The Dark Side of A.I.: Understanding the Dangers of Deepfake Images
blog.theabhishek.dev
blog.theabhishek.dev
There is absolutely nothing new, added to this discussion, by this article. It also seem gender provocative and charged, as the article choose to exclude male deepfakes as an equal probable problem, without even commenting on it? I suspect deepfakes to be just as valid for turning voters, blackmail, false imprisonment, and the likes - and that would include both genders.
This feels like bait/trash journalism, to me - On a topic that is worth a real discussion, in regard to real life problems,
What you should also do, dear, is assume your government is corrupted and that big tech companies are soulless Leviathans motivated only by extracting some easily quantifiable monetary profit from you, abusing you as much as it can within its accounting. At worst, you were wrong and actually your local government is only steered by competent benevolent people that will understand that trust doesn’t come without checks, and big tech organizations are actually all non-profit in disguise which are striving to make the world a better place for every single living being.
Also, the key issue here, as it at always been, is how people behave with each other, not which tools they use to actually implement these interactions.
The real challenge is not that much "how do you regulate tools", but "how do you make people act with mutual respect at societal level".
Am I missing something?
Doesn't really matter if the video is real, the people who this is done to have not signed up for this. And you know people are jerking off to it. I don't doubt for a second that it's absolutely repulsive to see yourself in a deepfake porn.
A couple of popular streamers who have had that happened to them have reported feeling similar feelings of shame and violation to the way they felt when they experienced actual sexual assault. That's pretty alarming and should count for something, no?
Also, this gives a completely new dimension to parasocial relationships. There are lots of people struggling with real realtionships that find solace in live streams. As they continue to engage in the community, they can begin to feel like they have an actual relationship with those streamers, which is obviously dangerous since it's such a one-sided thing. Deepfake porn elevates this even further and makes it that much more likely for people to develop an unhealthy obsession with people that they will likely never meet, instead of working on their real interpersonal relationships.
In other words, is the harm not in the production, but the wide distribution of the material, and the harm is in the reputation of the subject (even if explicitly labeled fake), and not in the actual contents?
And so this begs the question - if an AI could deep fake something on demand, without ever distributing it except to the requester, then is there any harm done?
But then the relationship breaks up and one of the partners publishes the footage. Or their account or device get compromised and the material goes public.
There's a series of new laws with regards to revenge porn, and deepfake porn will likely fall under that category. It's not about whether it stays private, it's the risk of it going public.
I mean this issue isn't anything new either, it's pretty straightforward and common to edit a celebrity's face onto a nude model.
Not really.
> And you know people are jerking off to it.
Why would I care what you're jerking off to? Maybe it's thoughts of me. Maybe it's photoshopped pictures of me. Maybe it's a deepfake video of me. But unless you phone me up in the middle of the night to tell me about it (i.e. harassment), I couldn't care less.
"Ah but what if someone emailed your grandma with a video of you having sex with ten guys" - well yeah, that would be bad but it would be bad if they emailed my grandma with a poem about the same. The harm there would be the intent to harass not some psychic damage caused by you creating images.
>A couple of popular streamers who have had that happened to them have reported feeling similar feelings of shame and violation to the way they felt when they experienced actual sexual assault. That's pretty alarming and should count for something, no?
It's bad that they had those feelings, yes. But feeling shame for what someone else does with your image is clearly irrational. Understandable maybe but I'm not sure we should making public policy on whether someone feels icky about something.
No. What's truly alarming to me is this notion that people have that they own and control "their" image and likeness. They don't.
Rich celebrities with lawyers get to extract rent from corporations for use of "their" likeness but there's pretty much nothing normal people can do against other people with a computer. At the end of the day it's just data that can be freely manipulated and like all privacy issues the only possible defense is to prevent that data from coming into existence in the first place. It's not even a new problem, even before stable diffusion you could easily find on the internet people photoshopping other people naked for kicks.
Which is why I don't have any sympathy at all for public figures, especially the ones who deliberately publish that data and profit off of the attention they receive whether emotionally or financially through advertising. They asked for the attention and they got it. Now they want government tyranny to control what kinds of attention they get? I don't accept that.
Not really?
> I don't doubt for a second that it's absolutely repulsive to see yourself in a deepfake porn.
Then don't watch it?
People jerk off to others all the time, it's harmless.
