Deep Thought Over Computers

‘Is it possible that the next generation of computers will solve chess?

Dominic Lawson

Among the many amusing literary inventions of the science fiction writer Douglas Adams was a computer called Deep Thought. Fans of The Hitchhiker’s Guide to the Galaxy will recall that Deep Thought had been asked by its programmers to compute the Answer to the Ultimate Question of Life, the Universe and Everything. It would take seven-and-a-half million years, but at the end of that span of time Deep Thought produced the answer: 42. This far from satisfactory response would have pleased the National Union of Philosophers, Luminaries and Other Thinking Persons, whose representatives, as Adams related, had been concerned that Deep Thought would make all of them redundant.

A similar anxiety must now be seizing the more apprehensive of the fraternity of chess grandmasters. Is it possible that the next generation of computers will solve chess? This would mean, in practice, finding out what the best moves were for both colours, beginning with the starting position. It might turn out that the initial position is a win for white, or (more likely, I suspect) a draw. However, in any event, once such a series of moves was discovered, chess could be said to have been “solved”. Game over.

Computer scientists, for all the extraordinary advances in programming and miniaturisation, still regard this as unfeasible in practice. They point out that the number of possible chess positions greatly exceeds the number of atoms in the known universe. To hack through all of that might indeed be the work of millions of years. Still, some grandmasters have a deep hostility to our silicon friend—Nigel Short for one. He told me some time ago that he would never play in an event against a computer, on the grounds that “they are trying to do us out of our job, and I don’t see why I should help make them any better.” 

Unfortunately for Nigel, they are increasingly better at a dramatic rate, even without his assistance. In 1978, the year in which Adams’s radio series was first broadcast, the chess master David Levy won a bet he had made in 1968, when he staked £1,250 (a lot in those days) that Silicon Valley could not, within ten years, produce a chess program able to beat him in match. Bear in mind that Levy, while a former Scottish champion—and also a noted authority on artificial intelligence—was not of grandmaster strength. It was considered a very rash bet at the time, but in 1978 Levy beat Chess 4.7 quite convincingly. He confessed, however, to being surprised at the strength of his opposition and, having originally said that “the idea of an electronic world chess champion belongs only in the pages of a science fiction book”, declared, “Now, nothing would surprise me.”

Levy was right to hedge his bets, this time. Ten years later, a chess program, called Deep Thought in honour of Douglas Adams’s fictional computer, and built by the AI team at Carnegie Mellon University, in Pittsburgh, Pennsylvania, became the first machine to beat grandmasters in tournament conditions. In the 1988 Long Beach Open in California, Deep Thought crushed Igor Ivanov and Bent Larsen. They had tried to out-tactic the silicon monster, which was exactly how not to tackle it. So I thought it would be interesting if I, a mere county-strength amateur, were to challenge Deep Thought, but to do so having taken coaching from Levy. Carnegie Mellon graciously agreed to give me an hour of Deep Thought’s time—so we would play a game of 30 minutes on each side. Armed with Levy’s advice to go for a very stodgy positional game I managed, by about move 30, to establish an advantage. But it’s usually necessary to engage in tactics to turn such a positional advantage into something more concrete. Sure enough, I miscalculated, and even managed to lose. Nevertheless, I felt, after this experience, that Levy had been right in his original assertion that no computer would ever be stronger than the best that carbon-based life forms could offer.

How wrong I was: less than a decade later, in 1997, IBM’s Deep Blue beat Garry Kasparov in a contentious six-game match—later the subject of Vikram Jayanti’s superb documentary film, Game Over. To Kasparov’s fury, IBM retired Deep Blue as undefeated champion, rather than agree to a rematch. The final demolition of humanity occurred in 2006 when the awesomely prepared Vladimir Kramnik, then world champion, was demolished by the commercial program Deep Fritz: not only did the computer not lose a single game, it did not even look as though it might.

The result is that anyone who has about £80 to spare can now have the world’s strongest chess player in his own home, on permanent standby: this is what it costs to buy Deep Fritz 11 in disc form. With some misgivings, I recently acquired this piece of killer software—and the results are as depressing as I feared. It makes analytical mincemeat of many games that you might once have treasured.

Last month, for example, I won a game in a vital club match, at one point playing what I felt was a devilish move. My opponent evidently felt so too. He looked most disconcerted by the move and subsided rapidly. The next day I ran the game through the Deep Fritz program. In less time than it takes you to read this sentence, it found a fatal flaw in the very move of which I had been so proud. For a few minutes, I sat dazed, and consumed with the thought: why bother?

We still do bother, of course. We remain hopelessly infatuated with the dream that one day we might play the perfect game, in which none of our moves, or our opponent’s, could be improved upon. In the real world however, this will be achieved only by two machines, neither of which could ever have had a dream in the first place.

Here is the game from that historic 1988 tournament, in which Deep Thought, playing White, demolished Grandmaster Ivanov:

1.e4 d5 2.exd5 Nf6 3.d4 Nxd5 4.c4 Nf6 5.Nf3 Bg4 6.Be2 e6 7.Be3 Bb4+ 8.Nbd2 0-0 9.a3 Be7 10.h3 Bh5 11.0-0 c6 12.g4 Bg6 13.Nh4 Nbd7 14.Nxg6 hxg6 15.f4 c5 16.g5 Ne8 17.Ne4 Nd6 18.Nxd6 Bxd6 19.b4 cxb4 20.c5 Bc7 21.axb4 a6 22.Qc2 Qe7 23.Qe4 b6 24.Qb7 Rfc8 25.Bxa6 e5 26.fxe5 bxc5 27.Bc4 Rab8 28.Rxf7 Rxb7 29.Rf4+-and Ivanov not only resigned, but abandoned the entire event. His mortification was made complete by White’s final move, which, if played by a human, could have been described as sardonic.

Underrated: Abroad

The ravenous longing for the infinite possibilities of “otherwhere”

The king of cakes

"Yuletide revels were designed to see you through the dark days — and how dark they seem today"