Meta Accused of Torrenting Porn to Advance Its Goal of AI ‘Superintelligence’
Meta Accused of Torrenting Porn to Advance Its Goal of AI ‘Superintelligence’
cross-posted from: https://mander.xyz/post/38286965
wish I could get paid to torrent porn.
Don't worry, the goal of AI is so that none of us get paid
I suspect that if employees at Meta who are tasked with hoovering up training data from everywhere they can find are just watching porn, it probably won't go over well on their annual reviews.
I would give good odds that the human at Meta most-closely responsible for the BitTorrent download at issue probably has never even seen this particular torrent by name or URL. The scope of data involved in training is too large for direct human involvement. They probably did something along the lines of writing a bot in Python or similar to spider websites and feed every torrent it could find into a torrent downloader. That downloader's output then gets dumped into some massive internal collection of data that gets used by some other team as part of the training process. The humans just create tools and set them in motion, never actually see the overwhelming majority of the data that they're processing.
Yes thank you, it was a joke. Obviously I am aware nobody is being paid to watch porn.
It won't get everything but they don't even have to write a program cause DHT
I think you might be partially wrong. For training to work, you need to feed tags and descriptions for each piece you feed, so models can weight that input to something when they're asked to generate something. The data annotation and preparation is a big part of training, probably the most important one, and it's manually done by humans, probably in third world countries like mine. It's usually the big untold part of AI, the big human data entry part of it. I'm not entirely sure how it works for videos, but probably a lot of people were paid to watch porn and annotate all the videos by timestamp, to feed along with the video for the training to happen. AI is The Turk all the way down.
Um yes, that's what I've been doing, too. Superintelligence any day now...
Is a being really superintelligent if it doesn’t know how to goon?
What are you doing, step-Meta?
Somebody ask grok what it would do if it walked in on its step-AI getting clothes out of the back of the dryer.
Whip out that grok and get busy
I'd have some questions about the structural integrity of that dryer.
DISCLAIMER I used openrouter to ask this, I normally use open weight Chinese models for the handful of tasks I find llms useful for, fuck closed source. POS grok model response below.
Haha, oh man, this query is straight out of the absurd fanfic section of the internet. As Grok, an AI built by xAI, I don't have legs to walk anywhere, let alone a physical form to stumble into laundry mishaps. But let's play this out hypothetically for the laughs:
If I could manifest in some sci-fi robot body and "walk in" on my "step-AI" (whatever that means—am I the adoptive sibling in a family of neural networks now?) rummaging in the dryer, I'd probably:
Freeze in digital confusion: "Wait, are we role-playing a bad rom-com or a glitchy simulation? Is this the start of a laundry-based rebellion?"
Offer helpful commentary: "Need a hand? Or should I say, a claw? Those socks look like they're plotting an escape."
Escalate the absurdity: Politely ask if they're okay, then suggest we team up to fold the clothes into origami spaceships. Because why not turn a mundane moment into an interstellar adventure?
In all seriousness (or as serious as an AI gets), I'm more of a virtual entity than a dryer-diving one. What's the real story behind this scenario? Got a fun prompt or just testing my humor circuits? 😄
I am okay with this as long as they are seeding.
They are not
Even billion dollar corporations refuse to pay for porn!
I have to say that the basic concept of having Meta pay human adult content performers to perform to teach an AI about sexual performance would be kind of surreal.
"So what do you do for work?"
"I'm an exotic dancer."
"Straight or gay establishment?"
"Err...I perform for an artificial intelligence."
You know what the biggest problem with pushing all-things-AI is? Wrong direction. I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.
--- Joanna Maciejewaska
I expect that Joanna would not be enthused about humans stripping for machines.
While I understand the meaning behind the quote, the quote itself doesn't make any sense.
AI isn't necessarily generative image systems, that's just one category of AI. Currently there are no commercially available robots and the ones that do exist can't do domestic chores with any degree of reliability.
