Comment by WatchDog
4 days ago
Whatever you think about X's image generation models, I don't see how it is related to the petition that the EFF is opposing.
Is generation of non-consensual imagery really a privacy issue?
If someone publishes a real naked photo of you, that was acquired without your consent, that would be a privacy issue.
If someone generates a naked photo of you, even if it looks identical to a real photo, it's not your private data.
I think the terminology here can confuse the issue a bit. And because its such a socially pernicious and stigmatized topic, it's hard to even talk about the phrasing without without raising doubts as to why you would get caught up in the weeds on such an issue. But, I would say there is use in making the distinction between something that is CSAM, where its a record of abuse that has happened to a real child, and sexualized depictions of children, or content promoting the sexualization of children. The social and personal harms are distinct, and if we are to firmly understand the arguments for/against guardrails on generative AI, then it's a distinction that needs to be made I think.
A problem you'd probably run into here is that it would be rather difficult to prove that no real CSAM was involved in the process of making the somehow-okay fake CSAM. Image generation models require training sets, after all. Do the companies training these models have the necessary data and evidence to prove every individual in every training image was over 18? Is generating fake CSAM okay if you trained the model on non-CSAM photos of real kids? I don't think so.
There are of course situations where being aggressive about this can hinder people's freedoms - like an adult who looks 'too young' having their freedoms curtailed because any photos or videos would Look Like CSAM - but I don't think they're common enough harms to justify holding back on regulation here.
> like an adult who looks 'too young' having their freedoms curtailed because any photos or videos would Look Like CSAM - but I don't think they're common enough harms
It is extremely common in some parts of the world. Age from visual appearance basically doesn't work inter-culturally, with or without AI, and ... there are people amassing something like 1/3 of the world that look like kids until they must retire from all work including use of gas stoves to boil water. Said people just literally look like kids, attracts wrong kinds of the rest 2/3, wrongly reject them if others come over, and gets banned on social media as being underaged, all the time.
edit: I have feeling that the concept of "adults by appearance" might be the case of suspicious discontinuities created by industrial revolution; it is often said that the modern concept of binary child/adults dichotomy is the result of industrialization though the concept of child always existed. For this reason, there might have been selective pressure towards attaining "definitely adult" appearances at younger ages in forerunner regions, and less such pressure at regions that followed it. IMO that makes more sense than assuming people from some regions are obsessed with certain things.
3 replies →
Personally I think that even if the training data, is made of images of people who are over 18, the state of the adult industry is that there is a lot of harm and exploitation involved in that even. Do we have the evidence necessary to prove every over 18 person who's images are in the training data are okay with that use? So there is no somehow-okay anything. But if were dealing with models that can combine concepts .. then "adult" + "sex" = "sexualized adult" so "child" + "sex" = "sexualized child" .. this is just a fundamental capability of the technology. Without even getting into nudity where its easy to see how non CSAM, non-sexualize medical images etc, can fill in the gaps for the model. This is why I think just calling it CSAM is confusing the arguments. I see the base harm as being the promotion of sexualization of children. And this is more a harm and risk to society and to people consuming these images. Of course if you extend this to "real childs image" + "sex" to get a sexualized deepfake then there is a kind of harm to that individual too, but its still a disctint thing from physical abuse.. This is just my instincts on this but I'd be really interested to see some real discussion between people with actual extensive understand and experience of all these areas.
Hand drawings and photoshopping a minor's face onto an adult body are already illegal. Having a machine do the work automatically doesn't change that.
1 reply →
Frankly, I don't think anyone cares that these images are generated using CSAM or non-consensual imagery. If I generate a perfectly ordinary picture of a person and the model somehow used CSAM or non-consensual imagery to do it, no one would care. Conversely, if we could prove a model generated gross ass images using nothing gross ass, everyone currently objecting would still object. So, let's just say what we mean here: we have a moral problem with people generating, storing, and sharing these images.
> If someone generates a naked photo of you, even if it looks identical to a real photo, it's not your private data.
"You see, your honor, it's not a picture of them, it's a picture of their reflection in the mirror."
I feel like this discussion is a question of what the exact structure of the hydrogen-filled blimp should look like, and not a discussion of the fact that THE BLIMP IS FILLED WITH HYDROGEN.
Like we got so deep into the lawyered-definition of words, that we skipped right over the clearly wrong/awful intent.
> Like we got so deep into the lawyered-definition of words,
Perhaps try reading "your private data" again - slowly.
What do you think my comment was responding to? Literally those are the words that I was responding to.
Saying that the provenance of a naked image of myself is relevant to the fact that it's a naked image of _me_ and I don't have ownership over my own image is exactly the kind of lawyer-brained wording I was referring to when I made the comment about the mirror. "It's not a picture of you, it's a picture of a reflection of you."
9 replies →
In an ideal world, generated non-consensual imagery should be illegal through invasion of privacy through misappropriation of name or likeness, but I think only a limited number of states have those laws.
In an ideal world, defamation should not be illegal through invasion of specifically privacy. Because that shields perpetrators from clean-room defamations, which is wrong.
I guess I live in such an ideal world? Defamation don't require proofs of breach of privacy where I am, only damages. I've heard that this setup is not universal.
Should it apply also to drawings? Detailed descriptions? Strong mental visualisations?
Is it private data if it was scraped from my social media profile marked private but leaked through a shared party? My expectation was that image would only be shared with those I wanted to see it (form of privacy).
That's fair. Is there any indication as to if xAi trains on private profiles?
Is there any indication it doesn’t. From what you know about LLM teams, products, and engineers, especially XAI, does the word “private” or “Disallow: <my-private-profile-path>” seem to stop them?
"CSAM" is a codeword for anime. Users of this term routinely reject focusing on real kids and abuses. I assume "privacy" must be therefore a codeword too, especially considering that nonconsensual shocking images can be handled by defamation laws than privacy laws and principles.
I would hope we can agree that aggressive policing of anime and cartoons is a bad thing without denying the real existence of CSAM - the actual thing - or denying the bad things that have to occur for it to exist
I never denied the real existence of CSAM, I'm just saying that there are people who use "CSAM" to specifically mean anime and anime related photos and avoids discussion of actually detecting or preventing sexual abuses. I think I've seen it blatantly said on organizational blog posts somewhere that trying to move focus to kids is disingenuous or something.
7 replies →
[flagged]