Also we really can’t ignore the lack of consent and the abhorrent violation of someone's image. This can’t be boiled down to growing “thicker skin” - there’s pretty major ethical considerations as well as the potential harm it can cause to a person's reputation and mental well-being.
That.
It's not something that's actually being done to them, unlike say, a snearing comment.
Edit: I worked in school over a decade, and I can definitely see the huge potential for harm here with digital bullying.
I can imagine they have an expert witness testify that it's fake or not, but they can and have been wrong as well.
Perhaps we'll see the rise of cryptographically verified videos: the camera signs the output of their sensors, proving it was captured from the real world.
This of course is pretty much impossible for all but most introverted or hermits.
Think of what happened with Hulk Hogan, who had a sex tape released without his permission. Though it was real, it left a lasting impression on a given audience, and it tarnished his reputation (further), and he successfully sued because, iirc, the video was “leaked” without his consent.
The thing is, you cannot guarantee that everyone will know it's fake, and when you weaponise it against people who can't distinguish it there'll be consequences.
Currently we see people found to be doing Adult Content online fired from their jobs, but what if a technically-blind boss gets sent a ton of deepfakes of an employee and is falsely told they do content like that on the side? What if a doubting partner suddenly receives deepfakes of their partner with someone else? What if your political rival starts spreading deepfakes with wildly rabid claims about yourself which the media picks up and reports on – bullshit has spread around the world by the time the truth has tied its shoelaces.
Think about the absolute inane content people believe on Facebook. Think about the misinformation that gets spread over Twitter just because the fake information would be funny. Think about the stuff you've seen online which turned out not to be so real.
Even if the images generated aren't real, the subject of them can't convince the world they're not, and there's no way for them to stop the violation of their self-image.
I'm having a hard time wrapping my head around this. Is the same true of doing so mentally?
My post specifically posits that the images are known to be fake and asks if THAT has any harm. Obviously framing people for things is a harm.
You might be thinking of celebrities where the idea that they have done porn is so far fetched that it's obvious it would be fake. But what about minor personalities like youtubers? It could tank a career if it's unclear something is real or not. Just like a false assault allegation, it will still cause a lot if harm even if the truth prevails.
Lastly, on an individual level, people are going to get hurt by sociopaths circling imagery of coworkers or whatever, it's messy human social business and the pragmatic person might imagine the truth is all that matters, but it definitely isn't. You can ruin someone's social life and mental health with mere rumours, let alone what will look like proof of ireputable activity.
Added benefit, as you say, is that any image can now be faked. So even a real photo of someone caught in the act, could be dismissed by the victim as fake if they wanted to make the problem go away.
It doesn't matter if they're fake or not for the purposes of blackmail though.
For any harm to happen, the status of the image as real needs to be confirmed and sustained over time.
If the victim says "it's fake", why would people doubt that or not care? That would just mean those who give oxygen to the fake images, are themselves in danger of spreading false information and losing all credibility. Only trolls and trashy people bypass concern for their own credibility.
As for "consent", who cares? It's a dirty fake image made by losers, and is the modern equivalent of words written on a public bathroom wall: "For a good time, call Jo". Jo never consented to that, but who cares, it's already content from the gutter.
I don't know what the laws should be, but I think there are some serious limitations on what's possible/practical at this point.
Why this focus on pornography?
Pornographic image editing has been around forever. Deepfakes is just another tool in the toolbox. This article fails to see the wider consequences of deepfakes outside of pornography as described by the Wiki page cited in the article.
I'd say the real danger is the potential to destabilize society by faking authority and spreading misinformation to manipulate people for financial or political goals. Pornography is a tiny part of that.
Even without logging in the majority of models market themselves with cute/scantly clad women. That’s why they focus on porn, because that’s what everyone is doing with AI image generators.
Although deepfakes, especially video + audio, can be more convincing than the older techniques.
We have here a case of the meaning of "indistinguishable" being deeply deepfaked, given the deepfake image shows naked v. the original clothed.
It seems the meaning of "indistinguishable" is being deeply deepfaked, given the deepfake image shows naked v. the original clothed.
Weaponised pornfakes could get people killed. Pizzagate happened.
Men are allowed to feel the same sense of privacy and intimacy over one's own body that we afford women, and pretending only women get harmed by the rampant abuse of photoshop on steroids doesn't do anybody any good.
It's not like this is some new occurrence. Just made easier by what's out. Anybody that will take the time to do these things would put the time into it through another medium. Just more AI fearmongering.
And of course... only mentioning women as victims. Right.