This is a very short article (by me) that links two very long articles (by people much better than me).
In short, today’s AI is statistics-oriented, and the huge practical success it has had so far has pushed the adoption of these methods in other fields as well. Where is the difference between the “old school” AI, more oriented towards rules and this one?
This is the subject of this interview to Noam Chomsky, which you can find here: http://m.theatlantic.com/technology/archive/2012/11/noam-chomsky-on-where-artificial-intelligence-went-wrong/261637/
Peter Norvig has also replied to it with an article just as thought-provoking and worth reading: http://norvig.com/chomsky.html
(None of this is new stuff, though I find this particular interview to Chomsky well made and going to a depth that the others I have read before were lacking. Norvig’s reply is linked for the sake of completeness to whoever does not know it and is interested in this argument, but it’s rather old - in fact, it precedes that particular interview by a year or two).