Separation from hyperexistential risk — LessWrong