The Ideological Turing Test — LessWrong