Deep atheism and AI risk — LessWrong