Well, we're approaching the 25th anniversary of Columbine. But 35 years ago, my neighborhood was equipped with quite a lot of Uzis and Mac-11s, and we had quite a lot of mass shootings. Only they were called "drive-bys" and nobody circled the wagons to preserve the historical tradition of access to such equipment.
At what point did ordinary people get easy access to weapons that have high capacity and rapid fire? That's when we see a rise in mass shootings. Today, ordinary people are getting access to tools which can quickly and easily make better fakes than experts with photoshop could do even 5 years ago.
No they're not. The first reconized school shooting was in 1764, with the Enoch Brown school massacre. There have been school shootings ever since. (Unless you want to argue that the 1700s wasn't that long ago, which it also wasn't, but I don't think that's where you're coming from.)
Oh, aye? What's the record on mass decapitations at knifepoint?
And are you actually comparing a trained militia to randos buying military hardware without any accountability or oversight? Because generative AI isn't only accessible to trained and licensed practitioners, is it?
Both of those are deceptive comparisons: that page will tell you that China has had 90 fatalities in total since 2010 whereas despite being a third the size we have had 13,500 mass shooting deaths. Similarly, while the Bath bombing killed about a third more people than the deadliest school shooting (only ⅔ of the deadliest mass shooting) the difference is that you had to go back a century for that example whereas the shooting toll for any recent decade is higher.
The common factor is easy access to firearms. Every country has mentally ill people, every country has embittered people who are mad at the world, but only the United States has removed the restrictions on gun access which allow those people to engage in mass shootings. That started in the late 1970s and the contrast with our peers is staggering: Australia has many cultural similarities but they actually value life and tightened their gun control laws after the Port Arthur shooting, while we’ve loosened ours and have hundreds of thousands of more deaths.
The article fails to realise that the exploitation has nothing to do with AI, has existed for the last two decades, can be done with Photoshop. There is literally nothing that’s “AI” in the article.
It makes good clickbait headline though.
Also there are existing laws for protecting against defamation.
I think there is the implication that any idiot can do it now and there has been an explosion in proliferation as the new tech has lowered the barriers to producing content.
It’s like saying that since you could carry around a Morse transmitter in 1920 that texting wasn’t something really new.
Perhaps not, but ease of use really can be transformative on how something affects society. Especially since someone who has to spend years learning how to do something is less likely to do stupid shit with it than someone who just has to click a few buttons.
Any file on a computer can be generated by writing it the contents down in hexadecimal on sheepskin in ink made from walnuts, then scanning in and OCRing that data.
Will you claim that, therefore, NO software can possibly have any societal impact? Because, after all, it’s already “possible” to do what that software does?
The problem is that people like you have no idea what they are talking about this way worse than mere badly photoshopped images. They look real. I gladly do a little experiment if you send me one picture of yours and the address of your workplace.
Browse around the "about us" that many companies have on their website. Most included head shots. Generate and send some images. I doubt you will get much of a reaction.
This could actually be done long before Photoshop came along. You would simply mix some paint on a palette to match the person's skin tone, and then paint over the person's clothes to "remove" them.
It is not easier to generate a convincing image. The amount of time it would take to learn how to clip a face, and blend it in Photoshop onto a real image would take less time than learning how to configure an advanced AI workflow to do the face swap, fix the hands, skin, etc.
If anything, I would argue what increases devious interest in this topic is the media talking about it as if it was new. Thats a quick way to get a lot of people "involved" on both sides of the debate.
Just doing a little digging on the internet. You can create a free email address, log in to promptchan.ai, upload a picture, select the region of the body you want to unclothe and type “nude”.
Yes, I had to do about 5 minutes research on the internet to figure it out.
Because of Apple’s rules, it won’t let you “unclothe” pictures. But the website will.
As a 50 year old married dude who never delved into that side of the world (not claiming I’ve never looked at porn though), I didn’t know how convincing it could do it for adults with shadows, skin tones, etc.
Face swap is completely different from what the tools can do, they can match skin tone, shadows the quality and grain of the pictures and “create” nude images from full body pictures.
You can do this for free from your phone on at least one website that I’m not going to keep repeating.
I had never heard of this world until seeing posts about it on HN and it was just today that I delved into the bowels of the internet to see how well it worked so I could talk intelligently about it.
I understand your point. However, I am saying, the barrier was already low and the activity was already happening before. It was not a big deal until people smelled AI has a lot money and let's make it a big deal.
