To the Editor:

I read with interest Boris Gulko’s article, “Is Chess Finished?” [July]. As an amateur player and professional chess journalist, I have the greatest respect for Grandmaster Gulko’s judgment in these matters. However, his smooth and plausible prose may be boiled down to two points: (1) that chess is an art beyond the ken of mere machines; and (2) that world champion Garry Kasparov lost to the computer Deep Blue because he was not in good form.

The first point is obviously false: the chess world has observed the steady up-ward progress in the performance of chess-playing machines for 40 years, with no observable flattening of their trajectory to date. (This is literally true, as graphs demonstrate.) Forty years ago, chess players were saying: “Yes, it’s possible to teach a machine the rules, but it will never be able to compete with skilled human players.” Soon the skeptics had to back off a bit: “All right, machines can beat average tournament players, but they’ll never beat masters.” Retreating to higher ground as the water rose, they then said: “Ah, but machines will never beat grandmasters—grandmasters are special; their play is qualitatively different from that of lesser mortals.”

Then the first computer beat a grandmaster. The skeptics, now treacling water, dismissed this as a fluke, a bad day for the human. But soon it was happening often enough that it became clear some machines were, indeed, playing at grandmaster level, and the skeptics were reduced to their last stand: the mystical belief that no machine would ever beat the world champion.

Now comes Deep Blue and beats the world chess champion in a set match, after playing toughly against him in another match a year earlier. How could this happen? Grandmaster Gulko says (and this is his second point) that Kasparov was not in good form. Notice that Mr. Gulko has refined the “bad-day” argument, by claiming that Kasparov had a “bad week.” How many bad days in a row must a human have before the skeptics will admit that, just maybe, there is something to this computer-chess thing?

But even if it is true that Kasparov was not in form, Mr. Gulko’s point is moot. It is obvious that machines are playing better every year, while the best human players are not perceptibly improving. In other words, if Kasparov had not lost this time, he would have lost next time, or perhaps the time after that. It is pure romantic bravado to pretend otherwise.

Romantics persist, of course. In addition to Mr. Gulko, three-time U.S. champion Lev Alburt (like Gulko, a Soviet grandmaster who emigrated here) published an article in the July 1997 issue of Chess Life, the national magazine of the U.S. Chess Federation, brashly titled “Computers Will Never Surpass the Best Human Chess Players.” Perhaps the article went to press before Kasparov’s debacle in New York. As Will Rogers once wrote, “In America, people who say it can’t be done are usually interrupted by someone who is doing it”—or in this case, by some computer that has just done it.

Some people will never admit the obvious if it clashes with their self-esteem or deeply held beliefs. It is not surprising that a man like Boris Gulko, who has devoted his life to what he calls the “art” of chess, should be reluctant to admit the truth. But unless IBM loses interest in this diversion, it seems obvious to objective observers that machines, even with their occasional dumb moves, will continue to improve beyond the level of their human teachers.

Boris Gulko asks, “Is Chess Finished?” His answer is no, but his entire line of argument suggests that chess would be finished if machines were indeed to surpass humans. My answer to his question is also no, but for a different reason. Long after the computer-vs.-human competition has been settled definitively in favor of the machine, chess will survive and prosper in the same form it always has: as a competition among humans. Human players will still battle among themselves for intellectual bragging rights, and chess fans will eagerly follow the fight vicariously, whether in real time over the Internet or later in the pages of a magazine or well-annotated book. May the best man—or woman—win.

Timothy Hanke
Editor, Chess Horizons
Newburyport, Massachusetts

_____________

To the Editor:

Boris Gulko is understandably reluctant to concede human-like abilities to chess computers, in particular the ability to plan. This has long been an assertion of strong players, and I believe it basically represents a state of denial. It is connected with the human propensity to reify concepts: we conceive of a thing called “planning,” and we believe that it requires an intelligent mind; by definition, a computer does not have intelligence; ergo, the computer cannot plan. Case dismissed.

