Computer scientist Dr. Tom Murphy has created an artificial intelligence called Playfun. It monitors you while you play games so that the AI can beat the game better than you can.
One time Playfun was running through a game of Tetris when things went sour. That is when the AI took it upon itself to pause the game so that it wouldn't lose. Who says teaching robots how to learn is a bad thing?
I wouldn't go so far as to say that this is reliable or useful tool for that matter but if PlayFun was tasked with something more important like flying a plane, and quit that process because of really bad turbulence in mid-flight, well that wouldn't be good. Then again, that's probably why it doesn't fly planes.
Murphy's goal is to develop a program that learns from the user. The program does this by logging jostick/keyboard input and applying that information to it's own routines. In his paper, "The First Level of Super Mario Bros. is Easy with Lexicographic Orderings and Time Travel . . . after that it gets a little tricky" he writes "The basic idea is to deduce an objective function from a short recording of a player's inputs to the game. The objective function is then used to guide search over possible inputs, using an emulator this allows the player's notion of progress to be generalized in order to produce novel gameplay."
For right now Playfun is only playing NES games and it doesn't do a very good job at most of them. It does a better job at at playing platformers like "Super Mario Bros." but that's about the extent of its capabilities. Call us when it's winning "DOTA 2" games by itself.
Watch Murphy run through his thesis below:
