General Game Playing (GGP) where agents learn to play strategy games from a formal rule description given at runtime. This distinguishes them from game-specific AIs, like Deep Blue for chess, that have their strategies coded in advance