h2g2 The Hitchhiker's Guide to the Galaxy: Earth Edition

How about Go - it's much, much, much more complicated than chess. There are no computer programs yet in existence that can match up to even a moderately capable player.

Nice oblate spheroid stones, too!

I have played Go online a few times. Generally it goes one of two ways. "Hi, I'm a game theoritican" and I beat the stuffing out of them, or "So you from America?... You much play this game of Go?" And I get hammered into the ground. (Fellow English speakers beat me quite often too, but I have never beaten anybody from a country where Go is really _played_.) A good game of chess is like lifting a car using telekinesis. A good game of Go is like assembling a car using telekinesis.

Hmmm... good point.

Another game which computers fail quite spectacularly to play is Bridge. This is because the sheer amount of number-crunching involved in making a number-crunching opponent would make every move take far too many universe lifetimes. So to write a bridge AI you either have to make it stick to simple rules, or put a lot of work into making something quite intelligent...

Chess is pretty much the ideal game for computers to play... no random elements and simple rules.

Computers are logic, hence, they have a wonderful understading of these sorts of games. The odd thing with chess though is that _no_ computer is "world champion": a human always wins. That isn't the case for Othello / Reversi though which computers have been winning for years. There will always be one that we humans will win though: snap. That is, unless they give computers arms, which would be a very bad idea, unless they also give them wheels and hoovers.

Good point. Actually, I was thinking of writing a snap computer game... cuts out the cheating, kinda thing...

But I don't think I'll let the computer play.

Current computer programs cannot really be said to 'understand' chess the way that humans do, and they certainly don't play the way we do. They solve the problem by just evaluating every possible sequence of moves and choosing the best - pure brute force approach. Grand Masters don't (and can't) do that. They use many psychological insights into their oponents' play patterns and gamble against them, which computers are incapable (as yet) of doing. They work out what the opponent is likely to be thinking and/or planning, and whether they might be thinking about whether their plans have been discovered or not, and whether to change them. Computer can't do this kind of meta-thinking very well yet. GMs don't actually look that far ahead either, compared to computers.

Until someone produces a neural net-based chess playing machine that assesses its opponent by apparent character and mood, and can 'get inside his/her mind', a GM will always win - it's cunning, which computers aren't.

I liked Poe's Chess/Go/car analogies!