Using a technique called reinforcement learning, a researcher at Google Brain has shown that virtual robots can redesign their body parts to help them navigate challenging obstacle courses--even if the solutions they come up with are completely bizarre.
Please to comment