Chess ace triumphs over the machine: Third time lucky as computer outflanks champion

The limitations of artificial intelligence were finally exposed in the semi-finals of the Intel World Chess Grand Prix in London yesterday, when the Pentium Genius 2 computer, which had eliminated two grandmasters from the competition, was beaten by Viswanathan Anand.

Mechanical greed brought its downfall in the first game. A pawn ahead, Genius stubbornly protected its ill-gotten gains, neglecting king safety and piece development. Anand cruised to victory with an irresistible attack. In the second game, another aspect of artificial stupidity was revealed as the computer blundered to defeat from a level endgame. Perhaps, knowing that it had to win to stay in the competition, the machine's operators had turned its 'risk' factor up too high.

After defeating the world number one, Garry Kasparov, in the first round, Pentium Genius 2 had gone on to eliminate Predrag Nikolic in the second round of an event restricted to the world's most highly ranked players. Nikolic's view, expounded with a cheerful smile was: 'At the moment we can still fight, but I don't know for how long. It has no feelings.'

'This will happen more and more,' said Nikolic. 'You cannot have luck against a machine.'

The machine has been ruthless at exposing human frailty. In the first game against Kasparov it was, by human standards, in difficulties. Under pressure on one wing, and with a bishop out of play on the other, it seemed to be slipping to defeat. However, a lapse by Kasparov allowed a surprising equalising manoeuvre, after which the champion seemed to lose the thread of the game completely.

'He wanted to crush the computer, not just beat it,' was Nikolic's diagnosis. And when crushing it became impossible, Kasparov was not able to adjust under the stern time limit of 25 minutes for each player for the whole game.

In the second game, the machine was again in trouble, a pawn behind for nothing, but it threw so many difficulties in the champion's path that eventually Kasparov weakened and threw away his advantage. The game was drawn, and meant Kasparov's exit from the tournament.

Warned by Kasparov's experience, Nikolic kept things simple and exposed some glaring weaknesses in the computer's chess understanding.

The machine, however, showed that it is a fine hustler. After losing an endgame, it began to move very quickly indeed and completely bamboozled its human opponent.

Nikolic saw his winning position deteriorate into a draw, then slip to a loss, and the machine simply calculated the possibilities far more quickly than any human could manage. 'It has no shame,' said Nikolic. 'It plays stupidly from time to time, but it doesn't care. It doesn't know.'

The second game was a similar story: an uneventful start, some pointless moves from the machine, a winning position for Nikolic, then everything falling apart as he ran short of time.

Both he and Kasparov had clearly demonstrated the machine's weaknesses, but the computer had been more efficient at exploiting theirs.

Grandmasters used to be smug about machines' fumbled attempts to play chess, but with computers now fast enough to be able to analyse millions of positions every second, players are having to reconcile themselves to the inevitability of brute force of calculation overcoming human intellect and guile.

While some grandmasters in London were still maintaining that machines would never play 'creative' chess, and that their wins were 'bad' games, most accept that this is the beginning of the end for human dominance.

Trying to find a more optimistic viewpoint, some see a new beginning for chess as a spectator sport, with the performance of grandmasters on stage being judged by perfectly informed spectators, each with a laptop computer, ruthlessly assessing every move. Generally, though, the mood is more resigned.

'I chose chess rather than music or art,' explained Nikolic, 'because at chess you win or lose. Your performance does not have to be judged by someone else. Now, when we lose to computers, we must accept the more unpleasant side of that fact.'

Nikolic guesses that humans may have another couple of years' grace before the machines take over, but when they do the news could be grim. Last year, Kasparov earned more than pounds 1m by playing the best chess on earth.

What price will the world human champion fetch when a piece of software retailing at less than pounds 150 can beat him?