I was recently invited to give a talk on “AI in games” to Jyväskylä’s AMK game programming students. The lecture could have gone better if I had prepared more material for it, but the catch was that I didn’t know what to prepare. These notes act as a guide for “next time”. I won’t go into any depth on the subjects here.
- Briefly introduce scientific AI, point out useful libraries [A*, speech recognition, opencv] then waft it away
- Focus on architecture of players, representation of game state and move logic
- Establish the simplest AI player for tic-tac-toe, then develop it to a perfect player [“smart as possible AI”]
- Make the tic tac toe player more fun to play against
- Consider expanding it to more complex games (chess, civilization) and optimizations
- Human behavior simulation – know the limits of senses, program them in. FPS/Stealth AI is a good example. [“dumb as a human AI”]
- Planning and decision making architecture
- Techniques for more realistic behavior
Additionally…
- Examples of AI in video games: planner in fear, neural networks aren’t used much in industry (notable exception, Supreme Commander 2) and only for decisions, Quake 3 doesn’t have neural net bots
- Data structures and algorithms are ever useful
- Observe nature and mimic behaviour
Bomadeno.com 