jump to navigation

Computers cannot play chess – no, I’m serious. December 13, 2006

Posted by dorigo in Blogroll, chess, computers, games, internet, news, personal.

Not kidding.

After a few weeks of absence, I went back to pay a visit to Tim Krabbé’s refreshing Open chess diary, http://www.xs4all.nl/~timkr/chess2/diary.htm . Tim is a web-friend of mine, a strong chessplayer, and inexhaustible source of chess trivia of the highest level, besides being a talented writer of novels (many of you have probably seen the movie “The Vanishing”, featuring Sandra Bullock and Jeff Bridges: it is based on one of his novels, although he is unhappy of the movie).

If you visit his site, besides getting hooked, you will find item 330, which is a jaw-dropping demonstration that indeed, computers -even the strongest ones in circulation, that is, including Fritz 10 and Rybka 2.2- cannot play chess. In two ways: first, by not seeing a mate in one or a queen en prise. Second, by showing they do not know the application of one of the most important practical rules of the game.

If you love chess, you are strongly encouraged to check Tim’s diary. If only to claim we humans are still superior to machines… However, the problem is in the way we humans program the darn silicon beasts!



1. D - December 13, 2006

this seems nitpicky at best – remember Korchnoi once famously had to ask if castling was allowed in a certain position.

In any case, that Fritz plays differently from Kramnik is obvious. What is dubious is arguing that therefore only the latter plays chess.

2. dorigo - December 14, 2006

Yes, but korchnoi did not allow mate in one when he was in doubt… The nitpicking, however, has to do with the fact that these programs are imperfect because of man’s choice of optimization. It is not an intrinsic limit of the machine…


3. hwasungmars - December 14, 2006

Oh, I thought computers were better than humans in chess. Hmm… if they are not better than humans now, I can bet they will be better in the future!

4. Lucian - February 9, 2007

a computer can’t play chess better than his creator.

5. dorigo - February 10, 2007

Lucian, that is not true of computers applied to chess any more than it is true in general. In fact, it is generally false.

6. Kannan Ramanathan - February 28, 2007

Computers are good at brute force / number crunching things. All the current chess programs now make use of weighted searching / pruning for finding an optimal move. This is why we think chess programs do not “understand” chess. This strategy has more or less hit its ceiling, except for the scope for some more optimizations.

I guess, only when the computers become massively capable enough to do AI / self learning algorithms, we will have a true chess program capable of being called something that “understands” chess.

7. dorigo - March 1, 2007

I think it depends on how you define “understanding”, kannan. For sure, something closer to the way we schematize logical reasoning is easier for us to call “understanding”, but if we get to the bottom of it, it is all a flow of basic units of information, in our mind as much as in a silicon chip.

8. There goes checkers « A Quantum Diaries Survivor - July 21, 2007

[…] ? Well, fear not. Despite the achieved supremacy of computers on over-the-board play (with some caveats, true), and despite the complete solution of five- and six-men positions, computers will never […]

Sorry comments are closed for this entry

%d bloggers like this: