Once AI systems start to self-improve their capabilities, ensuring they remain safe will require someone who is ready and ...