This has been the norm since computers could play board games. I read Blondie24[0] about a year back about 2 guys who build a complicated neural network for a checkers program which teaches itself to play by playing against itself and is eventually entered into 'Human' tournaments. It's a really fast read even if you have no experience in AI.

There is also Deep Blue[1], which is famous for once defeating Garry Kasparov in a high profile match.

[0] http://www.amazon.com/Blondie24-Playing-Kaufmann-Artificial-... [1] http://en.wikipedia.org/wiki/Deep_Blue_%28chess_computer%29