I think the biggest issue when thinking about AI is the possibility of creating something that comes to be smarter than humans and take over. In order to reach strong AI it will have to pass the Turing Test, which tests a humans ability to tell whether they are talking to a human or machine. If a reasonable person can be fooled into thinking that they are talking to another human being than the machine has passed the Turing Test.
If we are to create something this smart what is to keep it from evolving. Once it has become self-aware and has the ability to evolve of its own accord what control do we have over it? This seems to be the main theme of many tragic post-apocalyptic world movies.