My Updating Thoughts on AI policy — LessWrong