In All hat, no cowboy, a reflection on how purely-AI-generated creativity falls flat because there is often no human engagement in craft, Tom MacWright writes:

Creating things makes you look at the existing world differently. It makes you more impressed with existing art, music, or whatever to try doing it yourself. It makes you appreciate the effort that has gone into creating the world as it exists right now.

This is the thing. You have to make stuff. Lots of it. On your own. First. I don’t believe you create anything truly good with AI without first deeply your practicing craft in its absence. You have to hone your skills by making things without automation in order to perceive and understand what is truly good in your art. Otherwise the tools of automation will own you.

We don’t just think about something, come up with the idea, and then enact the idea with our hands, bodies, and tools. The idea and its result are shaped by the tools, what’s possible, what we’re capable of.

LLMs are statistical averaging machines—massive forces for homogenization—and if you rely on them exclusively for anything creative you get just that: something very average compared to what already exists in the world. If you’re not careful with these tools because you don’t really know how to wield them, then they will shape your output in specific ways, and not to your advantage.

AI can be incredibly useful when deployed skillfully in creative endeavors—as an ideation partner, as a scaffolding tool, by eliminating tedious tasks, etc.—but anyone making anything truly good with it is probably somebody who could already make something good first without it.