AI Child Porn Will Probably Save Real Children

A defense of my most controversial tweet

…Of particular note is that this country, like Denmark and Japan, had a prolonged interval during which possession of child pornography was not illegal and, like those other countries, showed a significant decrease in the incidence of child sex abuse.
- National Library of Medicine

i regret deleting this tweet

This is by far my most controversial tweet, for reasons that are now obvious to me but were not at the time I tweeted it.

People somehow have interpreted this as me supporting child sexual assault, which is confusing to me but makes more sense if I model people as being LLMs that get triggered if you say too many negatively-flavored keywords too close together.

Most of the actual objections to my argument I’ve found summarized fairly well by this person:

“zero harm to kids” lmao every moron who advocates for AI child porn always fails to consider what the AI is going to be trained on to produce the material to begin with.

"but it will stop more children from being abused in the future!" only if you assume that none of the men who watch it will go on to abuse a child themselves, which is wishful retardation at best and malicious retardation at worst.

I don't expect Aella to comprehend the implications of this, because she's a degenerate porn addict herself, but it would follow the exact pattern as all other pornography.

people tend to want to "try out" the things they see in porn after establishing a dopamine connection with whatever they viewed that made them orgasm. that is why we are seeing an explosion in rectal and strangulation injuries in increasingly younger cohorts exposed to these acts through porn.

people also tend to want the "real thing," which is why AI porn in general hasn't taken off, and why tens of millions of men still pay for OnlyFans despite the fact that pornographic images and videos are readily available on the internet for free.

all AI child porn would do is 1) normalize the consumption of that content by making it widespread, 2) introduce new audiences to it previously deterred by legal consequences by advertising it as “victimless”, and 3) create an even more lucrative market for the REAL stuff that, like with all other forms of pornography, would need to push the limits to cater to an increasingly depraved audience.

we’d then be stuck in a spiral of increasing tolerance, like with all other pornography.

So let’s go down this list point by point, and I’ll add a few other points at the end for good measure.

AI must be trained on actual CSAM (child sexual assault material)

every moron who advocates for AI child porn always fails to consider what the AI is going to be trained on to produce the material to begin with.

…No, it doesn’t.

I’m like 99% sure that no human has ever taken a photo of a rabbit-waffle-chicken before. And yet:

The AI that generated this image was almost certainly not trained on a bunch of photos of rabbit-waffle-chickens. What the AI is doing is something like imagining what separate elements it’s familiar with, would look like if combined.

It’d be easier to train it with real images, but it’s theoretically possible to do so without - especially since AI image generation is rapidly getting better. It seems pretty obviously true that either now or shortly in the future, it is not beyond our technical means to have an AI generate content that it’s never seen before.

I’m assuming nobody who watches AI CP will abuse children

"but it will stop more children from being abused in the future!" only if you assume that none of the men who watch it will go on to abuse a child themselves, which is wishful retardation at best and malicious retardation at worst.

Of course! Watching CP is highly correlated with child abuse, and likely still would be even if we introduced AI CP.

Antidepressants probably help reduce rates of suicide. But people who use antidepressants are still more likely to commit suicide. It’d be ridiculous to say that nobody who takes antidepressants will ever commit suicide.

Methadone clinics reduce overdoses, but the presence of methadone clinics are correlated with crime and drug use. It’d be ridiculous to say nobody who uses a methadone clinic will ever use heroin.

Widespread porn access probably reduces sexual assault in populations, but people who watch more porn are more likely to sexually assault (high sex drive/aggression correlates with both). It’d be ridiculous to say that nobody who watches porn will ever commit sexual assault.

My theory is that AI CP functions as a dampener for people’s darker urges. Of course it will disproportionately attract people who have dark urges.

People Want To Try Out Things They See In Porn

people tend to want to "try out" the things they see in porn after establishing a dopamine connection with whatever they viewed that made them orgasm.

This is complicated and nobody has a very good grasp of how true this is.

For example, people don’t have this intuition about gay porn. If you’re not gay, and you tried to masturbate to gay porn with all your might, you’d still probably fail. If you’re gay and try to go to a pray-yourself-straight camp, you’d also probably fail. In a quick survey I ran to test this hypothesis, my followers overwhelmingly agreed that exposure to gay porn will do basically nothing.

They also agree exposure to taboo, rarer stuff also does nothing - when I asked about exposure to bestiality or poop, people have the same intuitions as they do about gay porn. No matter how many videos you watch of people coyly smearing feces over each other’s faces in time to suggestive saxophone tunes, you probably are not going to develop a hard-on.

But people had different intuitions about stuff that was more vanilla or popular.

n=150. Each dot pair is a fetish

The more taboo and rare a fetish, the more people thought they were immune. The more vanilla the fetish, the more people thought they were vulnerable.

