Mark T. Tomczak on Nostr: Yeah, I'm not looking forward to it, but my understanding of the legal framework here ...

Yeah, I'm not looking forward to it, but my understanding of the legal framework here makes me a bit doomer on the whole thing in much the same way that overturning Roe v Wade was awful but I can understand the legal thinking.

We've known forever that you can't block ownership of CSAM on ownership grounds alone because of the First Amendment. The precedent is pretty clear vis-a-vis the Hustler trial on ownership and private, discrete trafficking of consensually-generated smut.

So CSAM being illegal has been couched in the notion that creation of it is automatically harm and that it is legally impossible for underaged people to consent; there is no possible way to generate it that didn't inflict harm on a child. The trafficking ban is based on slightly shakier ground imho, that it creates a market for harming children... I am not a lawyer so I'm not going to pretend to understand how secondary effect arguments like that hold water, I'm just going to accept they do.

... But that was always a justification for banning it that has been waiting on somebody to come up with a solution for creating without harming children.

And the wild thing about how AI generators work is that... Yeah. It's going to have to be figured out in a court of law whether they were trained on actual CSAM or are extrapolating. When it comes to image generation I don't actually know what the technology is capable of here. I am pretty certain that LLM text generation could extrapolate Lolita without having been trained on a Lolita story, though.

Mostly I would not want to be X going through the discovery process that will involve divulging their training data though, because I am real certain no human being has laid eyes on every single image that went into the hopper. So maybe we all get lucky and they fucked up bad.

Gross. I'm going to go think about something much nicer now.