What problem would you like to see Reinforcement Learning applied to? — LessWrong