Chess vs machine vs chess vs humans

[Michaël Samyn]

I've been playing chess against my iPod lately. It strikes me how different a game of chess against a computer is, compared to a game of chess against another person. In the case of the latter, a big part of the game, for me, is challenging and bluffing and also distracting en hoping that your opponent will not see certain mistakes or opportunities. If you know the person well, you can speculate on this, even. But the computer is a system.

Playing against the computer amounts to trying to understand this system and then, I guess, simply using the same system against the machine, and get better than the machine. Any kind of variety in the computer's play style seems random. Because the computer doesn't have a personality and therefore no extrinsic reasons or motivations for its behaviour.

I still find it interesting to play against a computer. Mostly to see how the machine-mind works. But it's a very different game.

I find it amusing to find that even game-games become something else on the computer.

[Kjell]

You're not really playing against the machine though, you're playing against a algorithm designed and / or written by the software engineer.

[Utforska]

I think faking intelligent and emotional behavior is one of the things computer game creators will keep being limited by for a pretty long time (I think real machine intelligence is possible, but it's far away and making better computer games is probably not the main reason to strive for it...).

Wouldn't it be interesting, by the way, to create a game that combined chess AI with Eliza-style conversations? It shouldn't be that hard to fake a conversation about such narrow a topic as chess. It could taunt you ("You better watch your rook!") and even "accidentally" reveal its strategy, depending on difficulty setting (are chess algorithms even strategy based, by the way, or are they based completely on statistics and brute force calculations?).

[Thomas]

This brings me to the topic of "AI Stupidity", which I think is what should have more research devoted to it. For a computer to act convincingly stupid is something that is needed to make it a engaging component and it is always what would make it closer to a playmate than some expert system.

Utforska: I like that taunting idea! It feels like someone ought to have tried that by now?

[Kjell]

The architecture of conventional processors isn't ideal for mimicking human-like intelligence. You end up with a thick simulation layer for dynamics that happen ( seemingly ) effortlessly in nature. Hopefully in the future we'll see hybrid chips consisting of traditional processors ( for number crunching ) and artificial synapses

[Thomas]

I do think we need to have any true AI to make the kind of experiences that Michael talks about. I think that when dealing with a very narrow subject it should be possible design an opponent to a "fun" player instead of just the best possible decision maker.

It is not really the same, kinda related though, but when I am programming AI for creatures in our game I am not trying to make them is as deadly as possible and that is a totally different task. Part of that is to add the appearance of personality or at least some kind of intent. This is hard stuff and I have way to little time I can spend on it, but often very small stuff can add to the behaviour. Many times just having some random stuff happening can give the appearance of some deeper intelligence, when in reality the AI is very shallow. For example, just small tweaks in the manner a creature tries to find the player after loosing track can have a huge difference.

[Michaël Samyn]

I didn't mean to imply that I wanted the machine to act like a human. Not at all. I find the difference interesting. Chess is a different game against a computer. But not an unpleasant one.

That being said, the topic of autonomous behaviour of computer characters is near to my heart as well. We've done a big project around this topic a few years ago. In general, I don't think we need actual artificial intelligence for making our characters in games seem alive. The main reason for this is the presence of the player. Our characters don't need to be alive. They only need to appear alive to the player. And if the player is willing to play the game, we're already halfway there without a single line of code. Imagination is key to this, as to many things regarding artistic creation. The imagination of the viewer is, to some extent, our true canvas.

[Michaël Samyn]

I have started experimenting with other chess apps. They all play in a different way. The one I'm playing now (tChess) feels a lot less rigid than the one I was playing before (Deep Green).

I lose a lot. But it's a different kind of losing as in other, regular videogames. In those games I lose because I fail. I fail to make a jump, fail to duck, fail to take health, fail to solve the puzzle on time, etc. In chess, however, I lose because my opponent was smarter than me. This makes losing a far more integral part of the game. In most videogames, "losing" is simply ending the game because you failed to do what you're supposed to. Game over. After a checkmate, I often feel that the game could continue. Maybe we'll go after the Queen, next? When you lose in chess, it's not like the world has ended, or you have "died". It's just your opponent being smart. You admire him for that. With chess, you don't get frustrated. You don't throw the controller at the screen. You never have to start over and over again until you get it right. When you lose, it doesn't mean you did it wrong. Just that your opponent was smarter.