gwern comments on Your best arguments for risks from AI