What's the state of AI safety in Japan? — LessWrong