I'm with you on this. It also occurs to me, though, that by rigorously labeling all offensive content, you give ai a huge boost in creating it with perfect fide

I'm with you on this. It also occurs to me, though, that by rigorously labeling all offensive content, you give ai a huge boost in creating it with perfect fidelity. Which inspires another really weird, super uncomfortable thought. If the CSAM producers had cheap, reliable methods of creating their awful content without the use of real people, would that reduce the harm done? I can't remember the last time I felt so conflicted just asking a question, but there it is.