time. A human of course. Named Marion Tinsley. He played thousands of games of checkers without ever having lost a single one. This was from the time he was twelve years old. He played more than 5,000 games. Won 'em all. And although it was thought possible to beat him -- he was not "programmed" after all -- nobody ever did. He'd always find a way.

He retired at 43, undefeated. This was around 1960. And he didn't play anymore checkers for the next 15 years. Then he was lured out of retirement to play the best of the next generation of checkers players and he beat every one of them too, winding up with 6,000 wins and no defeats.