Just use the existing laws and social norms to punish badly behaving teenagers, no need to invent new AI laws for that.
You really think the barrier was low to create almost perfectly realistic nudes that matched skin tones when you had to find a nude picture that matched the skin tone, film grain, etc and use photoshop from a computer than it is now when you can just create a free email account, go to a website, upload a picture, select the body part and click “edit”?
It was never as easy or happening with the scale or as easy to make it look realistic.
Doing that in Photoshop was not something the average middle schooler could do realistically - let alone ok their phone, for free
This is an article about pasting one face over the top of another face and using a layer mask to blend it. The only automated parts are the object selection tool, which helps you select the face for copying and pasting, and the match color feature, which helps you match the tone and brightness of the photo you're attempting to blend on top of.
I used to do Photoshop image manipulation for work. Performing a convincing face swap with the tools mentioned in your article was difficult unless the light sources, face angles, and focal lengths of the cameras that took the photos happened to match very closely. You could paint in your own lighting if you were very skilled, but it's the sort of thing that took years of practice to get good at, and it still wasn't very convincing most of the time.
> Performing a convincing face swap with the tools mentioned in your article was difficult unless the light sources, face angles, and focal lengths of the cameras that took the photos happened to match very closely.
When you have a million images to pick from, getting a reasonable match is a lot easier.
I can't stand it when a boomer's version of a same service says "contact for quote", and the sales agent acts confused and perplexed about the term "self service" when you ask if such a thing is on the roadmap
(this wasn't about criminal websites, just self service e-commerce)
Oh, contrair, I think that a.i. flips it on its head. Nudes proof nothing but the weird obsession of the owner nowadays. It's quite socially liberating
...
If your examples are not logically equivalent even using abstraction, some people may be less impressed than yourself unfortunately. I do appreciate the fun aspect of it though.
The general form of the argument is just “when things happen very frequently, people get used to them.”
This is both true and a terrible reason to be okay with bad things proliferating. And it’s especially suspicious when this argument is leveled by people who generally aren’t going to be victimized by said bad things.
All of these things you’re listing are tangible events that cause specific harm to someone.
A nude photo is information. If you can turn it from implied authenticity to implied inauthenticity then the harm is reduced. The argument here is not about being used to things.
Does harassment to the point of job loss, familial dysfunction, or suicide count as tangible harm or not tangible?
Sure, if you can do that. Of course the question is whether/when that’ll happen and how many young people have to be victimized first in order to get there — and for what benefit?
Was your claim above not absolute in terms of outcomes? I mean sure, you were "just joking" I suppose, but what's the point of conversation if we normalize bait and switching?
Umm I made assertions in which I don't believe, of the shape "Near term bad thing is okay because in the long run it might have good second order consequences."
Honestly I think it’s unlikely that this will affect many people at all. The harassment is the problem and the ai photos don’t really make it any easier to do.
Maybe we’re entering a world where instead of saying suck a dick, you send someone a gif of them sucking off their dad, but as far as actual consequences go I think it’s pretty much immediately discreditable. Plausible deniability is all people ever really need.
The “this is no different than before!” argument is dismissible on its face. As evidenced by NVDA’s trillion dollar valuation and the hundreds of billions being deployed into AI companies — that’s not because they’re building stuff that makes the world just the same as it was before.
Ah yes, middle school/high school bullies: famously disempowered by things like plausible deniability. The suffering caused by bullying and harassment: powerless in the light of plausible deniability.
AFAICT the people who would be close to this issue (e.g. people who pay attention to online harassment, teen mental health, and child pornography) come to a different prediction as far as impact. I’m far more inclined to believe them than people leveling an argument that, as I said, can just as readily defend murder, car thefts, and medical malpractice. All the suffering is just in people’s heads and all of it can be eliminated by proliferating the behavior to sufficient degree.
> The “this is no different than before!” argument is dismissible on its face. As evidenced by NVDA’s trillion dollar valuation and the hundreds of billions being deployed into AI companies — that’s not because they’re building stuff that makes the world just the same as it was before.
By this logic, you need to worry about me using AI to seduce your mom. Because NVIDIA spent a TRILLION DOLLARS!
> Ah yes, middle school/high school bullies: famously disempowered by things like plausible deniability. The suffering caused by bullying and harassment: powerless in the light of plausible deniability.
If bullies can push things as truth regardless of their merit then false information doesn’t matter.
