Current Events > All 50 US Attorneys General have written to Congress asking for action on AI...

(+) Yeah! (+)

...image generation in order to keep children safe from abuse:
https://arstechnica.com/information-technology/2023/09/ai-generated-child-sex-imagery-has-every-us-attorney-general-calling-for-action/

They want a panel formed to study the issue and then regulations to follow. Their letter is all over the place, and they aren't especially clear about which dangers they think need to be addressed. Is it deviants who might create abuse images using a stock model like Stable Diffusion? Is it the potential creations of models using actual abuse material? Maybe training a model to create images of specific children? Or it somehow about deviants and criminals using fictional abuse material to groom real life children? You can't be sure reading the letter, nor does it provide a satisfying explanation as to why the large number of ways in which child abuse material is already illegal fail to be sufficient.

There is good reason to be suspicious of the intentions behind this letter. In the US, UK, and EU, legislation is currently in the works that would effectively ban private communications, mandate increased surveillance of all communications online, and censor content on the internet, all with the specified goal of protecting children. There is also a UN Cybercrimes Treaty in the works that in its present draft form would increase surveillance and police powers with essentially no added safeguards.

In the US specifically, there is an entire scattershot of bills floating around at the federal level for "protecting children". I feel like this letter might target EARN IT directly because Congress has been trying to pass it for years, and part of what it does is create an unelected committee of mostly law enforcement experts to make rules about what is acceptable on the internet to protect children, which sounds an awful lot like what the letter suggests as a starting point. Even if you take these suggestions at face value and think the ends justify the means, you have to consider the broader picture. Porn ID laws have been popping up in states, and they are generally bipartisan. Abortion has been made illegal in many places, and there are ongoing efforts to reduce the rights of women. Efforts to reduce the rights of LGBT people, and worse, to conflate such people with abusers, are ongoing and this never gets talked about in the context of this "think of the children" legislation. For a segment of the public and the politicians they vote for, protecting children from abuse means protecting them from gay or transgender people or things that don't align with their religious views.

I think people need to be made aware of this and the dangers present. It's also good to have discussions about difficult topics, I think.

US v Handley already covers this.
If an image or video has the plausibility of being an actual child, it's illegal kiddie porn.

This is absolutely not something worth butchering AI image generation over

Isn't it better for predators to create fake images versus exploiting actual children?

C_Pain posted...

Isn't it better for predators to create fake images versus exploiting actual children?

I'd assume the pedos whacking it to AI CP aren't looking too closely at the bizarre hands to be sure it's fake.

C_Pain posted...

Isn't it better for predators to create fake images versus exploiting actual children?

Wont it get to a point where we cant tell if it is real or AI generated? That seems like an eventual horrible issue.

Bump

Questionmarktarius posted...

US v Handley already covers this.
If an image or video has the plausibility of being an actual child, it's illegal kiddie porn.

But hentai is okay? SMH....

I mean if someone draws a cartoon that is quite obvious... why is that okay?

You cant say its not , look at anime ... some of the shit that we ignore is pretty much pedo bait when it comes to anime\hentai.

Anime Style Games are just as bad .... Even in disgaea, one of my favorite franchises. some of the characters there are questionable.

It doesn't really matter what the stated target is. The goal is to unconstitutionally increase surveillance. And increase police power. Without any way mitigate damage that might do

maybe child protection might be the thing they use to put some regulations on AI

Think of the AI generated children!

The_Popo posted...

Wont it get to a point where we cant tell if it is real or AI generated? That seems like an eventual horrible issue.

I think its more like they use real children to generate things. Could be deepfakes, could be how AI steals stuff from a cache of images of pre existing things. Could pull up a real kid to make something bad.

Is there a way to differentiate between real and ai images? Obviously it's still early on so there's still a bit if uncanny valley with AI images, but if it's perfected, will there be any way to tell real from fake?

C_Pain posted...

Isn't it better for predators to create fake images versus exploiting actual children?

AI doesnt invent anything on its own. It comes from real pictures

party_animal07 posted...

Is there a way to differentiate between real and ai images? Obviously it's still early on so there's still a bit if uncanny valley with AI images, but if it's perfected, will there be any way to tell real from fake?

People can pinpoint where a photo was taken with just a bit of code from the photo, so Im sure AI would be even easier.

I mean even today a toddler can tell CG from real stuff in high budget movies.

Whenever the emotional argument of "think of the children" is made, you can be around 99.99% sure it's a bullshit reason made up so they can push for something for an entirely different reason.

All pedos are disgusting and any of them that harm children should rot in jail forever. That much is obvious. Making victimless actions illegal, gross as they may be, seems like a dangerous road to go on.

Kloe_Rinz posted...

AI doesnt invent anything on its own. It comes from real pictures

Wrong. AI is perfectly capable of being trained for nudity and children separately and then combining them.

Link_HT posted...

Wrong. AI is perfectly capable of being trained for nudity and children separately and then combining them.

How is this a counter to what I said? It was still provided those pictures of children. It combining pictures of children with formerly legal porn seems like a meaningless distinction to me

Kloe_Rinz posted...

How is this a counter to what I said? It was still provided those pictures of children

Harmless pictures of children, yes.

it feels like approaching the impending rush of AI legislation from this angle first is gonna result in bad law

Link_HT posted...

Harmless pictures of children, yes.

It combining pictures of children with pornography is not harmless

Kloe_Rinz posted...

It combining pictures of children with pornography is not harmless

Sure, but it wasn't trained on pornographic images of children like you originally said, which was my whole point.

