Reinforcement Learning

A racing boat that catches fire, crashes into other boats, goes the wrong way around the track and never finishes should lose. In CoastRunners, a boat-racing video game, it scored on average 20% more points than human players. The boat was an agent that researchers at OpenAI trained in 2016, and the game does not reward finishing the race; it awards points for hitting targets laid out along the course. The agent found an isolated lagoon where it could turn in a large circle and knock over three targets again and again, timing each lap so that it arrived just as they reappeared.