In Spain, dozens of girls are reporting AI-generated nude photos of them being circulated at school: ‘My heart skipped a beat’
In Spain, dozens of girls are reporting AI-generated nude photos of them being circulated at school: ‘My heart skipped a beat’
While the police investigate, the mothers of the affected have organized to take action and try to stop those responsible
Police investigation remains open. The photo of one of the minors included a fly; that is the logo of Clothoff, the application that is presumably being used to create the images, which promotes its services with the slogan: “Undress anybody with our free service!”
You're viewing a single thread.
This was just a matter of time - and there isn't really that much the affected can do (and in some cases, should do). Shutting down that service is the correct thing - but that'll only buy a short amount of time: Training custom models is trivial nowadays, and both the skill and hardware to do so is in reach of the age group in question.
So in the long term we'll see that shift to images generated at home, by kids often too young to be prosecuted - and you won't be able to stop that unless you start outlawing most of AI image generation tools.
At least in Germany the dealing with child/youth pornography got badly botched by incompetent populists in the government - which would send any of those parents to jail for at least a year, if they take possession of one of those generated pictures. Having it sent to their phone and going to police for a complaint would be sufficient to get prosecution against them started.
There's one blessing coming out of that mess, though: For girls who did take pictures, and had them leaked, saying "they're AI generated" is becoming a plausible way out.
There's one blessing coming out of that mess, though: For girls who did take pictures, and had them leaked, saying "they're AI generated" is becoming a plausible way out.
Indeed, once the AI gets good enough, the value of pictures and videos will plummet to zero.
Ironically, in a sense we will revert back to the era before photography existed. To verify if something is real, we might have to rely on witness testimony.
Politics is about to get WILD
Dwayne Elizondo Mountain Dew Herbert Camacho approves!
Shit's going to get real emotional
I'd vote for Terry Crews. No lie.
To verify if something is real, we might have to rely on witness testimony.
This is not going to work. Just because images and videos become less reliable that doesn't mean we will forget about the fact that eyewitness testimony is very unreliable.
You say "forget" like it's not still incredibly common as evidence.
There's lots of data showing that eyewitnesses aren't reliable but that doesn't mean courts actually stopped relying on it. Ai making another form of evidence untrustworthy will result in eyewitnesses taking its place.
Indeed, once the AI gets good enough, the value of pictures and videos will plummet to zero.
This just isn't true. They will still be used to sexualise people, mostly girls and women, against their consent. It's no different from AI-generated child pornography. It does harm even if no 'real' people appear in the images.
Fucking horrible world we're forced to live in. Where's the fucking exit?
It is different than AI-generated CSAM because real people are actually being harmed by these deepfake images.
I was replying to someone who was claiming they aren't harmful as long as everyone knows they're fake. Maybe nitpick them, not me?
Reak kids are harmed by AI CSAM normalising a problem they should be seeking help for, not getting off on.
Permanently Deleted
Not getting beyond your first sentence here. I am not interested in what fucked up laws have been passed. Nor in engaging with someone who wants to argue that any form of child porn is somehow OK.
Permanently Deleted
Permanently Deleted
Im addressing you because you made the claim they are equivalent when they clearly are not.
No I didn't. Go nitpick someone else.
Or better still, explain why you think AI-generated CSAM isn't harmful. FFS
Let's be real here:
Sure, it's not illegal. But if I find "those kinds" of AI-generated images on someone's phone or computer, the fact that it's AI-generated will not improve my view of that person in any possible way.
Even if it's technically "legal".
They tellin' on themselves.
People who consume any kind of cp are dangerous and encouraging thar behavior is just as criminal. I'm glad that shit is illegal in most civilized countries.
Sauce that allowing computer generated cp causes more harm?
How is this place infested with so many fucking nonces?
I made no claims about "more harm" so what imaginary claim are you referring to in your attempt to justify CSAM?
Oh, so you want more harm. Curious.
A bit off topic, but I wonder if the entertainment industry as a whole is going to be completely destroyed by AI when it gets good enough.
I can totally see myself prompting “a movie about love in the style of Star Wars, with Ryan Gosling and Audrey Hepburn as the leads, directed by Alfred Hitchcock, written by Vincent Hugo.” And then what? It’s game over for any content creation.
Curious if I’ll see that kind of power at home (using open source tools) in my lifetime.
I envisage a world where your browsing Netflix, and based on past preferences some of the title cards are generated on the fly for you. Then based on what you click, the AI engine warms us and generates the film for you in real-time. Essentially indistinguishable from the majority of Hollywood regurgitation.
And because the script is just a series of autogenerated prompts, its like a choose your own adventure book, you can steer the narrative the way you want if you elect to. Otherwise it'll be good enough to keep most monkey brains happy and you won't even be able to tell the difference most of the time.
Then the real money will be in hipster retro human-generated movies
And it will work, because we've grown used to Hollywood being so repetitive.
I know it's impossible to perfectly predict future technology, but I believe AI will exist alongside traditional filmmaking. You'll NEVER get something with the emotional impact of Up or Schindler's List from an AI. You'll be able to make fun action or fantasy movies though, and like you said, fully customized for the viewer. I imagine it'll be like CGI vs traditional animation now - you only see the latter for passion projects, but for most uses, CGI works well enough.
This is already starting to happen for digital illustration. With better models and enough images saved, you can already train a model to replicate the art created by an artist.
Not so much replicate as simulate or produce art on the style of that artist.
AI can't replicate a piece of art unless it's only trained on that one piece of art, at which point you don't need an AI to make a copy anyway.
If you trained an AI on two paintings by the same artist, it will never produce either original painting, only blends of the two.
That is why I wrote replicate the art (color composition, image composition, style elements, tone, etc.) not the art pieces. I also never wrote one or two images but enough images ....
Cool, we are on the same page
Holy shit, I never thought of the whole witness testimony aspect. For some reason my mind was just like “well, nothing we see in videos or pictures is real anymore, guess everyone is just gonna devolve into believing whatever confirms their bias and argue endlessly about which pictures are fake and which are real.”
Witness testimony and live political interactions are going to become incredibly important for how our society views “the truth” in world events in the near future. I don’t know if I love or hate that.
Not necessarily, solutions can implemented. For example, footage from private security cameras can be sent to trusted establishment (trusted by the court at least) in real time which can be timestamped and stored (maybe not necessarily even stored there, encryption with timestamp may be enough). If source private camera and the network is secure, footage is also secure.
Permanently Deleted
I don't think that will matter very much considering how many real time video modifications we can do already today. Not to mention synthesizing video before the time it is supposed to take place.
Network security is a pretty big ask though - just look at how many unsecured cameras are around now. And once an attacker is in anything generated on that network becomes suspect - how do you know the security camera feed wasn't intercepted, manipulated, or replaced altogether?
Maybe there will be cameras as well that sign the pictures they take?
To verify if something is real, we might have to rely on
witness testimonyflagrancy.FTFY. Witness has never been that good a means to verify something is real.
Thats why we need Blockchain Technology..
Check Blockchain Camera for example: https://github.com/sv1sjp/Blockchain_Camera
Abstract:
Blockchain Camera provides an easy and safe way to capture and guarantee the existence of videos reducing the impact of modified videos as it can preserve the integrity and validity of videos using Blockchain Technology. Blockchain Camera sends to Ethereum Network the hash of each video and the time the video has been recorded in order to be able validate that a video is genuine and hasn't been modified using a Blockchain Camera Validation Tool.How exactly does that prevent someone from uploading a fake video?
The point is to know the time that a video has been uploaded as well as the previous and next videos from it for uses as security cameras, accidents in cars etc to be able to trust a video. (More information can be found on paper).
It won't, you'll just be able to verify a source
Not even that. It only allows you to verify that the source is identical to (the potentially wrong information) that was claimed at the time of recording by the person adding that information to the block chain. Blockchain, as usual, adds nothing here.
It proves that the video could not have been created at a later time.
