Steven Byrnes comments on Learning human preferences: optimistic and pessimistic scenarios