The AI Control Problem in a wider intellectual context — LessWrong