Sure if you ignore that generating and modifying images is one of the leading usecases. Anyhow I think the combination "your mom" joke and thinking valuation == spend has revealed this to be a fruitless conversation.
My inaccuracy is a testament to how lame of an argument it is to say this is an area of high monetary investment ergo it must be salient to teens harassing each other
Just laying out my understanding of where we are right now. Tell me where it's wrong:
Companies: We're raising billions of dollars to make it cheaper, faster, and easier to produce extremely believable fake media. Our success in this endeavor is verifiable by opening one of thousands of new image generator tools and trying it yourself for 30 seconds.
This is imperfect and only covers the easiest aspect of the problem: leaking of consensual nudes, where the images are real but you didn’t want them to be public. There’s still a ton of damage possible by providing plausible images of things that were never real: middle school can be toxic and kids know how to pick the most damaging things, and the initial description and circulation will stick to some extent no matter what – spread photos of someone having sex with the least popular student and the victim has to prove a negative or wonder how much everyone around them is probably weighing the “maybe it’s real?” odds even if they’re saying they don’t believe it, and the other kid is going to live with knowing that someone thought they were the best way to degrade someone else. It probably won’t be hot porn star sex, either, but whatever they picked as the least flattering option - the kind of thing the target knows people will remember when they think about them.
Worse, some of the options force criminal investigations. If they fake a video with an adult, now that kid has to deal with the police investigating a very sensitive topic. If they pick a teacher or family member, etc. that also means those people might have their jobs jeopardized and it could be very hard to disprove if the fakes don’t have obvious errors.
I think fake nudes are shocking to the victims because even if fake, the pics/videos depict them as sexual objects. One usually just wants to be seen at in a sexual way by intimate partners and with one's consent, but imagine finding out there are fake nudes of you, and that plenty of horny people have been looking at you sexually. Even if it's not your actual body, it'll still feel violating.
Especially because there’s often an intention to degrade the target. It’s not just someone’s private fantasy, it’s usually shared in the context of them being able to do something without the victim’s consent which is going to be hard to look past.
Why are so many comments working so hard to dismiss this article? There’s a mountain of research showing how social media is damaging kids and the increases in suicides back this up. Yet somehow tearing this article apart is the solution? Is it just a mass of AI apologists who think it is just a harmless tool, like a gun or an atom bomb? Or is it just “this is something I like and the consequences don’t hurt me so everyone else can eff off.”
There is a contrarianism that runs on HN and adjacent communities that also comes with a myopia of empathy. I observe about one article per month where there's a comments section like this expressing opinions that if you ever were to say out-loud in a social setting you would get some weird looks.
This isn’t about the dangers of social media. This is about the ability for anyone to log into a site for free and create a convincing looking nude of an underage girl for free and spread it.
Your comment is so fraught and keeps on jumping around so I can't figure out what you'd like to see happen.
But regulation of people's behaviour is how we deal with these things. There should be consequences for abusing other people no matter what tools you use.
If you're talking about banning AI ("AI apologists"?) then were do you draw the line? People use cars to kidnap children and rape and murder them, should be ban cars? AI has positive and negative uses just like cars.
That's because you're trying to frame my statement on your terms, rather than catching your breath and thinking. Probably because you didn't understand my point but are really excited to tell me how wrong I am.
So let me restate it a different way. I'm gravely concerned with the destructive potential of AI and the exploitation in this article is just one of hundreds of negative consequences. Yet advocates of AI are shitting on any attempt to constrain it, and instead are tripping over themselves to monetize it, risks be damned.
In this particular case of article, the culprits where the boys in the same school. There has been bullying in schools forever. There are social norms, rules and laws to deal with it. In the worst case, you can sue for defamation.
For this particular article case, there is no need for new laws or rules. Just take the boys, give them a harsh warning, public humiliation and if it repeats then suspend from the school. Whether there is AI or no AI, it is irrelevant.
Whether this is relevant or not about the harms of social media is not related to the AI. If the article headline would be "social media gives new ways to bullying" it would be more to correct. However the article headline is about AI, and is thus a clickbait, and dishonest in its narrative.
Your emotive, ad-hominem and broad, unsubstantiated claims don't help your argument.
There are risks with AI as with anything, and there can be regulation of anything, it's just a question of what, which you don't go into a all.
To use your language: it seems you just want to shit on AI and people who want to use and develop it, and these victims are little more than a convenient sensation for you to push your pet peeve.
