AI is engaged in two significantly different enterprises: a `science' of human intelligence; and an engineering discipline concerned with building smarter physical systems.
In either case, the problems which AI addresses are ones connected with representing and reasoning about the nature of the world. The interests addressed range from how to recognise a smile to how to replan a space mission after a mechanical malfunction. The issues range from the philosophical (can machines feel emotions?) to the technical (how do we guarantee the correctness of a database update scheme?).
The problems addressed are typically hard ones (for a variety of reasons). Solutions are usually partial. Even limited success can lead to a strange effect: a successful technique developed may be adopted as a standard computer science technique and therefore no longer AI!
AI draws on work done by people from many different disciplines. These include: philosophy; psychology; computing; linguistics; anthropology; and logic. The problems addressed by AI can be found in all disciplines from software engineering to film theory. Methodologies range from formal, deductive analyses through to empirically-based studies. The techniques used in AI research are very varied.