Can a computer ever really think like a human? Ross Bentley reports on the latest battle between a chess legend and his would-be nemesis
He is probably the most famous chess champion in history, and certainly the most successful grandmaster of recent times. But lately Garry Kasparov has been cast in an even loftier role - standard-bearer for the human mind in the battle with computer-generated intellect. As Newsweek put it, he embodies "the brain's last stand".
Kasparov made his name in a series of classic battles with Anatoly Karpov in the 1980s, but a new generation of chess fans know him as the frowning Russian who pits his wits against the most sophisticated chess software around.
The latest of these contests took place at the New York Athletic Club last month when Kasparov drew a six-match encounter with a PC-based computer program called Deep Junior
Winner of the world computer chess championship three times and noted for its aggressive attacking style and tremendous tactical ability, Deep Junior is the creation of Israeli programmers Amir Ban and Shay Bushinsky.
Computer chess software has a rich heritage stretching back to IBM's Deep Blue computer, which Kasparov first played - and beat - in 1996, and IBM's Deeper Blue, which famously defeated the Russian in 1997.
But, according to Ban, this latest creation marks a quantum step in the evolution of chess software.
"Compared with Deep Blue's awesome powers of number crunching, almost 300 million calculations per second, Deep Junior is a babe," he says. "The strength, however, of Deep Junior is that it processes in more abstract human terms. Deep Junior's capability of three million moves per second may not sound like much, but it also analyses and assesses the best future moves at the same time," he says.
Supporters of Deep Junior say its humanoid qualities were best illustrated in game five of last month's tightly fought contest, when it took the speculative step of sacrificing a bishop in order to further its progress.
"What it did was audacious and unprecedented in the history of man versus machine - its seemingly human-like cunning took everyone's breath away," says Bushinsky.
But while Kasparov admits he had to muster all his powers of concentration after this unconventional manoeuvre, he denies that the contest signifies the ascendancy of the computer over the human mind.
"What this tournament has shown is why a computer, in my view, cannot be more like a human, at least not in the immediate future.
"Whatever Shay and Amir say about Junior's ability to run through millions of possible strategies, I, by contrast, might consider only a few strategies in any one game.
"But you can bet your life they will be the very best ones," says Kasparov.
With several titanic struggles against high-powered chess computers under his belt, Kasparov is keen to expand on his theme.
"The mind's essence is creativity and the ability to improvise - this is what demarcates humans brains from computer software.
"Chess computers have something I call a 'black limit'. Based on its pre-programmed algorithms, a chess computer will calculate six, seven or maybe eight moves ahead. The problem is that once the computer has decided upon a course of action, it is very hard for the machine to change that instantaneously. We humans - even when playing a strategic game like chess - will nearly always feel the danger even before it happens."
But if chess computers are always going to play second fiddle to the human mind, do they have any use in the wider world?
While academics have, for the most part, rubbished the idea of chess programs being important testing grounds for the future development of artificial intelligence, Bushinsky says he is in negotiation with a number of parties to develop the commercial potential of Deep Junior.
At IBM where the original Deep Blue program was developed on an RS/6000 Unix server, the high-powered, number-crunching technology honed for the Kasparov contests is now being used to compute large quantities of data in areas such as molecular modelling and oil exploration.
"Deep Blue may make big headlines but the real work of Deep Blue's RS/6000 technology takes place quietly," says Mark Bregnan, general manager of IBM's RS/6000 division.
"Day after day, the IBM RS/6000 helps pharmaceutical companies develop innovative drug therapies, auto makers design cars and petroleum companies explore for oil under the ocean. It also helps forecast the weather."
But whatever the context, computers have one habit that places them below humans in the food chain, as Kasparov is quick to point out. "People think that chess is a calm and gentlemanly game. But in fact it is a sport that can, on a psychological level, be very aggressive and dynamic," he says.
"I like to joke that I managed to psych out Deep Blue in our second tournament in 1997. Not a lot of people know that Deep Blue actually crashed three times - something IBM's PR department initially tried to deny."
Checkered past: the history of human chess champions versus chess computers
February 1996: Philadelphia: Kasparov plays a six-game match against Deep Blue, a computer programmed by IBM. He declares his 4-2 victory a "win for mankind", conceding that the match was one of the most difficult of his career
May 1997: New York: Kasparov plays a rematch against an improved IBM program named Deeper Blue. In one of the most famous chess matches in history, Deeper Blue defeats Kasparov by 3.5-2.5 BOXTEXT: October 1999, Kasparov takes on the whole world, via the internet, and wins in 62 moves over four months
October 2002: Bahrain: Sheikh Hamad bin Isa al-Khalifa offers a $1m (£633,000) prize for the winner of a match between the human world champion, Vladimir Kramnik from Moscow, and the world's top chess playing program, Deep Fritz from Hamburg. The eight-game match ends in a 4-4 draw
January/February 2003: New York: Kasparov picks up $700,000 for his part in the 3-3 tie with Israeli program Deep Junior.
Additional reporting by Jonathan Webster
This was first published in March 2003