Remember that the human won't only be thinking "better" or "just as good". They almost certainly can't avoid thinking "If I say 'better' here, what direction will that drive the algorithm in?" They don't just learn how to drive the plasma, they're learning how to drive Optometrist as well.