Quote Originally Posted by Pancho Villa View Post
We have adaptive programs that can change how they operate based on a strict set of rules, but I don't think they "learn" in the same way we do. You'd have to write a different set of learning codes to cover every possible situation for the robot, and if it ever encountered something not coded for it would just take a shit and not know what to do. Overly complicated and VERY fragile.

Those adaptive programs are neat; a friend who was a day trader used one to automatically predict trends as stock prices went up/down. Had some pretty phenomenal returns for a while, and then equally bad losses.
Like I said, there are robots already that carry out experiments, reason about the results, and plan the next experiment without the help of humans.