Garry Kasparov is probably as famous for the thousands of matches he won during his 15 year reign as world chess champion as the one match he lost, in 1997, to IBM’s chess-playing computer Deep Blue.
He’s still not completely over it.
“So looking at the games and the moves you understand Deep Blue was far from perfect,” Kasparov tells the audience at Gartner Symposium “And I was probably stronger objectively when you look at the chess rankings.”
He was “under pressure” the Russian-Armenian grandmaster says, recalling the front page headline on Newsweek days before of the match: 'The Brain's Last Stand'. “The future of humanity is on the line” one news anchor said at the time.
“I was not well prepared for the match,” he tells the Gold Coast conference crowd. “I have to give all credit to the IBM scientific team, they did a great job. Though of course the conditions weren’t perfect and I’m not complaining about it…”
He reminds the audience that the 1997 loss was actually the rematch “just for the sake of the truth – stick to the facts”; Kasparov beat Deep Blue a year earlier four games to two (two were drawn).
“I’m a sore loser. And that’s why I stayed on top for so long because you have to be a sore loser, to feel this pain when you lose,” Kasparov says.
Weak moves
Kasparov has the misfortune of being reigning champion at chess through the period when computers went from being “laughably weak” to unbeatable at the game.
Only a few years earlier, in 1985, the grandmaster took on 32 machine opponents from four manufacturers at once, and won all 32 games “no draws, no losses”.
A photo of the exhibition is projected behind him. “This reminds me of the golden age of human machine relations. Machines were very weak, but my hair was strong!” he jokes.
By ’97 however, the machines had the man considered to be the game’s greatest player beat.
The match has been scrutinised move by move ever since. Like the 44th move of game one, allegedly the result of a bug in Deep Blue which left it unable to determine a desirable move, so it defaulted to a safety play. It threw Kasparov, who attributed it to “superior intelligence”. Or the 46th move of game two, which led Kasparov to accuse IBM of cheating by taking tips from a grandmaster rival working behind the scenes.
Whole books have been written on the contest. A movie has been made. Kasparov says it was “the watershed moment; that was the turning point”.
The defeat prompted some soul-searching from the world champion.
“I was sitting there, pensive, ruminating what happens next,” he recalls.
Better forever after
Chess had long been thought of as the pinnacle of human intelligence, but it seems strange now to think of a human beating chess engine in ‘end of the brain’ terms.
Chess game apps can easily checkmate most human players, and there are unbeatable programs that run on standard multi-core CPUs.
“Deep Blue was not intelligent. It was as intelligent as your alarm clock. A very expensive one. It was not a chess playing machine per se – it was more of a parallel processor project, but it could also play chess,” Kasparov says.
Chess is no more the big test of AI. Since it can be successfully won by computers with a brute force approach (or Good Old-Fashioned Artificial Intelligence), AI supremacy is now focused on more open, strategic games like Go.
AI has already beaten the best human players at Go however. In 2016, Google DeepMind’s engine AlphaGo took on and defeated legendary Go player, Lee Sedol, in what the company founder called “a major milestone for artificial intelligence”. Rather than brute force it used deep neural networks and advanced "tree search" programs to succeed.
“The cycle is repeated in every closed system,” Kasparov says. “First they say it is impossible. Then machines are laughably weak. Then it’s the very short window when we can compete, and then machines are better forever after.”
One by one, AI will become better than humans at many tasks. And that’s okay says Kasparov, as long as humans use the capability wisely.
“Machines will become more intelligent. It means we will have to make sure that we know how to use this power because with these new powers, the impact of each mistake is amplified,” he says.
“It is for us to understand how we coordinate these machines, how we cooperate with them, how we use them for our advantage,” Kasparov adds.
The author travelled to Gartner Symposium as a guest of Gartner.