Not so smart (Kasparov, Computers and chess)

The Economist ^ | Jan 30th 2003

[Sawdring]

THE idea that chess-playing skill is a proxy for machine intelligence is not new. It goes back as far as 1770, when Wolfgang von Kempelen, a Hungarian inventor, unveiled a wooden, clockwork-powered mannekin at the court of Maria Theresa, Empress of Austria-Hungary. This machine, known as the Turk because of its exotic costume, could play chess, moving the pieces with a mechanical arm and defeating even the best human players. It was, of course, a trick—a hidden human operator controlled the automaton's movements—but some observers equated its chess prowess with intelligence.

This notion was revived in the 1950s, when the building of a genuine chess-playing machine was seen by artificial-intelligence researchers as a stepping-stone towards a general theory of machine intelligence. Claude Shannon, a computer scientist, explained why, in an article published in 1950. “The problem is sharply defined. It is neither so simple as to be trivial or too difficult for satisfactory solution. And such a machine could be pitted against a human opponent, giving a clear measure of the machine's ability in this kind of reasoning.”

You would have thought that such speculation would have ended in 1997 with the defeat of Garry Kasparov, the world's top-ranked chess player, by Deep Blue, a computer built by researchers at IBM. Mr Kasparov made much of the notion that he was defending humanity's honour; Newsweek called the contest “the brain's last stand”. In fact, it was no such thing. Far from being a step towards machine intelligence, as theorists had hoped in the 1950s, building a world-class chess computer has proved to be surprisingly easy, thanks to the plummeting price and soaring power of computer chips. Rather than emulating the complex thought-processes of human players, computers simply resort to mindless number-crunching to decide what move to make. Throw enough microchips at the problem—Deep Blue contained hundreds of specialist chess-analysis chips—and it does indeed become trivial. Quantity, as Mr Kasparov noted after his defeat, had become quality. He demanded a rematch, but IBM said no.

The World Chess Federation hosts the match between Garry Kasparov and Deep Junior.

Now Mr Kasparov has, in a sense, belatedly got what he asked for. Still the world's top-ranked player, he is playing a six-game match in New York against today's top computer program, Deep Junior. Inevitably, there has been renewed speculation about the implications for mankind if the best human player loses again to a computer. This is silly, for two reasons.

First, Deep Blue, Deep Junior and their sort are human creations. The real victors, if Mr Kasparov loses again, will not be machines, but the humans who designed and built them. Since machines are—so far, at least—unable to design and build improved versions of themselves, there is no need to worry about the world being taken over by chess-mad robots.

The second, more important reason is that we now know that chess-playing skill does not, in fact, equal intelligence. Nobody minds that cars can outrun the fastest athlete, or that cranes can lift heavier weights than the strongest man. Playing chess, it turns out, falls into the same category, despite its outward complexity: it is possible to get a dumb machine to do it better than any human. The equation of chess-playing with intelligence is centuries old, but it is time to lay it to rest.

[jgorris]

He lost to Krammnik in the World Title Match. Deal with it.

Kasparov is the highest rated player. Deal with it.

So because it's not FIDE, your assertion is that somehow it's not legitimate?

Don't take my work for it, go ask FIDE and Ruslan Ponomariov.

Krammnik was in his late teens when he lost many of those matches, and Krasparov was in his prime.

OK, following your logic, Kramnik is the one now in his prime and Kasparov is not. Therefore, Kramnik should now be dominating Kasparov, but he is not-- neither in terms of rating (which is a measure of playing strength) nor in terms of games against Garry.

Whatever. Hopefully we will see a re-unification match (which all three parties, Ponomariov, Kramnik and Kasparov have already committed to) and the question of who is the real world champion will be decided.

[MHGinTN]

The game to which I'm referring is one played by Fischer in the US Championship, 1956, I think, against Donald Byrne. He made a Queen sacrifice at about move 17 or 18 an mated his opponent in 24 more moves, forcing the action with six or seven checks culminating in mate at about move 41 or 42! I may have confused the name with another game of nearly equal brilliance ... Fischer played so many brilliant games, it is hard to keep them all in mind.

[Mr. Mojo]

Hopefully we will see a re-unification match (which all three parties, Ponomariov, Kramnik and Kasparov have already committed to) and the question of who is the real world champion will be decided.

I'd like to see that. Who do you think would (or will) win?