University of Wyoming Computer Science Professor Jeff Clune saw his research published this week showing that robots’ problem-solving skills can be improved by encouraging ‘creative thinking’ in artificial intelligence.
The research was accepted in ‘Proceedings of the Genetic and Evolutionary Computation Conference,’ a peer-reviewed publication.
The robots Clune and his team experimented with were rewarded when they ‘had ideas’ they never had before—basically when their simulated neurons displayed new patterns.
Clune says robots can get into a rut when trying to solve a problem—just like humans.
“No matter how hard you think about it or how hard you work on it, you can’t come up with a solution,” Clune said. “And then you might decide to go for a walk or do something completely different and all of the sudden you’ll think about that problem in a totally new way. And once you come at it from that new angle, you kind of have a creative inspiration and you’ll suddenly solve the problem. So we thought to ourselves, if that works for humans, maybe that can also work for robots.”
Clune says robots with more intelligent problem-solving skills could be used to do things like fight fires or clean up after nuclear disasters.
“What’s limiting us in terms of having robots help us in the world is not hardware," said Clune. It’s really making software that’s smart. So, if we can get robots that are much more intelligent and capable, then we can have all sorts of benefits for society.”
He says his team’s research is a step in that direction.
The lead writer behind this research was Jingyu Li who was a Laramie High School Senior while she was working on the project. She’ll attend MIT in the fall.