Just read an interesting article on Times Online about some research being undertaken by the US Navy regarding AI in field robots. Interesting stuff, although it's all been talked about for quite a while now.
Two key questions come up:
1. How do you stop the robots from going all Skynet on you?
2. Who's responsible when the robots go wrong?
You can see the article here: http://technology.timesonline.co.uk/tol/news/tech_and_web/article5741334.ece