Computer heralds draughts of change

Click to follow
The Independent Online
MAN against machine for the official world title in a game of pure thought. This science-fiction scenario became reality at the Park Lane Hotel in London last week.

Draughts is often seen as the weedy little brother of chess, but for both humans and computers its possibilities are vast. If the Chinook computer program can add to its two victories and go on to beat the world champion, Dr Marion Tinsley, over 40 games, it will be a significant achievement.

However, the match took an unexpected turn yesterday when Chinook seized up in the middle of the 18th game. Play was suspended as its programmers and engineers tried to sort out the fault, but they were unable to start it again immediately and had to resign the game. Chinook was later able to play two more games, both drawn, leaving the score at two wins each with 16 draws at the halfway stage.

Dr Jonathan Schaeffer, Chinook's programmer, said: 'It's like asking what is the scientific contribution of climbing Mount Everest. It's nothing. But it's a milestone. What we can really say is that we now know we can solve a certain level of complexity and therefore there's no point in tackling anything simpler.'

Dr Schaeffer, 35, specialises in machine intelligence at the University of Alberta in Canada. In the mid-1980s he wrote one of the world's strongest chess- playing programs, but had good academic reasons for changing games. 'Computer chess was so competitive. The bottom line in computer chess is the result: 'My program finished half a point ahead of yours in a tournament' - that was all they cared about. That's not research. It's not advancing anything.'

Competitiveness, he says, is the worst thing that has happened to computer chess. He cites some recent work proving that, from any normal position, king, rook and bishop can always defeat king and two knights, but it may take at least 220 moves. 'Absolutely fascinating, but utterly irrelevant,' he says. Apart from the fact that either player could claim a draw after the first 50 moves, that ending has never occurred in a century of grandmaster practice. 'Millions of dollars of computing time are being spent on computing databases, and all it does is further our knowledge of chess. It doesn't further our knowledge of anything else in any way.'

Last year, Chinook (named after a wind in the Rocky Mountains) won the right to challenge Dr Tinsley for the world title. 'The moment that we earned the right to play Dr Tinsley it became a tremendous pressure to win. All of a sudden it didn't become so much of a research effort.'

Draughts (called 'checkers' in the United States) is, to Dr Schaeffer, a measure of the present limits of complexity that computers can handle. 'There are five times 10 to the twentieth (that is five followed by 20 zeros) possible checkers positions. There's absolutely no doubt that today, with the proper computing equipment, we could solve the game of checkers. We could enumerate every single one of those checkers positions, store them somewhere and the computer would be 100 per cent perfect.'

All he needs is 10 years and a multi- million-dollar computer. And the work could have applications. 'There are a lot of problems in physics, for example, where you want to find the optimal solution to a problem and there are billions and billions of possibilities. Finding the optimal move (in draughts) is comparable to these problems.'

Chinook's successes, particularly its win in the eighth game against Dr Tinsley, have been praised as evidence of computer creativity; but Dr Schaeffer says: 'My program has zero intelligence. What I do is not create intelligence, but the illusion of intelligence. My computer is dumb. It'll only do what I program it to do. It plays checkers pretty well, but you want to ask it to play bridge? Until computers are capable of independent thought, there's no point in even talking about intelligence.'

But what if a machine does become the best draughts player in the world? Will chess be next? With 10 to the 44th possible positions, chess is a considerably bigger problem.

As for proceeding to the complexities of real life, Dr Schaeffer has still less faith in computers: 'The problems are so difficult that I don't think the things people are afraid of - like intelligent machines that can talk back and reason and do all sorts of things like the computer Hal in the film 2001 - will happen in my lifetime. In 100 years from now, maybe, but I'm not optimistic.'

(Photograph omitted)