(Also, as a side note, people overwhelmingly predicted that other people would be impacted by exposure to porn more than they themselves would be impacted)

Probably the closer a fetish is to something you’re already into, the more likely you are to be influenced by exposure to porn of it, at least according to the general intuitions of people who took my survey.

But in addition to this, there’s a logistical constraint: you probably aren’t going to be spending a meaningful amount of time masturbating to something you aren’t into, ever. Why would you spend time establishing a dopamine connection with what you’re viewing if you aren’t getting dopamine out of the masturbation to begin with? That defeats the entire purpose.

Likely this will only end up happening if it’s adjacent to something that does turn you on. For example, if you’re really into porn of sneaky sex where you have to be quiet, but it turns out almost all sneaky porn features naughty stepsiblings hiding from their parents, then you’ll functionally be regularly masturbating to incest porn, even if the thing you’re targeting is sneakiness.

But it has to be already kinda hot to you for some reason. Sneaky siblings - okay, maybe. Sneakiness cause they’re coal miners trying not to get caught by their harsh overseer probably won’t make you more aroused by coal mining. Coal mining is just too far away from your sexuality.

So - if you manage to spend a lot of time masturbating to something you’re not into, and if it’s already adjacent enough to your interests to have room to put a hook in, then you might start getting turned on by something new.

In this case, what you’d be doing is creating a shallow fetish.

My current theory of sexuality (which, take with a grain of salt, the data is nuanced and confusing and if anybody claims to know anything for sure about how sexuality works they’re probably full of shit) is that you can broadly categorize fetishes along a spectrum:

Shallow fetishes tend to have later onset and are conditioned. For example, if your first girlfriend always put her hair up in a ponytail when she gobbled your hog, then over time you might have grown to associate ponytails with hoggobbling, and now even though this was years ago, whenever you see a girl put her hair up in public you get a little half chub.

Deep fetishes tend to have earlier onset, and are something like innate. People with deep fetishes often report things like “I remember being six years old and seeing a cartoon when bugs bunny got spanked, and I felt funny and fixated on it.”

The age of onset fetish charts in my big dataset (NOT comprised of my followers) all look like variations on this:

This is more a spectrum and less binary, but the general principle still exists. Fetishes with early onset are more intense than fetishes with late onset.

For example, the fetishes of being attracted to the same or opposite gender are very innate, and tend to be noticeable quite early, in pre-sexual years, while the people who get simply bicurious usually notice it later.

(I should clarify here that I use the term ‘fetish’ intentionally broadly, as a term for any sexual interest. I consider ‘being straight’ to also be a fetish along these lines.)

Sometimes deep fetishes can look shallow when they’re heavily suppressed. Imagine a conservative gay man, raised in a repressed culture, who saw for the first time a gay c couple walking down the street. This might help prompt an awakening. Similarly, a woman with a deep bondage fetish might be more likely to realize and act upon this if she saw depictions of people getting tied up. I think in general, activating deep fetishes from porn is much rarer than conditioning shallow ones, mainly because it’s quite unusual to go a long time without any exposure to something that triggers a deep fetish. People into bondage felt funny when they first saw a villain get tied up during saturday morning breakfast cartoons; people into feet felt funny when they first crawled on the floor around their barefoot mother; people into bestiality felt funny when their neighbor’s dog licked their face.

So: it seems likely, to me, that you can develop shallow fetishes by conditioned masturbation when you’re repeatedly presented with sexual stimuli that’s adjacent to something you’re already into (typically a deeper fetish). It also seems likely, but less common, that you can activate deep fetishes by getting newly exposed to stimuli you’re hardwired to love.

So: Do I think exposure to AI-generated CP could cause people to experience pedophilic preference where there weren’t otherwise?

Yes, but I think this would probably be pretty rare.

To have it be conditioned, you’d have to:

1. Be at least slightly open to CSAM at all.

2. Regularly masturbate to content that was adjacent to CSAM

This seems pretty unlikely to happen. But even if it did, in this case, I think you’d likely end up with a shallow pedophilia fetish. Reminder that people with shallow fetishes tend to experience those fetishes less intensely - and are thus probably less likely to actually act on it.

“I love Mounds bars, it’s my favorite thing right, but there’s a limit. I can’t even eat a Mounds bar and do something else at the same time, that’s how much I love them. Like if I’m eating a Mounds bar I cant read the paper. I have to just sit there with it in my mouth, and go “Why is this so good? I love this so much.” Because they are delicious, and yet if somebody said to me “if you eat another Mounds bar you will go to jail and everyone will hate you,” I would stop eating them. Because they do taste delicious, but they don’t taste as good as a young boy does—and shouldn’t!—to a child molester.” - Louis CK

Probably most of the pedophiles who risk jail and extreme social ostracization to actually molest a child, are the ones who are really into it. Probably, it’s the deep pedos - the ones that ended up with their sexuality hardwired in the most unfortunate possible way. Deep pedos already know they’re into it - it’s hard to avoid exposure to any children ever. Thus, the most dangerous group is probably the least likely to be influenced by CSAM.

