David Scott Krueger (formerly: capybaralet) comments on When is unaligned AI morally valuable?