Part of my inspiration to make these icons was me thinking about the qualitative difference between my own use of ML technologies in my no-budget cartoon I'm making, and the """"AI anime"""" type stuff that rightfully disgusts any sensible person. The most obvious difference I can think of is that my own ML software I use — Demucs to make instrumental versions of songs, Retrieval-based Voice Conversion to let me voice act outside my normal vocal range — don't involve prompts. Because if you're writing prompts, generally speaking, you are functionally not the artist so much as you are the commissioner, right? A skilless, creatively bankrupt, entitled consumer.

Things are of course not this black and white in reality, though: you can be skilless and creatively bankrupt using non-prompt-based software, or even skilled and creative using prompt-based software, simply because there are so many different tools and different ways to use them. Just speaking generally, though, I do believe that prompt-based software is used to make way more slop than non-prompt-based software, just because of the way you interact with prompt-based software. But the fact that "AI" isn't black and white is why, as I was saying in the post body, an "AI" classification system needs a lot of refinements and testing to figure out what actually is the most important information and how to communicate that information most effectively.