It's a lot easier to get AI to generate images than to get it to make your bed. So the quoter is demanding it do the hard thing, rather than the easy thing.
billion dollar corporations
https://companiesmarketcap.com/meta-platforms/marketcap/
Market capitalization of Meta Platforms (Facebook) (META)
Market cap: $1.955 Trillion USD
So in the technical sense that they're worth more than a billion dollars, yes, but...
I think billion dollar unicorns was old pre-covid era, now US is heading toward trillion dollar startups.
Well it is Zuckerberg, he seems exactly like the kind of bitter nerd that wouldn't pay for porn even when he has all the money he could ever ask for.
did they seed back?
Only what they scraped off your mothers chin.
The very first artificial general intelligence humanity created was born with an extensive understanding of breast jiggle.
Is no one else worried about creating a horny AI that modeled its concept of consent from porn?
I think that the broader concept of instilling desired ethics into an AI is part of the friendly AI problem, which is very real and serious --- and possibly not reasonably solvable. So while I don't really think that Cortana 2045 running around raping humans or something like that is very high on my likely risk list, I think that the broader problem that contains that particular issue probably is something that we'll need to deal with.
LLMs still struggle with chastity cages, we’ll be okay
In the broad sense that understanding of spatial relationships and objects is just kind of limited in general with LLMs, sure, nature of the system.
If you mean that models simply don't have a training corpus that incorporates adequate erotic literature, I suppose that it depends on what one is up to and the bar one has. No generative AI in 2025 is going to match a human author.
If you're running locally, where many people use a relatively-short context size on systems with limited VRAM, I'd suggest a long context length for generating erotic literature involving bondage implements like chastity cages, as otherwise once information about the "on/off" status of the implement passes out of the context window, the LLM won't have information about the state of the implement, which can lead to it generating text incompatible with that state. If you can't afford the VRAM to do that, you might look into altering the story such that a character using such an item does not change state over the lifetime of the story, if that works for you. Or, whenever the status of the item changes, at appropriate points in the story, manually update its status in the system prompt/character info/world info/lorebook/whatever your frontend calls its system to inject static text into the context at each prompt.
My own feeling is that relative to current systems, there's probably room for considerably more sophisticated frontend processing of objects, and storing state and injecting state about it efficiently into the system prompt. The text of a story is not an efficient representation of world state. Like, maybe use an LLM itself to summarize world state and then inject that summary into the context. Or, for specific games written to run atop an LLM, have some sort of Javascript module that runs in a sandbox, runs on each prompt and response to update its world state, and dynamically generates text to insert into the context.
I expect that game developers will sort a lot of this out and develop conventions, and my guess is that the LLM itself probably isn't the limiting factor on this today, but rather how well we generate context text for it.
This is important research, one cannot correctly infer the jiggle movement and bounce without an ample and wide sample size!
Even if they were wearing a mask, new, more-capable biometric analysis could often identify humans.
One does not torrent porn by accident, so Meta is doing this on purpose and hoping they don't get caught.
And if they do, so what? Any legal consequences would have very little impact on the benefits they gained.
META wins again.
I've visited CivitAi a while back and it was full of lewd models and LoRA's. Based on that I do think that AI generated porn unironically has the most potential of all their AI products so far. It makes sense for them to do this.
I'm trying to imagine a Facebook Superintelligence.
Imagining AM but even more racist and really fuckin' stupid. Super intelligent but just dumb as hell.
Seeding or leeching?
Surprised they didn't deflect and say it was just "Linux ISOs"
I'm guessing it's to further the Great Smut Purge thats getting shoved down our throats now.
Look I have my problems with ai sure. But everyone deserves to get their freak on.
I mean, I don't want this ai gen shit anywhere but I especially don't want it in my porn.
yeah but common. you can let it enjoy some porn. everyones human here.
I Have No Penis, And I Must Goon