The field of machine learning and intelligence is a serious and important area of scientific research right now. Machines that can learn from their mistakes or learn entirely new tasks promises a future of improved industrial automation and electronics that are literally smarter. Still, it's hard not to chuckle just a little bit when the researchers at Rutgers choose to use video of a learning algorithm figuring out how to play Pitfall! to show their work in progress. Of course, only a fool laughs in the face of his future lord and master, and I'll no doubt either be dead or working in the deepest darkest parts of the magnesium mines once they figure out the launch codes. Check out Post Position for more of the nuts and bolts behind the Object-Oriented Markov Decision Process that drives this AI, then check out this video to see a machine figure out how to play Pitfall! You'll have to decide for yourself whether to smile or be profoundly disturbed when it does its victory dance.
Giant Bomb News
50 CommentsRobot Plays Pitfall!; Next Up, Thermonuclear Warfare
by Ryan Davis on
The only winning move is to, uh, avoid the scorpion?
50 Comments