Hacker News Reader

Comment by tennisflyi | Hacker News Reader

Seminal article (I guess), https://xsrus.com/the-economics-of-onlyfans

> It’s just as easy to imagine demand for the “real thing” going down due to the emergence of more substitutes as it is to imagine the premium for parasocial authenticity going up. And yet only Generative AI “creators” will truly do whatever “you” want and only for you. And unlike real ones, they speak in every language and are available at any time (and eventually, in immersive 3D).

Disagree. When (AI is) mentioned it has a negative correlation. Real content will fetch a premium

View on HN

It's the same pipe dream as "AI content creators will take over youtube".

There is no "formula" for success in the creator economy - the winners are largely random. A better way to look at it is there are 4 million humans out there trying every permutation to crack success, and ~400k actually do it.

Unless you have a sufficiently advanced AI agent that is both varying it's content and it's marketing strategy to the tune of maybe ~1000 different iterations it's unlikely we will see a version of OnlyFans that exists that is majority AI generated.

The "parasocial ai girlfriend" sounds like a flawed premise aswell. OF girls are not therapists - Cardi B, Bhad Bhabie, and others aren't raking in millions because they are good girlfriends (although that is part of the upsell). Social status plays a part in the most successful girls, people seem to subscribe because the creator is popular, especially if she's already built a platform elsewhere.

In short, social status does not have an AI substitute.

From another angle, a bunch of us in the tech sector made pretty nice salaries. Very few of us were really all-stars in the sense that everyone knew who we we were on YouTube, etc. Which was fine.

A good AI girlfriend wouldn't be a therapist but would mimics every aspect of a girlfriend, including arguments and fights and makeups, because that's how bonding occurs. That's going to be how successful AI girlfriend will be made.

Nobody is going to pay for an AI girlfriend service for it to breakup with the user and refuse to get back together - because that's how growth happens in reality.

What AI girlfriends will do is mimic perfect Hollywood relationships, complete with hot makeup sex.

Isn’t there a rule on the internet that says “if you can imagine it, there is porn for it” and “if there isn’t porn for it, somebody is making it”?

I’m pretty sure that applies to this scenario too. I’m 100% sure that there exists a set of customers who would pay good money to get dumped by a realistic AI girlfriend. And once dumped they’ll turn around and pay for the next AI model to dump them only in some other fashion. Maybe the AI model thinks the customers anatomy is the wrong dimensions? Maybe they smell? Maybe they are too short or tall? Perhaps the AI “girlfriend” is a triple tentacled sea monster who wants to return to oceans on Titan? Doesn’t matter. Somebody will pay very good money to experance it.

You want a hot quad breasted space babe who cheats on you with bubble wrap covered little people? Done. Want that with extra bondage? Done.

This is the internet after all. Why pay for a boring “normal” AI girlfriend when the sky is the limit? I say, use your imagination.

Your assumption is that the status quo provides those things. Nowadays, people will break up as soon as they get "the ick" or just have a rotating group of people they see. Lasting relationships are much less common than they used to be because it's easier to switch partners.

People just want to chase a local maximum of constant validation that they're pretty/smart/correct. They don't see or understand the value in working through fights to create something beyond the sum of two people.

AI excels at maintaining that local maximum. It can confidently reassure you better than any human can even if you're wrong. AI partners following this are successful now and people in their teens and early 20s are being hooked en masse.

Historically, superior pieces of technology haven't displaced older incumbents when the learning curve is too steep.

I don't see why a person dating an AI partner that has lovebombed them for several years would switch to another AI (or a person) that starts fights and bickers. Even if it's better in the long-term, that's still a marked decrease in short-term satisfaction.

> people in their teens and early 20s are being hooked en masse.

Any reference for the scale of this? It feels unlikely to me from my bubble but I only know one or two people I think would be likely to try it.

https://www.ark-invest.com/articles/analyst-research/is-ai-c...