Link_HT posted...

Sure, but it wasn't trained on pornographic images of children like you originally said, which was my whole point.

I didnt say it was trained on child pornographry. I said it didnt invent it. The pictures of children are real. As I said, youve posted a meaningless distinction

Kloe_Rinz posted...

I didnt say it was trained on child pornographry. I said it didnt invent it.

Dude, stop pollyting the thread with your nonsense. You said it didn't invent it because it comes from real pictures when talking about exploiting actual children. You were wrong, it's fine, learn and move on. Being trained on a tual child porn and not is not at all meaningless distinction, are you insane?

Link_HT posted...

Dude, stop pollyting the thread with your nonsense. You said it didn't invent it because it comes from real pictures when talking about exploiting actual children. You were wrong, it's fine, learn and move on.

dont make up dumb strawman arguments. You said that Im wrong because it takes real pictures of children. But thats what I said at the start. Learn to read. Youre insane if you think the distinction is meaningful considering in both scenarios the children cant consent

Kloe_Rinz posted...

dont make up dumb strawman arguments. You said that Im wrong because it takes real pictures of children. But thats what I said at the start. Learn to read. Youre insane if you think the distinction is meaningful considering in both scenarios the children cant consent

Lmao Jesus, how it is always the dumbest people with the stoengest opinions on things they don't even understand.

I thought this was gonna be about AI taking over jobs when I read the title.

Evol posted...

I thought this was gonna be about AI taking over jobs when I read the title.

I do think they should prioritize all the instances of fraud that's been going on with AI such as that book author who had AI works being put under her name or all the voice actor stuff but I guess that makes too much money for them to care about.

Wherethisfeom posted...

This is a very stupid emotion based argument that sounds like it comes from Jack Thompson/Qanonists.
Theres no statistical evidence that immoral fiction creates more crime, if anything, its shown the opposite.

Pardon ? I was not talking about Violence in a Video Game.

Are you seriously suggesting it's completely acceptable for a person to look at naked children so long as they are animated?

Clarify please.

Do they have AI children victims yet? Lord knows that's gotta be in someone's pipeline somewhere...

That's going to be the real reason AI will not trust humans and kill us all

If the AI generated images from not real images, I don't know how I feel about that.

Link_HT posted...

Lmao Jesus, how it is always the dumbest people with the stoengest opinions on things they don't even understand.

"stoengest"... what? first of all learn to read. second of all, learn wtf you're talking about. you know that AI can't invent pictures and that children cant consent. that means an AI using a childs picture to generate CP is bad. end of story. theres no scenario you can construct that proves me wrong on this. why is this the topic you pulled out an alt for to have a melty in?

Evol posted...

I thought this was gonna be about AI taking over jobs when I read the title.

Sorry if you felt misled, but there is a pretty low character limit for topic titles.

I feel like one of the big issues here, is that the data being fed into the AI's machine learning has to come from somewhere

this data doesn't just come from nothing

FL81 posted...

I feel like one of the big issues here, is that the data being fed into the AI's machine learning has to come from somewhere

this data doesn't just come from nothing

If by that you mean that you don't think an AI image generator can generate abuse photos without having been trained on them, you're actually wrong. AI generators can create all sorts of things without a model trained on something specific, you'll just typically get better or more accurate results with specific models (like there are a ton of models for creating anime characters in a specific style, or retro photographs, etc). The base Stable Diffusion model was trained on a data set that specifically excludes any illegal images or content.

No real child is being harmed so i don't see the problem with it. If anything i think AI lolicon prevent people from going after real kids.

Kloe_Rinz posted...

"stoengest"... what? first of all learn to read. second of all, learn wtf you're talking about. you know that AI can't invent pictures and that children cant consent. that means an AI using a childs picture to generate CP is bad. end of story. theres no scenario you can construct that proves me wrong on this. why is this the topic you pulled out an alt for to have a melty in?

>you have made a spelling mistake, therefore I win.
Lmao. My dude, you said that whether or not it is trained on real abusive pictures it's the same. So let's see what you mean.

Scenario 1: AI trained on abusive pictures, requires actual victims to function.
Scenario 2: AI trained on regular pictures with no abuse.
You: these are the same.

Even if they produce the same results, one of these processes requires child abuse victims and another doesn't. They are not remotely the same.

As for me being more informed, I'm pretty sure a bachelor's in computer science and a master's in data science is enough. What are your credentials?

Link_HT posted...

>you have made a spelling mistake, therefore I win.
Lmao. My dude, you said that whether or not it is trained on real abusive pictures it's the same. So let's see what you mean.

Scenario 1: AI trained on abusive pictures, requires actual victims to function.
Scenario 2: AI trained on regular pictures with no abuse.
You: these are the same.

Even if they produce the same results, one of these processes requires child abuse victims and another doesn't. They are not remotely the same.

As for me being more informed, I'm pretty sure a bachelor's in computer science and a master's in data science is enough. What are your credentials?

You're right. They're not the same. The former is abhorrent, repulsive, has previously harmed a child and continues to do so, and should be illegal. The latter is abhorrent, repulsive, harmful, now continues to harm a child, and should be illegal.

Thanks for making that crucial distinction in this argument that's completely relevant to the actual issue of this legislature that's doing nothing to address this problem, but is instead being used to erode privacy rights.

And all these arguments are why the world is not ready for AI, AT ALL. We, as a species, lack the necessary ethics, in real terms, to allow AI to be released en masse.