from the wouldn't-it-be-nice-if-people-had- dept
Here’s an article about a developmental robot that learns “good” and “bad” actions by researchers pressing the “good” or “bad” button on the robot (or by saying “good” or “bad” out loud). Right now, the robot is learning to wander the halls of buildings. I realize there’s a lot of similar research going on out there, but this one sounds pretty cool. They compare it to the Aibo robot dog. Of cousre, the problem with all of this is that the robot is only learning a “bad” thing after they’ve done it. I could easily see this creating some very bad results before the robots are trained properly.