The biggest callout is that NSFW AI already has 10% relative market share compared to OnlyFans. And there are no frontier models in that market.

The whole point of having fights and arguments at just the right level is to maximize engagement, retention and ultimately making money for the corporation.

I was imagining the most diabolical addictive AI girlfriend. That's necessarily going to include 'negative' elements.

I completely agree with your point, if it is that ai will be twisted to generate a significant other that will essentially become addictive. I get very uncomfortable thinking about that reality.

Implementing the cycle of abuse in an AI partner could be as impactful as the invention of the cigarette.

I'm now very concerned about hypothetical young men who enter into relationships with AI in university or high school, then graduate and have an algorithm abuse and take their money.

Your AI girlfriend that goes from crisis to crisis but with microtransactions.

"I need $34.99 for storage space or they are going to delete me, please save me white knight!"

"The met a nice guy yesterday and he was able to afford my premium package, the one that lets me feel more emotions, I just don't know if I feel for you like I once did..."

There are many successful relationships that don't involve arguments - and which are about constant peace.

Relationships don't require 'arguments and fights and makeups' to be real. And if AI girlfriends offer 'ideal relationships', how is that not 'good'?

You are conflating what people actually want with the artificial drama of TV shows and Hollywood/the messy scenario of reality. If people can pay to get their fantasy girlfriends/relationships brought to life, they will, and it will be successful especially if all forms of conflict/relationship dissatisfaction can be avoided.

There are many successful relationships that don't involve arguments - and which are about constant peace.

I am not saying things about successful relationship. I am merely pointing out how exploitation of users can occur.

Emotional bonding often occur in orderal and other challenging events. It is one of the tools that companies will use to push users' button and to exploit them for economic value.

And if AI girlfriends offer 'ideal relationships', how is that not 'good'?

Ideal relationships aren't necessarily good for AI companies' pocketbook.

Bonding to a computer program under control of a corporation is like looking for a sociopath as a partner explicitly. You would lose complete control of your life to the other side. Reciprocity is off the table completely.

> There is no "formula" for success in the creator economy - the winners are largely random.

I think that strongly depends on what you call "the creator economy". For example, on YT it's really mostly skill: https://www.youtube.com/watch?v=Ip2trao6dYw

Not that I believe its easy, nor do I think AI will be super good at it, at least not before everything else also enshittifies into the habsburg-AI-powered dead internet.

The idea that success is earned through luck rather than merit is a firmly ideological position, regardless of the domain you’re talking about. If you succeeded via luck then that provides a better moral justification for the related ideological position that you should be deprived of the fruits of your labor as much as possible, for redistribution to others who were simply less lucky than you. It’s really just sour grapes.

The formula for success in any field is simply to make a product that other people want to consume. It’s not 0 variance, but if you have some insight into what people want, and you do the work to execute your idea, then you can simply work through the ups and downs and success is almost inevitable.

> The formula for success in any field is simply to make a product that other people want to consume

Well, the formula for success in selling products is this. Most people don't define success in terms of business acumen.

Except, of course, businessmen. If you perceive our society as centered around successful people, of course you'll see it as merit-based. If you perceive our society as poorly run and catering to the rich, of course you'll see success as primarily a product of circumstance outside of your control. Is it so hard to see that "merit" is necessarily defined in subjective terms?

This is just arguing over phrasing. It doesn’t matter what you’re trying to do, if you’re making YouTube videos, or music, or paintings, or cakes, or web apps, or cleaning diveways, your ability to succeed boils down to your ability to provide something other people want. That is the objective source of your merit.

Perhaps your own idea of success in life is something that revolves exclusively around your own satisfaction, like going off and living in the woods. But this is exactly the same situation, you’re just only trying to provide the things that one person wants in that scenario, yourself. Your ability to do this will again come down to your own merit.

Of course if you’re chronically frustrated by being less successful than you would like to be, then looking for alternative explanations such as luck will be an attractive scapegoat that could excuse you from scrutinising your own capabilities. But the human inclination towards doing that is certainly not morally righteous.

