In Germany, at the Robot Learning Lab, which is part of the Max-Planck Institute for Biological Cybernetics, Dr. Jan Peters and his team are trying to design robot that learn complex task by showing them what they need to do.
The robot then try to repeat the task and improve by itself on a trial-by-trial basis. Even though Dr. Peters's work is still preliminary, their robot was able to learn how to play the ball-in-a-cup game.
The following video show their robot in active.
Holy shit we're screwed!
ReplyDeletespiffy
ReplyDeleteThought it usually are the Japanese who come out with this kind of stuff.
ReplyDeleteThats looks cool, but its like a bit scary, isn't it? Maybe I didn't have to watch Animatrix...
ReplyDelete