Sports

The Man Who Beat Kasparov

The greatest chess mind meets its match in Dr Hsu's ultimate chess machine-Deep Blue

Advertisement

The Man Who Beat Kasparov
info_icon

IN Arthur C. Clarke's short story Quarantine, a higher extra-terrestrial intelligence blows up planet earth, because every computer probe they sent in became hopelessly infected, and, soon, dysfunctional. "'They encountered a—problem—that couldn't be fully analysed within the lifetime of the Universe. Though it involved only six operators, they became totally obsessed by it.' 'How is that possible?' 'We do not know: we must never know. But if those six operators are ever re-discovered, all rational computing will end.'" The last line of the story reveals the identities of the six operators: king, queen, bishop, knight, rook, pawn.

"There are more possible chess moves than there are atoms in the universe," says Dr Fenghsiung Hsu. Neither his placid Taiwanese features nor his bland designation—research staff member, grand challenge systems research division, IBM—give any indication that he's the man who built Deep Blue, the first computer to beat a world chess champion, that too, Garry Kasparov, arguably the greatest chess mind ever. It's only when he starts describing Deep Blue's final showdown with Kasparov in May this year—an event Newsweek described as "the brain's last stand"—that you notice a disconcerting gleam you've seen before in the eyes of 'believers'. And when he talks about 'Garry', you get the feeling it's more than a scientific problem to be cracked; somewhere along the way, it became personal: one mind against another in a savage fight to the end. Kasparov had won the first match last year, but he lost the May rematch against an improved Deep Blue 3.5 to 2.5.

Advertisement

Hsu came to chess late, since in his native Taiwan, Chinese chess and Go were far more popular. While pursuing his computer science PhD at Carnegie Mellon University, US, he became interested in chess programmes. A strong correlation had already been established between a computer's processing speed and its playing strength. "I realised that if we could speed up processing 1,000 times, we could beat the world champion. But it would not be an easy problem: it would take five to 10 years." It ended up taking 12.

Hsu and friends built a machine called Chiptest, primarily with components surplus for other projects. In 1987, Chiptest won a nationwide computer chess competition. A year later came Deep Thought, named after the computer in Douglas Adams' Hitchhiker books, which could answer the "ultimate question of life, the universe and everything" (The answer is 42). It was the first machine to play chess at GM level. IBM hired Hsu in '89, and he got cracking on building "the ultimate chess machine": Deep Blue.

Advertisement

 In '96, the Association of Computing Machinery thought it would be a good idea to kick off its 50th anniversary celebrations with a Deep Blue-Kasparov match. Kasparov won 4-2, but Hsu's team learnt a lot. "It was quite clear to us that Garry knew more about chess than Deep Blue did. Also, Garry was changing: altering strategy mid-course in games, and from game to game."

The rematch came in May this year in New York. Armed with 286 processors and the power to evaluate 200 million chess positions per second (as against Kasparov's three), the 1. 4-tonne, 6-foot 5-inch machine was now faster, smarter, meaner. GM Joe Benjamin had helped refine Deep Blue's chess knowledge and correct its inability to handle midgame strategy changes. Its processing power had been doubled, and its evaluation function, which allowed it to see several moves ahead, was now even stronger. And you see that gleam in Hsu's eyes: the hunter recalling his biggest kill.

 "Game 1, Kasparov played white. He won, but it was very different from '96. Garry wasn't controlling the game any more, it was negotiation. We felt good, and we knew that if Kasparov was going to suffer like this in every game, then Deep Blue could win the match." But after the game, Kasparov said that he never thought he "had a chance to lose".

 "Game 2, Deep Blue with white, and Garry was squeezed throughout the game. By the end, he couldn't stand it any more. His face was red—he wasn't happy!" Hsu laughs. Kasparov lost, and later called Game 2 "the decisive game of the match which left a scar in my memory and prevented me from achieving my usual total concentration in the following games". He wrote in Time: "In...Game 2, we saw something that went well beyond our wildest expectations of how well a computer would be able to foresee the long-term positional consequences of its decisions. The machine refused to move to a position that had a decisive short-term advantage—showing a very human sense of danger." He even went to the over-the-top extent of saying that Hsu's team deserved a Nobel Prize.

Advertisement

Game 3 and 4 were drawn. "Game 5 was very crucial because this was the last time Garry would be playing white, so we knew that he would be trying very hard. And then in move H5, Deep Blue did something totally unexpected. Garry did a double take. It was a somewhat psychological move, a move that said: 'If you do this, I'm gonna go after you.' Humans don't do that to Garry Kasparov, but Deep Blue didn't know who Garry was." Game 5 too was a draw.

Deep Blue won Game 6 and the match. And Hsu has to prove to you that he won fair and square. "Commentators said Garry had made a mistake on one particular move, but Garry himself later said he regretted the decision to make that move. So it was a conscious decision! Deep Blue simply played better!" Vishwanathan Anand, though, had a calmer view. "By trying so hard to avoid any position where Deep Blue might be able to calculate its way through, (Kasparov) effectively self-destructed," he wrote. "By the 6th game, he was a pale shadow of himself. Suffice it to say that the trap he fell into in the sixth game is a well-known one."

Advertisement

 Unfortunately, the rematch ended on a bitter note. Wrote Kasparov in Time: "Whether they intended to or not, (IBM) created a hostile atmosphere that was very difficult for me to bear. There was something negative in the air. It was a Deep Blue show, and Deep Blue had to win." Bad vibes or not, it's undeniable that for IBM, it was the publicity event of the decade, especially for its (and Deep Blue's) RS/6000 hardware. IBM had created massive global media hype, painting this as the final shoot-out between man and machine, comparable to the '72 Fischer-Spassky match, which the media turned into a surrogate capitalism-communism war.

Advertisement

IBM also refused to give Kasparov printouts of Deep Blue's thought processes. The champion pointed out that in case of other opponents, he'd have access to all tournament games they had played, and form an idea of their style and thinking. Anyway, the Deep Blue he was playing in '97 was different from the one he played in '96. But, counters Hsu: "If Vishwanathan Anand goes away for six months to prepare and improve his skills, can Kasparov object?"

 Kasparov challenged IBM to another rematch, but with certain conditions. Among them: he be given in advance the log of 10 games Deep Blue plays with a neutral player or another computer in the presence of his representative; and that IBM should not organise the match.

So is Deep Blue the end of the brain's reign or just another bleep on the technology timeline? Has the transistor finally vanquished the neuron, and are computers soon going to turn sentient and make slaves of our children? Hsu has—rightfully—earned his place in computer history as the first man to climb Mount Chaturanga, but the truth is that Deep Blue is as far away from 'intelligence' or 'consciousness' as your doorbell. It has brute processing power, but all it can do is play chess. It can't even learn to play better chess over time as any child can do. As George Plimpton, editor of Paris Review, put it: "I think to call it, even if Kasparov himself said it, the end of mankind, is pushing it."

Advertisement

If you ask Hsu these questions, you meet a blank wall. He isn't interested in discussing computers and consciousness; he hasn't heard of Roger Penrose, the British mathematician who has sparked off a global debate with his theory that consciousness works through certain quantum leaps in brain circuits, something computers can never attain. "That's philosophy," Hsu shrugs. "I deal with reality." Reality.

Two days after we met Hsu, IBM announced that it was "retiring" Deep Blue. There will be no rematch with Kasparov; the world's finest inorganic chess mind will never play again. It has served its publicity purpose.

Advertisement

Tags

    Advertisement

    Advertisement

    Advertisement

    Advertisement

    Advertisement

    Advertisement