Daniel Kokotajlo answers What problem would you like to see Reinforcement Learning applied to?