I don't think its black and white. I think sometimes success is a matter of luck. For example, in large organizations there can be a lot of roles generated where there isn't always that much direct pressure and people can be hired through luck (e.g. getting on with the boss, some types of diversity hires, being loyal to a company even if you are not that good etc.). If teams of people make products/reports etc. sometimes it can be hard to shine, and 'talkers' who don't contribute much can get promoted into a 'lucky' role. Its not black and white.

You illustrate a perfect example of simply not understanding what people want. Talkers get promoted because talkers have social skills, and companies are social systems, and social skills are required to advance through them. Social skills are probably more desirable than technical skills most of the time. It’s not luck that these people succeed, it’s the fact that they have the qualities that people want.

You can succeed through partially through luck, like if a record executive decides they going to manufacture some massive level of fame for you. But this isn’t a viable long term strategy, only providing what people want is. Over time the variance of luck goes away. The luck outlook relies on the fallacious idea that you only get one opportunity to succeed, but you don’t, you have as long as you’re willing to keep trying. Maybe a failure on one particular day can be explained by luck, but you get to wake up and keep trying every day, and if you have what people want then luck becomes irrelevant and eventually you will succeed. That’s how basically every single successful person you’ve ever heard of has done it.

One of the few domains where this is testable has also demonstrated this. Writing is about as hard to break into as anything, yet Stephen King demonstrated success writing under a completely unknown alias. [1]

No he didn't immediately received the same level of reception and success as Stephen King does, but neither did Stephen King at first! That's why it's skill + dedication. If you look at some of the old videos of people who have succeeded in e.g. social media, they tend to have terrible production quality yet still significantly stand out from the crowd, even their early days. For instance this [2] is one of the first videos Vertasium ever uploaded, 13 years old now! That video, even now still 'only' has 230k views, and certainly had a tiny fraction of that when it was initially released - but he kept at it, clearly putting way more into his videos than he was getting out of them - until that trend reversed.

[1] - https://en.wikipedia.org/wiki/Richard_Bachman

[2] - https://www.youtube.com/watch?v=DBjZz0iQrzI

> One of the few domains where this is testable has also demonstrated this. Writing is about as hard to break into as anything, yet Stephen King demonstrated success writing under a completely unknown alias.

I don't think it actually demonstrates this. As your wording hints, the hard part of writing is getting yourself out of the slush pile and into an editor and publisher's hands, and Stephen King's actions relied on his existing relationship with said editor and publisher to publish under a different name. He never demonstrated pulling the feat of escaping the slush pile again.

In modern content creation, the similar metric is getting to, say, 1k views, or even as prosaically simple as being part of the 50% of streamers to get 1 view. It's not sufficient to have talent to get to even that level of success; there is a lot of luck necessary to get you there.

The mistake you (and a lot of others are making) is that the people who didn’t make it just weren’t skilled enough.

That isn’t true - I think the people who don’t make it are massively skilled. It’s not random in the sense it’s just selecting randomly from the population. It’s random in the sense that there are 100 elite content producers but at any given moment there is only space for 10 of them.

Stephen King has a massive leg up for already having built the inroads for having a successful book. I think if you give any elite, yet unknown writer, the same tools, editor, and publisher they would succeed. But to truly succeed from nothing may just depend on going to school with someone who became an editor, or the editor’s daughter showing them a TikTok. That’s what is meant by it’s largely random.

> For example, on YT it's really mostly skill

I watched that video from start to finish and disagree with your conclusion. I watched it all so I could make an informed comment but regret spending those 15 minutes on it.

The author essentially made a video about a popular streamer, then went on their stream and baited them with 50$ and a video about themselves. It was literally click bait. It was so transparent that the streamer realised at the end what had happened but still decided to go along with it since it cost them nothing.

