By taking a look at current virtual agents and by extrapolating trends in game design, this feature reveals the numerous benefits of expanding current game AI. Learning is identified as one of the key components, and possible implementations are discussed along with their drawbacks.

http://ai-depot.com/GameAI/Agent-Intelligence.html

There's been a lot of activity on this topic recently, so I felt compelled to contribute ;)

If anyone would like to experiment with using the Ai-Foundation's AIML chat bot code base to give a game agent a brain (things like "Predicates" can be set to handle the state variables and learning mentioned in the article) I would be happy to discuss the use of my company's AIML Editor for the project.

You may have access to the version of the Editor which produces chat bot AIML files at no charge. All we ask is to receive credit for contributing. [url]www.ai-dealer.com/AIML_Editor.html[/url]

Once you are comfortable with it, we also have a version of the Editor which can handle expert-based decision making and state tracking of a user's interaction with a system. It is not free, but we can work something out for the right application. We are interested in demonstrating what is possible.

We developed it for a commercial software application (complex eCommerce transaction - having a dealer sell a vehicle online), but would be interested to see someone apply it successfully in gaming.

Watching my sons play their video games, I know there is a big future for gaming agents that are not "dumb as a post."