If you want to know how the creature worked Richard Evans wrote an article for AI Wisdom called "Varieties of learning" but if you want to try to write AI, think of the simplest game concept you can (tic-tac-toe is always a good starter) and think about how to program it. If you get lost, post a more specific problem here (such as "how would _you_ program tic-tac-toe") or look up tic-tac-toe AI on the net.

I know that when you approach such a large body of work it can be confusing and you won''t know where to start and you''ll have no "feel" for AI in games, but if you start small and build up you should get a sense of what methods work for what problems.

Making it up as you go along is a good way of truly learning something. That way you make lots of mistakes and you know what works and why, which is more than you get than from reading a book.

Mike

Don''t get hung up thinking you "have" to use existing AI systems and algorithms. Who knows, by trying something new you can make a major breakthrough. Also, kepe in mind that the kind of AI you make is totally dependent on the game or application you are making. Neural nets might work great for A, but will suck for B.

Anyway, good luck. AI is rough. I had no previous experience and it was hard, but when you first watch your AI start following you and shooting ahead of your location trying to intercept you, it is a proud feeling. But then it is heart breaking when he gets stuck on the side of the hill and doesn''t realize his engine doesn''t have the power to go up a slope that steep. =) Alot to take into consideration, just take it one step at a time.

quote:Original post by wenching
I had failed my game project as i do not have enough experience in AI and C#.

You need to practice coding. It''s not needed to design an AI, but you probably won''t get real far implementing things without a certain level of coding knowledge. Like learning to walk or ride a bike, it''s not something someone can give to you, it''s something you pick up from practice. Sorry about that

quote:Original post by wenching
I want to be a game developer and one day i hope can create very amazing AI like in Black and White and Halo game.

Well thank you thank you. Sorta. Long story but the short version is that B&W appears to use an AI design i created and published on the Web as open source 4 years ago. There''s been some debate about whether it was "borrowed" from me or if it was simultaneous invention, but it''s been truly bizarre to see how much press the B&W AI has gotten

In any event, you can waste time reading my site or you can find Evan''s article on Gamasutra.com on how he made it work. None of the articles are probably at the point of walking you through the code but you''ll get a high level of how it works

quote:
As there are 2 many areas in AI, i want to stay focus on couple of areas and will make as many researches i can.

Here''s the bad news. Both my LightBlue and B&W''s creature AI are hybrid systems based on cognitive modeling. Meaning it''s not as easy as saying "go learn neural networks" or "study production rule systems". It''s a lot of things. None of them hard, unless you''re not used to object oriented coding and modeling. Just a lot of common sense stuff


Captain Wenching has a good point. It would be nice if there were some code samples of AI for some simple but realistic games (ie, not a perfect information perfect action board game like tictactoe). i bet a site like that exists somewhere but i can''t personally name one

-b