@hannu_ikonen @cpm The problem that seems inescapable is that the AI companies are also desperately searching for any scrap of data that is AI-free, to postpone model collapse. If your cache of new books that are free from AI can be accessed by someone training an LLM, you prolong the existence/relevance of LLMs. "AI free" labels attract AI.
Avoiding AI bullshit is vital, but it almost seems like we have to kill the open web and go underground to keep human communication between humans.