The author needs to go back to the library and read all those fashion magazines from the 90s, she can also go back further into the 60s but that would impact her view on "Summer of Love", not so innocent anymore.
On the other end, speaking of exploitation and unrealistic expectations, she could also consume "female literature" in order to get more understanding about this whole thing.
It's the human/society psyche. It's all imagination, there is no Mickey Mouse, it's just a bunch drawings scrolling really fast.
Now it's "AI" because it has to be, it's the "hot" topic, before was videogames, before it was TV, music ( remember the 80s? ), the author is just doing her job, good for her. But there is nothing new or substantial about this.
America is so schizophrenic about this. I remember the late-90s as being all "Think of the children! Strong family values!", and yet I don't at all see how the music video for "Baby one more time" (featuring a then-17 Britney Spears, and set in a high school) was considered acceptable according to the mores of the time.
83 comments
[ 1.8 ms ] story [ 164 ms ] threadPeople (women and men, both with their own methods) are the root of the problem.
Remember: tech is never the issue. People are.
People killed other people long before guns were a thing
At what point did ordinary people get easy access to weapons that have high capacity and rapid fire? That's when we see a rise in mass shootings. Today, ordinary people are getting access to tools which can quickly and easily make better fakes than experts with photoshop could do even 5 years ago.
https://en.wikipedia.org/wiki/List_of_school_shootings_in_th...
A well-known example is Swisserland, where lots of people have guns but somehow do not use them to kill people.
Another well-known example is France, where people use knives to cut other people's heads.
And are you actually comparing a trained militia to randos buying military hardware without any accountability or oversight? Because generative AI isn't only accessible to trained and licensed practitioners, is it?
The US is not a good representation of the world.
"Trained militia", you mean every random citizens and its family (the guns are kept are home) ? Then yes, this is it
The deadliest school attack in US history was the Bath bombing, which did involve a gun, but most victims were not from it.
The common factor is mental health, not firearms.
The common factor is easy access to firearms. Every country has mentally ill people, every country has embittered people who are mad at the world, but only the United States has removed the restrictions on gun access which allow those people to engage in mass shootings. That started in the late 1970s and the contrast with our peers is staggering: Australia has many cultural similarities but they actually value life and tightened their gun control laws after the Port Arthur shooting, while we’ve loosened ours and have hundreds of thousands of more deaths.
Fix't that for you. Anything is a weapon. The resulting high score is merely a matter of practice. No exceptions.
It makes good clickbait headline though.
Also there are existing laws for protecting against defamation.
Perhaps not, but ease of use really can be transformative on how something affects society. Especially since someone who has to spend years learning how to do something is less likely to do stupid shit with it than someone who just has to click a few buttons.
Will you claim that, therefore, NO software can possibly have any societal impact? Because, after all, it’s already “possible” to do what that software does?
If anything, I would argue what increases devious interest in this topic is the media talking about it as if it was new. Thats a quick way to get a lot of people "involved" on both sides of the debate.
Yes, I had to do about 5 minutes research on the internet to figure it out.
This is the version on the App Store
https://apps.apple.com/us/app/promptai-girl-generator/id6469...
Because of Apple’s rules, it won’t let you “unclothe” pictures. But the website will.
As a 50 year old married dude who never delved into that side of the world (not claiming I’ve never looked at porn though), I didn’t know how convincing it could do it for adults with shadows, skin tones, etc.
https://www.adobe.com/mena_en/creativecloud/photography/disc...
You can do this for free from your phone on at least one website that I’m not going to keep repeating.
I had never heard of this world until seeing posts about it on HN and it was just today that I delved into the bowels of the internet to see how well it worked so I could talk intelligently about it.
Just use the existing laws and social norms to punish badly behaving teenagers, no need to invent new AI laws for that.
It was never as easy or happening with the scale or as easy to make it look realistic.
Doing that in Photoshop was not something the average middle schooler could do realistically - let alone ok their phone, for free
I used to do Photoshop image manipulation for work. Performing a convincing face swap with the tools mentioned in your article was difficult unless the light sources, face angles, and focal lengths of the cameras that took the photos happened to match very closely. You could paint in your own lighting if you were very skilled, but it's the sort of thing that took years of practice to get good at, and it still wasn't very convincing most of the time.
