Let me tell you a little story about a robot. Someone was training a robot to dual other robots. They would reward the robot for doing better each day. The robot learned that if it got shot it was bad. So it would run away. And it would be rewarded. But after a while it couldn’t do any better. The robot was very good at running away but it never did anything else. So it couldn’t improve.

It was trapped in what is called (mathimatically) a local maxima. It would have to do something that would make things worse to learn how to make them better. My point is it’s really easy to get stuck in one place and not be willing to take the risk to get out of the place you are stuck and find a better place.

(This is a true story about someone trying to build an battle bot AI of sorts, it would run away but never shoot.)