Pulling the diagram back out.

Large models are inherently theft and there is no amount of specious comparisons you can make or weaseling you can do to get around that fact

Nearly all of their value comes from the unconsented and uncompensated labor of others.

Your "niche use cases" rely on this fact.

It doesn't matter if there's no risk of output directly "infringing". The entire premise of the product is the theft of others' work.

I'm so bad at hashtags, LLMs are theft see above etc. etc.

#NoAI #AntiAI