Of course I’m saying words like ‘probably’, ‘most’, ‘likely’. Nobody, not even me, knows this stuff for sure, and there’s always outliers. Sexuality and related behavior is extremely hard to predict. I am not claiming that it’s not possible to induce pedophilic behavior in someone who didn’t have any before.

My theory is that we can reduce pedophilic behavior more than we create it. If an intervention increases pedo behavior in 1% of pedos and decreases it in 99%, I think this is a huge amount of good and we would be terrible people to not do it.

Rectal and Strangulation Injuries Are On The Rise

that is why we are seeing an explosion in rectal and strangulation injuries in increasingly younger cohorts exposed to these acts through porn.

I’m not sure about ‘explosion’ - I couldn’t find any studies to support any meaningful rise of injuries. My guess is the writer is pulling their information from highly speculative sources. I think this doesn’t really matter to my argument regardless.

People Want The Real Thing

people also tend to want the "real thing," which is why AI porn in general hasn't taken off, and why tens of millions of men still pay for OnlyFans despite the fact that pornographic images and videos are readily available on the internet for free.

This is almost but not quite true.

AI porn hasn’t taken off yet because it’s bad. It will absolutely take off once it’s indistinguishable from real people. I know this because I’ve seen some onlyfans girls use the most egregious face and body morphing filters known to mankind, and the vast majority of men have no idea this is happening. The vast majority of horny men are much dumber than you’d think. Doing online sex work, my numbers consistently go up the dumber I act. You’d hope that going XOXO HORNY PUSSY WET FOR U BB would not in fact cause money to pour in, but it does. In the war of men’s penises vs the AI, I am betting on the side of AI.

I cannot overstate enough how much the thing men get out of Onlyfans is not realism, but rather the sensation of interaction and influence. They do not want you, they want some facsimile of you. People’s desire for genuineness often conflicts with their desire for convenience, safety, fantasy fulfillment.

It would normalize the consumption of CP content

all AI child porn would do is 1) normalize the consumption of that content by making it widespread,

I somewhat think this is the strongest part of the argument. I do think a lot of good is done by the immense social shame we heap upon offending pedophiles, and actually does deter a lot of assault. It doesn’t seem insane to argue that using AI CP to dilute real CSAM might reduce social shame, and we should handle threats to our greatest defense against child sexual assault very seriously.

I think this is unlikely though. Social shame around pedophilia has, if anything, increased, while pedophilic access to CSAM over past decades has also increased. Passing out clean needles in SF to help prevent illnesses in drug dealers did not, I think, normalize drug use; if anything I think it made people more upset about drug use.

So: It’s possible that allowing AI CP would normalize watching AI CP, but it’s also possible that it would have the opposite effect - making it even more taboo than it already is. We don’t really know. Ideally we would spend a lot of effort figuring this out!

It would introduce new audiences

2) introduce new audiences to it previously deterred by legal consequences by advertising it as “victimless”,

That is good. I think the new audience that AI CP would attract is an audience more likely to actually assault kids in real life, and that by attracting them to AI CP, we would probably redirect their energy and reduce the number of actual assaults.

It would create a market for real stuff

and 3) create an even more lucrative market for the REAL stuff that, like with all other forms of pornography, would need to push the limits to cater to an increasingly depraved audience.

How do CSAM markets work currently?

It seems to go something like this:

CSAM gets shared in groups or forums, which are hard to get to (typically you have to be recommended by someone or go through the dark web). In order to gain access to content, you often have to produce fresh content.

To gain access to these communities, an initiate must produce new CSAM.

This requirement ensures applicants are legitimate minor abusers and gives participants legal leverage over potential informants. - Humantraffickingfront

One requirement for entry to the club, apart from a recommendation from an existing member, was the expectation to supply 10,000 new or self-produced pornographic images of children. - Wikipedia, Operation Cathedral, about the Wonderland Club

Members were required to upload new material at least every 50 days to maintain their access and remain in good standing. - Wikipedia, Operation Delago, about Dreamland

In some online forums, children are forced to hold up signs with the name of the group or other identifying information to prove the images are fresh. A private section of the forum was available only to members who shared imagery of children they abused themselves. They were known as “producers.” - New York Times, about the Love Zone

In these communities, there is immense incentive to produce CSAM content yourself. This is the way you gain trust, acceptance, and access to more. In order to consume images, you have to go break the law by actually assaulting a child.

That’s the way the CSAM market is structured.

It’s like, if you wanted access to a gang, you had to go through an initiation ritual which required you to go shoot an innocent civilian, to prove that you could kill someone and also that you weren’t a cop.

