Garrison Lovely on the risks posed by AI, why company self-regulation is not enough, and what we can do about it.