That’s fair. We can also expect proper moderation from social media sites. I’m okay with a light touch but It shouldn’t be floating around if you get what I mean.

Banning the tech, banning generated cp on the internet or banning it at home?

I’m a big advocate of AI and don’t personally want any kind of banning or censorship of the tools.

I don’t think it should be published on any kind of image sharing sites. I don’t hold people publishing it in high regard and I’m not against some kind of consequence. I generally view prison as unproductive though.

At home, I’m not sure. People imo can do what they want behind closed doors. I don’t want any kind of surveillance but I don’t know how I would react if it got brought up at a trial, as a kind of proof if the allegations have something to do with that theme (child molestation).

I also don’t think we need much of a reason to ban it on the web.

Kids will do things if they see other children doing it in pictures and videos. It’s easier to normalize sexual behavior with cp then without.

Although that’s true, such material can easily be used to groom children which is where I think the real danger lies.

I really wish they had excluded children in the datasets.

You can’t really put a stop to it anymore but I don’t think it should be something that’s normalized and accepted just because there isn’t a direct victim anymore. We are also talking about distribution here and not something being done in private at home.

We first use the DE-COP membership inference attack (Duarte et al. 2024) to determine whether a particular data sample was part of a target model’s training set. This works by quizzing an LLM with a multiple choice test to see whether it can identify original human-authored O’Reilly book paragraphs from machine-generated paraphrased alternatives that we present it with. If the model frequently correctly identifies the actual (human-generated) booktext (for books published during the model’s training period) then this likely indicates priormodel recognition (training) of that text.

I’m almost certain OpenAI trained on copyrighted content but this proves nothing other then it’s ability to distinguish between human and machine written text.

It’s sadly already happening in regards to stack.