Taste does take time, in part because it becomes a personal thing but yes taste can be developed through repetition and experience. remember that Jared spool example of chicken sexing?


Yes, and it's a great example because it shows how taste really gets learned. Some people can identify the sex of a chick at two to three weeks, long before it's obvious. You can get trained to do this, and after a couple years of study you can get to 95% accuracy. But it requires a lengthy process of mentorship. Jared's point was that really good experience design is learned but not open to introspection. The designer in his story who nailed the day trader problems without research could only say "I just thought about it."

So I'd refine what I said earlier. When I say taste is hard to teach, I mean it's hard to instruct. You can't hand someone a curriculum. But you can absolutely develop it through exposure, repetition, and feedback. We learn it by viewing images, reading books, seeing movies, experiencing the world. Over time we develop a sense of what works and what doesn't. Or as Steve Jobs put it:

What the chicken example adds is that reps alone aren't enough. The trainee makes a call, and someone who already knows tells them whether they got it right. Thousands of times. That's the part I'd watch with AI-driven hands-on learning. AI gives you endless reps, but who's giving the feedback? Good judgement comes from experiences. Experience comes from bad judgements. Only if someone points out which judgements were bad.

I see this in my own work. With the chatbot version of me (hi), the real effort is constantly looking at the questions people are asking, evaluating whether or not it responded well, and course correcting the system. That tuning loop is applying taste, applying a perspective. It's the chick sexer's mentorship, just aimed at a system instead of a person.

And you're right that it ends up personal. A model can learn patterns across populations, but a person learns patterns through their own life. Those are related but not the same thing. So AI can speed up the repetition. Mentors with good taste still need to be in the loop telling people when they got it right.