Rodney Brooks on How Shallow Deep Learning is

Rodney_Brooks

 

WBUR’s Radio Boston:

On whether he foresaw today’s robots as a student at Stanford:
“I thought we’d be further along in some ways. And, I didn’t imagine lots of other things. At that time, I knew of three mobile robots in the world that were professional mobile robots that people actually had. I then went on here in Boston to co-found iRobot. And iRobot has sold over 16 million mobile robots. I couldn’t imagine that number back then…But I probably thought they would be smarter than today.”

On how far along artificial intelligence actually is:
“I think there’s a bunch of scientists and engineers who are not in AI who have misinterpreted AI and I think people in general misinterpret AI…When a person does some performance of some sort, we’ve got a pretty good model of how that performance generalizes to other things. When a computer does the performance, it may not generalize in the same way. Go back to the ’90s, we had Deep Blue beat the world chess champion, [Garry] Kasparov. And now, you can get any number of different programs for a laptop or even your phone which have a higher chess rating than any person in the world in history. But a Chess Grandmaster is pretty good at teaching someone how to play chess better and coaching them. All these programs can do is say, ‘best move,’ ‘not best move;’ They don’t understand in the same way a person understands. If someone’s a Grandmaster, you’d expect them to be able to play tic-tac-toe. Well those chess playing programs can’t play tic-tac-toe. They can’t play any other game. They don’t even know they’re playing a game.”

On what robots we’ll see in our daily lives:
“I think right now what we’re seeing is the robotification of cars with driver-assist features. And I’m not as optimistic as some people that we’re gonna have fully self-driving cars everywhere anytime real soon. But other things in our lives will become robotic. It won’t be the science fiction robot, the Rosie of the Jetsons, but there’ll be robotic capabilities.”

On worries about future robots:
“Some people get all worked up about it and say we’ve gotta worry about these super-intelligent robots but they’re not even closed to being poorly intelligent yet.”