Thoughts about an unequal battle

A question that is not heard enough these days, is: should we actually be glad with all these matches between Man and Machine? Of course there are many perspectives from which you can try to answer this question.

The sensationalist in me looks in particular at the entertainment-level of the match. In the Kramnik-Deep Fritz match, that level is relatively high so far: the World-champion overlooks a mate-in-one - what more can you wish for? As a chess lover I am indifferent for the moment: no surprising opening novelties, three little endgames, interesting but not of lasting value, and one blunder which every true chessplayer will understand with a slight feeling of shame.

From a philosophical point of view, one can criticize the match as well. What exactly are we ‘measuring’? Is it right to let the computer use external help sources such as an opening book and a table base? Is it fair to limit Kramnik’s thinking time? Wouldn’t it be better to facilitate a kind of consultation between humans during the game? After all, two humans know more than one, just like two processors can calculate better than one.
Maybe we shouldn’t worry so much. You take a human and ‘a box’, as one of my chess friends put it, you let them face each other, and you watch who comes out victorious in this ‘cage fight’. That’s the strongest player. But it’s not that simple. If everything is allowed, I would, as a human, consider simply pulling the power cable from the ‘box’. Good luck with your opening library! (I am informed that the Dutch Grandmaster-writer Donner once answered the question ‘How do you beat a computer?’‘ with: ‘With a hammer!’) Why shouldn’t a human be able to use al his available possibilities to win the cage fight? Besides, when does the ‘box’ stop being a box? Suppose I make the box bigger, and put two little boxes inside of it. Or ten. Or ten thousand. Will the fight always be a fair one? Should the fight actually be fair?

The match is most unsatisfactory to me when I look at it from the perspective of my own profession: software programmer. As every programmer knows, debugging and testing of the software is the most important part of the job. To get a good view of the strenghts and weaknesses of your program, you should test as much scenarios as possible. To measure the strength of your chess engine well, you should check your engine at work in all phases of the game. That means also in the opening, and also in theoretical endgames.

But this never happens. It’s certainly a noble decision of the organisers to give Kramnik insight in the opening library of the computer, but it’s curious all the same. It looks as thought the Fritz-programmers are not even interested in what their program itself thinks of, say, the beginning position. Have you ever looked at what your computer program says about the initial position? The thing recommends the Petroff Defence, Guico Piano, or Queen’s Gambit. As a programmer, I would be proud!

If they’re really afraid that Fritz’s evaluation is not correct, or not good enough to compete with Kramnik, of course it’s handy to use 400 year of chess opening theory. Well, handy it certainly is, but it’s also kind of lazy. It’s a little as if I claim to have written a beautiful program, whereas I actually just stole a large part of the source code straight from the internet. In Dutch we call it showing off with another’s feathers. As a programmer I wouldn’t want to see the quality of my software judged by something like a pre-programmed, instant ready-made opening library. Can’t the thing just do it by itself?

Photo: Andreas Schwartmann. Look at more of Andreas’s photos, for example in a slideshow.

Perhaps it can, but a computer playing without an opening book raises other questions. After all, Kramnik is allowed to use his opening knowledge? Well okay, he simply cannot switch off his opening knowledge. (Maybe therefore we should now simply conclude that a fair battle between Man and Machine is never possible.) To solve this dilemma, they should perhaps play Fischerrandom chess. No opening theory. Only then, it will be possible to test how the engine measures up to a human chess player. But then the evaluation will ignore one of the biggest powers of the human brain: combining known patterns and information.

And besides, Fischerrandom? Isn’t that that little game where world class players start blundering exchanges and pawns at move three, as Kasparov recently showed in New in Chess Magazine? Apart from the fact that the big public rightly can’t ever be interested in such a Man vs. Machine Fischerrandom match, I have a sneaking suspicion what the final result would be, and it wouldn’t make humans very happy.

And suddenly I get this evil thought. It’s not the Machine that needs an opening book, but Man! What would happen if Fritz would play 1.a3 with White, and 1…a6 with Black? Kramnik would get an opening advantage, the position would become complicated, but in the tactical confrontation needed to profit of the opponent’s unorthodox play, Kramnik would probably be outplayed. The opening book is just about everything Kramnik can still hold on too!

And there we see the real problem. Secretly, deep down, we know very well that computers in many aspects play much stronger chess than humans. In fact in almost all aspects, especially in a practical game. In the end chess is pure calculations, and computers happen to be better in it than humans. Everybody knows that. In blitz, where tactical skills matter most, humans have already been without a chance for years - even against relatively ‘weak’ computers.

So what do Man (Kramnik) and Machine (Chessbase) want to achieve with this match? Of course, this classical battle has always captured the imagination, from The Turk to 2001: A Space Odyssey. Does Kramnik want to show that he doesn’t have to lose, provided he doesn’t blunder? But we also knew that already. With correct play from both sides, every chess game will end in a draw. Do the programmers really want to test the strength of their engine, or do they want publicity mostly?

It’s a fact that the so-called battle between Man and Machine is starting to look more and more like the cliche of the runner against a race car. We know that speed is, in the end, just a matter of physics, and that in this aspect we simply cannot compete with machines and some other animals. Thus we also know that chess is ultimately a matter of calculations, and that we cannot compete in this against a computer.

With all kinds of patchwork (artificial opening libraries with ‘human’ openings, months of specific preparation, teams of seconds) we can probably give ourself the illusion that we keep up quite nicely. But we will lose the battle. Big deal. Let’s use the computer for cracking and improving old analysis, for making beautiful new analysis, for calculating theoretical endings and - in due course - for rewriting opening theory.

Psychologically, we’re probably not quite ready yet for giving up. Let’s at least console ourselves with the thought that no computer can even begin to understand that very concept.

>> replay games 1-3 of Kramnik-Deep Fritz