Ask HN: What's the Fuzz about Taste in AI?

Recently, I've encountered CTOs and CPOs who argue that 'taste' is the key differentiator humans bring to AI processes. I'm struggling to grasp this concept. For instance, in self-driving technology, I prioritize performance over aesthetics—does 'taste' even apply here? While I understand that 'taste' may refer to the lack of sufficient data for AI models, I question whether human judgment truly makes a significant difference. Is this just a way of saying we believe we know better than AI? What are your thoughts on the role of human taste in AI development?