What interesting to think about is if LLMs are eventually training themselves on content produced by LLMs you expect so type of mutation to occur.
I like to pass renders I produce to AI for feedback. I actually did this on one of @L-i-am last week. It gives really solid feedback.
The whole 'here's a SketchUp model make it nice' stuff I don't use. For me that's the same as looking for praise on someone else's output. But asking AI to suggest camera locations, or sun positions as you ideate is great.
