Hypothetical: suppose you live in a world where Zootopia was never created. Would it be possible to prompt an AI image generator to create Zootopia-style characters? What would your prompt be? How would you prevent it from looking too much like previous styles? How about a world where Seinfeld never existed. Would it be possible to prompt an AI image generator to reproduce Costanza's expression here? What would you even say? I wonder if maybe some things are just unpromptable. Maybe text isn't actually the universal interface. Maybe a picture isn't worth 1000 words, maybe it simply can't be compressed into words at all.

I was also thinking about this in regard to script-driven versus storyboard-driven cartoons. A script can say so-and-so makes a confused face, but it seems like that will never be able to capture the exact funny face the artist would conceive. The script is actually lossy compression of the art.