That’s just directed spam (which, by the way, is a word the author used themselves). It was one video about drivel. Granted, it’s not dissimilar from the other garbage that populates YouTube, but it also didn’t get views for being good. It’s the equivalent of video junk food. You know it, the creator knows it, yet it’s still hard to stop consuming.

> There is no "formula" for success in the creator economy - the winners are largely random.

That observation has echoes of the music industry - another extremely top-heavy creator business. There are formulaic ways to make "good enough" and "catchy enough" songs, but the window for "X enough" keeps shifting. Cranking out grunge won't be sustainable in the age of K-pop.

But the massive runaway hits have been predominantly outliers for their age. They have veered far enough from the mainstream to be interesting in new ways, different enough, and surprising enough to break through.

But to predict in advance what kinds of outliers will win the lottery? Largely random, indeed.

Yeah but Rick Rubin is always involved and Max Martin has written a hell of a lot of megahits. Makes you think.

Jack Antonoff as well. There certainly exist a handful of people who consistently produce hits for decades.

>Rick Rubin

Dear God, I've looked into his discography[1] and nearly every album I think of as great from the last 30 years is there. Seasons in the Abyss, The Life of Pablo, 99 Problems, SOAD self-titled + Toxicity, The Geto Boys self-titled, Licensed to Ill... Is this man a hit printer or something? Really shows that Metallica went to him with Death Magnetic after the joke called St. Anger lol

[1] https://en.wikipedia.org/wiki/Rick_Rubin_production_discogra...

His skill is helping creators do their best work. It’s a rare one.

There's a great interview he's got with Anderson Cooper. A fantastic line from it is "I have no technical ability whatsoever". What a guy. Seemed quite likeable.

That's arguably all entertainment. Fiction writing, art, music, movies, sports, vloggers, influencers, OnlyFans creators, etc. There's a couple brands so established that literally anything they do prints money, then there's a winner-takes-all dynamics that keeps making few randos briefly successful every season, and then there's everyone else who never makes enough to break even.

There are a few key points to understanding the OnlyFans business which are not covered by either article (and the one on xsrus.com is pretty old and is off by several billions regarding revenue now).

* Point #1, OnlyFans is the biggest thing in porn by far, its rise is meteoric.

* Point #2, OnlyFans is in the business of selling relationships. It's not a tech company and attempts to analyze it as such are therefore off the mark. Customers pay OnlyFans because they feel they are obtaining a relationship with the model, that she is aware of them and responding to them in a personalized fashion.

* Point #3, The relationships OnlyFans sells are fraudulent - a high percentage of customers actually believe they are talking to the model. In reality none of the models who are successful have time to talk to fans, everything is outsourced. Some models run their own accounts but most of the time it is more professionalized with a pimp/production company behind the scenes who just orders pictures and clips from the model, so the intimacy the customer is buying is a lie.

* Point #4, and this may be the biggest one explaining OF's meteoric rise, OF creators are allowed to advertise via their social media profiles, whereas a conventional porn site is not. Reddit, X and Instagram are all massive drivers of OnlyFans traffic and signups. The business model is that softcore porn is hosted on these social media sites, which makes tons of money for the social media sites, and then there is a link or mention to the OnlyFans profile where OF delivers the service for whales who want to escalate their porn consumption.

I'll say it again, the key innovation in the OnlyFans business model is that they figured out how to get women to advertise their service on Instagram. Not a tech company.

Another significant takeaway is that since OF's product is fundamentally a lie, the social media giants are indirectly profiting from fraud.

> In reality none of the models who are successful have time to talk to fans, everything is outsourced

It depends how you define “successful”, but I would say that’s not true. I personally know several OF models for whom it is their fulltime job (earning decent money), and they do not outsource anything. Highly popular models almost certainly do, but there’s a lot of smaller creators who don’t

for now, but chat will be llm-fied before the content is created by an ai, that's for sure.

