Computer Program Plays 'Super Mario World' Better Than You Do

The winner of the ongoing Mario A.I. Competition must create, using a learning algorithm, a controller (or agent) that can complete the most difficult levels of the game 'Infinite Mario' without any human aid. According to Joystiq, a video of one recent entry looks like it may be tough to top.

Creator Robin Baumgarten, a PhD student at Imperial College, London, writes on his site that he entered the contest because he was "bored," and wanted to hone his A.I. programming skills. Typically, watching others play video games can get boring and incredibly frustrating, but that isn't the case here. The video, and the mesmerizing gameplay, is simply amazing.

Nintendo is also reportedly developing 'Demo' capabilities for games, which would complete difficult levels for lazy and impatient gamers. With autopilot game levels, A.I., evolutionary neural networks, genetic programming, temporal difference learning, and predictive texting, the next generation of humans may not have to actually do anything for themselves. Sweet. Maybe it's time to go back to school and pursue a degree in programming and A.I. What's that old line about monkeys, typewriters, and Shakespeare? Who needs monkeys if you can program an awesome blogging agent? [From: Robins Pages! via Joystiq ]

Tags: AI, controller agent, ControllerAgent, fun, mario, mario ai, Mario AI competition, MarioAi, MarioAiCompetition