ARTelevision

People and computers pair off in games of chess occasionally. They've been doing it for decades. Of course, computers are evolving exponentially. Each successive round of Grand Master vs. CyberBrain brings this all into focus.

It's happening again and it's worth our attention, I think.
..........................................................................................

Kasparov 'obliterates' Fritz with strategy

10:52 17 November 03

NewScientist.com news service

World number one Garry Kasparov "obliterated" the chess computer program X3D Fritz on Sunday with a strategy that proved too forward-thinking for the machine to defend against.

"He steamrolled right over Fritz," says John Hernandez, chess consultant for Fritz. "It was a masterpiece." Kasparov achieved a winning position after just 16 moves. But playing cautiously, he did not finish the program off until the 45th move and over four hours.

In the previous match, Kasparov lost by making a blunder a computer would never make. But on Sunday, the position was reversed, with Fritz displaying a weakness that dogs chess programs - being incapable of understanding a long-term strategy.

"Almost from the start Fritz did not understand what was going on and just shuffled his pieces around aimlessly," says Jonathan Schaeffer, in the computer science games group at the University of Alberta, Canada. "Kasparov won effortlessly without giving Fritz an opportunity to do anything."

Commentators made it clear during the match that a human player would have known how to counter Kasparov's play. He used a similar approach to break down program Deep Blue in 1996.

Sunday's result means Kasparov and Fritz are now tied in the four game series in New York, following a draw in game one and the victory for Fritz in game two. "I am in a very good mood," Kasparov said directly after the match.


Inherent weakness


The match went straight to the heart of inherent computer weaknesses and human strengths, says Schaeffer. A human player is capable of implementing a plan that takes over 40 moves to execute, whereas a computer can only take each move as it comes. The program does its best to improve its position, but each time all it has is to refer to is the moves it is capable of probing in advance.

Fritz analyses three million moves per second and uses smart software to probe promising moves much more deeply than less interesting ones. This is in contrast to the "brute force'" approach of programs such as Deep Blue, which attempts to calculate all possible moves, relying on massive computing power to achieve deep searches.

Fritz actually probed deeper this game than ever before, sometimes looking 19 half moves ahead. But nevertheless it still could not rival the long-term strategic skills of the human brain.

Without understanding what Kasparov was doing, Fritz was incapable of a defence. "All the searching in the world couldn't have made a difference," explains Schaeffer.


Closed game


Kasparov also played a very "closed" game by building a wall of pawns early on. This exploits another computer weakness. In a closed game, pieces interact very closely, meaning there are fewer choices per move.

This puts the human player at an advantage, because winning then relies more on having the best strategy. In contrast, computers play better in open games where the sheer number of possibilities to consider often foils humans.

Man and machine will face off one in the final match on Tuesday. Kasparov will be confident after Sunday's victory, but will have the disadvantage of playing black.

......................................................

Besides the thrill of wild abstract mental competition this story evokes, it also allows us to think about the cyberbrains we interact with every day. Can you feel yourself sort of adjusting to the way your computer(s) think?

I can. I can feel it every day. I think more and more like a logically organized device. And I think that's a good thing, in general.

How about you?

__________________
create evolution