Blockchain, as usual, adds nothing here.
it can add trust. If there's a trusted central authority where these hashes can be stored then there's no need for a blockchain. However, if there isn't, then a blockchain could be used instead, as long as it's big and established enough that everybody can agree that the data stored on it cannot be manipulated
but false, nonconsensual nudes are not collectible items that need to have their authenticity proven. they are there to destroy peoples' lives. even if 99% of people seeing your nude believe you it's not authnetic, it still affects you heavily
nonconsensual nudes are not collectible items that need to have their authenticity proven
of course not, but that's not what this comment thread is about. It's about this:
Ironically, in a sense we will revert back to the era before photography existed. To verify if something is real, we might have to rely on witness testimony.
that's where it can be very useful to store a fingerprint of a file in a trusted database, regardless of where that database gets its trust from
yeah but the problem is mere existance of tools allowing pornographic forgery, not verifying whether the material is real or not
How is that better than an immutable database where you guarantee trust simply by gettin your own public hash receipt for the database every time you introduce a new item? Why obfuscate things by riding the "Blockchain" hype bandwagon?
Who manages and guarantees that immutable database?
A nonprofit with multiple synchronized copies of the database and you can get your own copy, synchronize, fork it if you have the space, like a GitLab repository. Remember this is not for secure transactions and to prevent double-spending like a currency. It's just an additive database. You don't need to overkill with a blockchain.
Can you name a nonprofit you'd trust to manage court admissable evidence? How do you resolve differences that can pop up when forks don't agree?
Look, Git exists and image or document registration in an official onine database is Git diffs with less functionality because you can't remove previous commits: you just append new lines. This is a solved problem. If you're trying to solve a double-spend problem, then you need more than that, but it's overkill for your problem.
PS: maybe I'm oversimplifying it, but here's more discussion on this:
https://stackoverflow.com/questions/46192377/why-is-git-not-considered-a-block-chain
https://stackoverflow.com/questions/59509764/is-git-distributed-or-decentralized
because you can't remove previous commits: you just append new lines
I know it's frowned on to modify the history of a remote branch, and I haven't done much research on it because of that, however I'm fairly certain you can modify the history.
So....who hosts the gitlab/GitHub server that you'd trust to never manipulate the git history?
You still haven't answered my question of which nonprofit you'd think everyone would agree with should host such a service.
There are examples like DNS or the Mozilla foundation or all sorts of repos. Due to the receipt system you can verify if the commit history has been tampered with (your image has been removed from the database or edited). For court documents each court could host its own database where checksums are verified periodically, by "oracles".
There are examples like DNS or the Mozilla foundation or all sorts of repos.
Are you suggesting that these folks be in charge of maintaining the database? Seems like a very techbro solution. I personally wouldn't trust them to be responsible for all court admissable evidence because that's nowhere near their wheelhouse, but I do know of their positive track record. Good luck convincing the layperson to trust them.
Due to the receipt system you can verify if the commit history has been tampered with (your image has been removed from the database).
What do you mean? What receipt system? Afaik that's not a gitlab feature.
For court documents each court could host its own database where checksums are verified periodically, by "oracles".
And who would these oracles be, and how do you resolve differences?
I don't want to only ask questions, without contributing to the discussion myself, so I'll say the following.
I suspect your answer to how to resolve differences would be that the majority of databases that agree would be considered the "truth". How very.... block chain.
And as far as finding a solution that everyone would trust, what if it were truly decentralized, across citizens or even the world, like.... a block chain. No one organization would be in a position to edit anything.
I'm not saying it's the only solution, far from it. I'm just saying that while things like NFTs for art are dumb, there actually are a few applications where the features of a block chain actually fit quite well, and keeping of immutable, objective public records definitely strikes me as one of them.
Same goes for any deepfake. People are loosing their shit because we won't know what's real and what's not!.
We should have been teaching critical thinking a generation ago. Sagan was pleading for reform in the 90s. We can start teaching the next generation how to navigate the Information Age. What we can't do is make the world childproof.
Yeah, what I see happening is people end up not caring as much because there's going to be so much plausible AI generated crap that any real stuff will be lost in the noise.
Quelle für das angesprochene Gesetz bitte. Das will ich im Detail lesen.
Fang mit dem relativ neuen Fall hier an, und von da solltest du dann genug Info haben um selber zu suchen was die letzten Jahre passiert ist - das ist exakt das wovor damals gewarnt wurde, aber wer den hysterischen Irren die alles was irgendwie mit "Teenager entdecken Sexualitaet" mit dem Strafrecht erschlagen wollen mit durchdachten Argumenten kommt ist dann ja direkt auch ein Paedophiler.