Topic: The evolution of AI

[foolish03]

http://forums.gamespot.com/gamespot/show_messages.php?board=909091266&topic=19669922&message=209857872#M209857872

Its an interesting article about the future of AI.  Although i dont believe next gen consoles will be able to achieve any significant improvements in ai over this gen pcs will definitely see a massive change.  What do you guys think.  

[KDR_11k]

Depends on what kind of AI you're talking about . Proper AI will only improve with larger neuron nets. Fake AI that's used in games might see a slight boost as previously impossible techniques become attainable but the major steps will still be made by intelligent design rather than raw power. (Fake) AI is not a problem of scale, it's as problem of application, namely to find new algorythms that help the AI. Unlike, say, graphics which can easily be advanced by increasing polygon counts and shader capabilities.

[couchmonkey]

I agree KDR...finding patterns of interesting behaviour for game AI to follow is a big part of what makes for good or bad AI.  I read an interview with Will Wright on Gamespy last week where he discussed creating "emergent" games.  The idea is that when he built the Sims, he worked out all kinds of patterns of behaviour for the characters and then ran simulations to see if the behaviour was interesting.  Creating interesting game AI is a matter of brainstorming and testing.  Sometimes the AI might not even be intelligent at all.  I'm sure lots of people remember the opening scenes of Goldeneye when you sneak into the base and there's a guy on the toilet.  There's nothing even that "intelligent" about that behaviour, it's just a scripted event, but it might seem smart the first time you see it because it's a realistic type of behaviour that you wouldn't normally expect in a game.

Now I should read the article.