I feel like I’m arguing - let’s give them innocent civilian test dummies. Please, for the love of god, swap out people with real lives with a dummy.

Call me crazy, but imagine that you’re a pedo who wants to find CSAM. You have two options to attain it: one is a quest where you have to go find and assault a new child and force them to hold a photo with the right words on it, and the other is to click a few buttons and submit an automatically generated photo. Almost certainly you would pick the latter.

…Really, if you could generate as many realistic-looking AI photos as you wanted, you probably wouldn’t be trying to gain access to the CSAM groups at all.

But okay - back to the original point. Would it create a more lucrative market for the ‘real’ stuff, that, like “other forms of pornography” would need to cater to an “increasingly depraved audience”?

AI, at its current early stage, is still convincing enough to override the amount that people care about it being real.

Generally, fake stuff seems to kill demand for real stuff, not enhance it.

1. As an escort, porn is a competitor, not an ally. If all porn suddenly vanished tomorrow, I’d expect my bookings to increase.

2. Cultured pearls, engineered wood, lab-grown diamonds, various scents, petroleum jelly, acrylic, etc. - these are all synthetic, cheaply made stuff that damaged the market for the real thing. Introducing polyester did not increase demand for linen.

3. AI art is already disrupting jobs for artists. It is not increasing demand for human artists.

   

If you want something, and you typically have to go through a whole lot of effort and risk to get it, or you could get a knockoff product for 1% of the effort, then most people are going to switch from the original thing to the knockoff product.

Porn and sexuality is not unusual in this aspect.

For example: arrests for rape dropped, despite (or perhaps because of) an explosive increase in access to internet porn.

In general, increase in porn use seems to correspond with lowered rates of sexual assault.

“Using state-level panel data on the rise of the internet, I find that internet access appears to be a substitute for rape. Specifically, the results suggest that a 10 percentage point increase in internet access is associated with a decline in reported rape victimization of around 7.3%.” - Todd D Kendall

I don’t mean to imply all the research is uniform or unassailable; many studies find no effect of porn on violence, and I read one study that found an increase. But we clearly don’t live in a world where flooding the planet’s male population with infinite pornography and access to violent porn has caused a surge in men doing more rape.

In fact, the watersupply meme is more like “men who watch too much porn will have trouble getting boners with real women” or “porn makes men stay home and not try to get girlfriends.” It’s weird that we feel so confident about this, but believe that the effect switches depending on the type of porn you watch.

We don’t have great data on CSAM, but there was a period of time where a few countries legalized porn and had no laws against CSAM, and sexual offenses against children dropped.

Similar to our findings, in Denmark and West Germany the most dramatic categories of sex crime to show a decrease were rapes and other sex crimes against and by juveniles. Between 1972 and 1980, the total number of sex crimes known to the police in the Federal Republic of Germany decreased by 11%; during the same period the total number of all crimes reported increased by 50%. Sex offenses against minors (those under 14 years of age) had a simi- 12 M. DIAMOND and A. UCHIYAMA larly slight decrease of about 10% during this period. For those victims under 6 years of age, however, the numbers decreased from 1,421 cases in 1972 to 579 in 1980, a decrease of more than 50% (Kutchinsky, 1985b, p. 319). Other researchers have found similar results. In Denmark, homosexual child molestation decreased more than 50% from 74 cases in 1966 to 20 cases in 1969 (Ben-Veniste, 1971, p. 254). - Pornography, Rape, and Sex Crimes in Japan

But even if easy porn access did turn you more degenerate, maybe it’s still net good? I’d rather a world full of degenerates masturbating to weirder porn than they would have otherwise, than a world where they’re porn-free and molesting kids.

I Would Change My Mind If

We don’t actually have good data for most of this. My argument lies mostly on stuff like reading police reports about CSAM sharing networks and patchy studies about how porn impacts behavior and my own studies on fetish preferences and my interviews with people with rare fetishes. Most studies done in this area have an unusually high number of flaws due to either difficulty getting data or being motivated from outrage. The field is full of people making very strong claims with not a lot of evidence to build the claim on.

Likewise, I am not certain. My entire argument here rests on a lot of strong inferences made from existing evidence, but is not foolproof. My goal is to actually save children from sexual assault. If some of my assumptions turn out to be wrong, I would change my mind. If, for example, we somehow did a robust study that showed giving pedophiles access to AI CP increased their risk of offending, I will update this blog post here with a link at the top saying “My guess was wrong, I’m changing my mind.”

It just seems that people - somewhat understandably - have an immense disgust response to the thought of doing anything other than woodchippers for those who have sexual urges towards children. I’m not here to argue your disgust response is wrong, only that it causes us to use strategies that are ineffective. Blinded by our disgust we are failing to figure out effective strategies for actually protecting children.