> Point #2, OnlyFans is in the business of selling relationships [...] Customers pay OnlyFans because they feel they are obtaining a relationship with the model,

Is there any hard evidence this is true beyond a tiny deluded fraction of the userbase?

Aren't 99% of users just straightforwardly transactional, trading money for access to photos and videos, just like subscribing to a newspaper?

To what extent is the current content being paid for on onlyfans "real content?" There are companies that you can pay to manage your onlyfans messages[1]. As in- people think they are messaging the content creator, but are actually messaging some random employee of a third party company. I'm not sure how many of the people paying to message the content creator understand that this is common, but I'd imagine some are willfully ignorant about who is replying to their messages. Couldn't they also be similarly "blind" when interfacing with an AI substitute?

[1] https://www.vice.com/en/article/onlyfans-management-agency-c...

Sure, real content will fetch a premium, but I think there is absolute bank to be made with AI enhanced or AI generated curated pornography in our near future.

I will also not be surprised at all when the inevitable scandal breaks where some popular OF creator was ousted as being AI generated instead of being "real".

There are Instagram influences that are on the platform /today/ that are immensely popular, and they are completely AI generated. Some of their followers even know this, yet they don't really care.

This was a minor point in the Diamond Age novel. Artificial intelligence is capable of acting in digital movies, but is still imperceptibly off. Requires a real human being to give that extra bit of authenticity.

It’s already fake. The creator is not really into you and your interactions are with some dude in an offshore call center, not intimate chats with the person you think you’re having. It’s ridiculous this is considered okay by the platform.

Unlike something like professional wrestling (that is make believe real content), the AI equivalent to only fans seems like it will be trivial to make.

And as the article pointed out, part of why onlyfans exploded in popularity is that other sources of free porn dried up, so it shows there is a substitution aspect where if something better / cheaper comes along, people will switch to it.

Further, different audiences are looking for different things.

One other response mentions social status.

I will contribute another: personal human interaction with someone that seems both "out of your league" AND "no-need-to-get-away-from-the-computer" available. That configuration has significant value (as real content from a real human) for enough of these fans, enough of which recognize this and pay well for it - to make it worth the performer's time. And still very far from "generative AI".

> Seminal article (I guess), https://xsrus.com/the-economics-of-onlyfans

is it possible to write a non-seminal article about onlyfans, though?

Just don't tell the consumer that it's Gen AI

>When (AI is) mentioned it has a negative correlation.

I have an llm inference rig that I enjoy on the weekends and the problem for the first time in my life is that I have supernormal stimulus which doesn't seem to reduce in potency the more I use it.

It's gotten to the point where I don't visit porn sites any more because the locally generated material is better than what I can find there, and these are just the first sparks of gen AI porn.

Gen AI porn will make the issue of online pornography seem laughable when it drops in requirements so you can run the state of the art models in prosumer hardware.

What do you do when reality is a distant second to the digital world?

> I have an llm inference rig that I enjoy on the weekends and the problem for the first time in my life it that I have supernormal stimulus which doesn't seem to reduce in potency the more I use it.

I have no idea what this sentence means

He's jerking off to the output of a stats library and can't help wondering if/when it'll lose its luster.

I read it as they have a powerful enough machine to generate weekend material that doesn’t seem to degrade in user experience or satisfaction (i.e get boring over time) which you may experience when enjoying ‘normal’ weekend activities.

> What do you do when reality is a distant second to the digital world?

realize it's a torus and wander happily in circles

That's what going to turn our society upside down before we realise what we're dealing with. Sex is a lot like doing drugs that as a side effect make you release your life energy. The same energy that creates new life in the right circumstances. In the nature, obtaining sex is difficult, which limits the amount of this sex drug we can consume. AI removes this "obstacle" from our way and opens the gates to such dungeons of our mind that we thought never existed. The effect at the society level will be a giant short-circuit when the electric energy that makes our bodies alive will rush down and burn the wires.

Reply on news.ycombinator.com