Operationally, however, this is by no means so clear-cut. For example, imagine you are playing a fairly simple king-and-pawn endgame against a computer—and give the computer a passed pawn (i.e., a pawn whose path to its queening square is not blocked by your own opposing pawns). Invariably the computer will attempt to move that passed pawn toward its queening square. The closer the pawn gets, the more the computer strives to push it forward. It is evident that the motivation for the computer’s “planning to queen” is provided by the large advantage it foresees if the pawn actually does queen.

Now go back to the initial position, turn the board around by 180 degrees, and you are the one who has the passed pawn. You find that you too will push the passed pawn and “plan to queen.” In fact, your chosen sequence of moves may be exactly the same as that of the computer. As a chess player would say: the “logic” of the position demands such a plan.

We can expand and generalize this little example to formulate a rough, working definition of planning in chess: the player (1) examines many different possible sequences of moves; (2) evaluates each step of each sequence; and (3) selects the sequence that gives the best final evaluation.

Of course the precise details of how these sequences are selected and evaluated are quite different for a silicon chip on the one hand and a human brain on the other. But it is clear that the overall process is the same for both types of player. And as of May 1997, it is also clear that the best computer is on a par with the best human in this process.

Let me gently suggest to Grandmaster Gulko, to his fellow-grandmasters, and to other chess players that perhaps now is the time to stop berating computers, but instead to use them more actively in analyzing the areas which continue to fascinate devotees of the Royal Game: things like finding opening novelties, clarifying complex mid-game combinations, and calculating deceptively simple-looking endings.

The game of chess is not finished—but the war is over.

Joseph Faucher
Carlsbad, California

_____________

Boris Gulko writes:

I completely agree with Timothy Hanke that computers have made astonishing progress in playing chess over the past 40 years. I also agree that this progress will continue until it reaches a degree of strength beyond what I can foresee. But I still disagree with Mr. Hanke’s evaluation of the match between Deep Blue and Garry Kasparov.

My opinion, in common with that of most grandmasters, is that Deep Blue played at a level indistinguishable from other top-of-the-line chess computers. Kasparov simply played uncharacteristically badly and lost. Deep Blue’s performance hardly represented a breakthrough of some kind.

The Deep Blue-Kasparov match would have far greater significance if its outcome were to be repeated in a rematch, or in a match with other top chess players who have subsequently challenged IBM: world champion Anatoly Karpov (of the Fédération Internationale des Échecs), for one; women’s world champion Susan Polgar, for another; or, still more interestingly, Kasparov, Karpov, and Polgar at once (a computer can play simultaneously against any number of chess titans without becoming flustered or fatigued).

IBM has rejected this approach, and, despite the tremendous publicity the corporation received from the spring 1997 match, it has no plans for further engagements with its machine. Having perhaps the clearest view of exactly how Deep Blue managed to subdue Garry Kasparov, IBM understandably prefers to rest on its laurels.

In his thoughtful letter, Joseph Faucher accuses me not of racism or sexism but of machinism: “now is the time to stop berating computers,” he writes, and to stop assuming that they cannot perform tasks, like planning, that we humans “reify.” This is an interesting line of argument, but I am afraid the very example he cites undercuts his argument.

Yes, in simple king vs. king-and-pawn endings, computers have been programmed to advance their pawns to the other side of the board so as to turn them into queens. And yes, this may make computers appear to be following a “plan.” But in fact they are only carrying out calculations, looking far enough ahead until they arrive at a decisive result. In more complicated positions, with more pieces on the board, mere calculation, for the reasons I set forth in my article, does not suffice. It yields only the kind of aimless play that we saw on occasion even in the Kasparov-Deep Blue match.

In short, I will “never admit the obvious,” as Mr. Hanke puts it, particularly when the obvious happens to be wrong. And as for whether I am in what Mr. Faucher calls a state of denial, I deny it, unequivocally.

_____________