Shmi comments on When is unaligned AI morally valuable?