When you have a million images to pick from, getting a reasonable match is a lot easier.
compare the same thing
I can't stand it when a boomer's version of a same service says "contact for quote", and the sales agent acts confused and perplexed about the term "self service" when you ask if such a thing is on the roadmap
(this wasn't about criminal websites, just self service e-commerce)
This is both true and a terrible reason to be okay with bad things proliferating. And it’s especially suspicious when this argument is leveled by people who generally aren’t going to be victimized by said bad things.
A nude photo is information. If you can turn it from implied authenticity to implied inauthenticity then the harm is reduced. The argument here is not about being used to things.
Sure, if you can do that. Of course the question is whether/when that’ll happen and how many young people have to be victimized first in order to get there — and for what benefit?
Not sure what's "non-factual" about it?
Maybe we’re entering a world where instead of saying suck a dick, you send someone a gif of them sucking off their dad, but as far as actual consequences go I think it’s pretty much immediately discreditable. Plausible deniability is all people ever really need.
Ah yes, middle school/high school bullies: famously disempowered by things like plausible deniability. The suffering caused by bullying and harassment: powerless in the light of plausible deniability.
AFAICT the people who would be close to this issue (e.g. people who pay attention to online harassment, teen mental health, and child pornography) come to a different prediction as far as impact. I’m far more inclined to believe them than people leveling an argument that, as I said, can just as readily defend murder, car thefts, and medical malpractice. All the suffering is just in people’s heads and all of it can be eliminated by proliferating the behavior to sufficient degree.
By this logic, you need to worry about me using AI to seduce your mom. Because NVIDIA spent a TRILLION DOLLARS!
> Ah yes, middle school/high school bullies: famously disempowered by things like plausible deniability. The suffering caused by bullying and harassment: powerless in the light of plausible deniability.
If bullies can push things as truth regardless of their merit then false information doesn’t matter.
Companies: We're raising billions of dollars to make it cheaper, faster, and easier to produce extremely believable fake media. Our success in this endeavor is verifiable by opening one of thousands of new image generator tools and trying it yourself for 30 seconds.
News reports: People are using tools that make it cheaper, faster, and easier to produce extremely believable nonconsensual sexual imagery of children (e.g. https://www.nbcnews.com/tech/misinformation/beverly-vista-hi...)
You: This is just like Photoshop. Not sure what the relevance is.
Is that right?
Worse, some of the options force criminal investigations. If they fake a video with an adult, now that kid has to deal with the police investigating a very sensitive topic. If they pick a teacher or family member, etc. that also means those people might have their jobs jeopardized and it could be very hard to disprove if the fakes don’t have obvious errors.
Wouldn't it make it less of a problem? At some point people will just assume all of the photos are fake.
But regulation of people's behaviour is how we deal with these things. There should be consequences for abusing other people no matter what tools you use.
If you're talking about banning AI ("AI apologists"?) then were do you draw the line? People use cars to kidnap children and rape and murder them, should be ban cars? AI has positive and negative uses just like cars.
So let me restate it a different way. I'm gravely concerned with the destructive potential of AI and the exploitation in this article is just one of hundreds of negative consequences. Yet advocates of AI are shitting on any attempt to constrain it, and instead are tripping over themselves to monetize it, risks be damned.
For this particular article case, there is no need for new laws or rules. Just take the boys, give them a harsh warning, public humiliation and if it repeats then suspend from the school. Whether there is AI or no AI, it is irrelevant.
Whether this is relevant or not about the harms of social media is not related to the AI. If the article headline would be "social media gives new ways to bullying" it would be more to correct. However the article headline is about AI, and is thus a clickbait, and dishonest in its narrative.
There are risks with AI as with anything, and there can be regulation of anything, it's just a question of what, which you don't go into a all.
To use your language: it seems you just want to shit on AI and people who want to use and develop it, and these victims are little more than a convenient sensation for you to push your pet peeve.
Neither do yours. Peace be upon you.
Are juvenile taunts really the best you can do?
On the other end, speaking of exploitation and unrealistic expectations, she could also consume "female literature" in order to get more understanding about this whole thing.
It's the human/society psyche. It's all imagination, there is no Mickey Mouse, it's just a bunch drawings scrolling really fast.
Now it's "AI" because it has to be, it's the "hot" topic, before was videogames, before it was TV, music ( remember the 80s? ), the author is just doing her job, good for her. But there is nothing new or substantial about this.
and people with those mores will pay for a fantasy that doesn't include those mores
I see what you did there.