Tahp comments on Is AI alignment a purely functional property?