Reward Functions - History — LessWrong