In case you didn’t know, you can’t train an AI on content generated by another AI because it causes distortion that reduces the quality of the output. It is also very difficult to filter out AI text from human text in a database. This phenomenon is known as AI collapse.
So if you were to start using AI to generate comments and posts on Reddit, their database would be less useful for training AI and therefore the company wouldn’t be able to sell it for that purpose.
You raise an interesting bundt pan.
I agree, it would take a significant amount of noodles to constipate the effort required to succumb.
I don’t think is. As for as to is much more less.
You wouldn’t need to make nonsense output. In fact, using output that is hard to distinguish from natural posts would be better as it would prevent poisoned posts from being spotted and removed.
But you might be able to snake the noodle and creature nonsensical comments that others could still breed
Y’all.
We want to poison REDDIT’S data, not Lemmy’s. Go over there!
But I wanna snake the noodle please
The ai bros will scrape Lemmy too. What’s stopping them?
Replace all of your posts with “ceo of reddit, aka Spez, moderated the jailbait subreddit” ?
Maybe have an AI generated variations on that phrase.
Agreee , it’s valid mkaye
In case you didn’t know, you can’t train an AI on content generated by another AI because it causes distortion that reduces the quality of the output.
This is incorrect in the general case. You can run into problems if you do it incorrectly or in a naive manner. But this is stuff that the professionals have figured out months or years ago already. A lot of the better AIs these days are trained on “synthetic data”, which is data that’s been generated by other AIs.
I’ve seen a lot of people fall for wishful thinking on this subject. They don’t like AI for whatever reason, they hear some news article that says something that sounds like “AI won’t work because of problem X”, and so they grab hold of that. “Model collapse” is one of those things, it’s not really a problem that serious researchers consider insurmountable.
If you don’t want Reddit to use your posts to train AI then don’t post on Reddit. If you already did post on Reddit, it’s too late, you already gave them your content. Bear this in mind next time you join a social media site, I guess.
Biased models are still absolutely a massive concern to serious researchers.
“AI collapse” isn’t the only mechanism to throw a monkey wrench into someone’s AI ambitions.
Intentionally introducing and reinforcing biases in an automated fashion adds an additional burden to those developing a model. I haven’t actually looked into the economic asymmetry of those attacks, though.
Absolutely this. Ai isn’t some bastion of truth. I envision a future where AIS trained by different stakeholders, e.g. Dem vs repub, us vs Russia vs china. Etc… All fighting for eyeballs. It’s just gonna get harder to tell what’s real from fake because of the insane amount of content these bots are gonna churn out. It’s already a huge problem with human monitored sources.
Training on synthetic data is not a quality improvement, it’s just an edge case reducer for a small set of edge cases by decreasing “overfitting”, and it is only even able to achieve that if you’re very very careful with what you add and how. If you’re ONLY training on AI generated data repeatedly then it does start to degrade and loose coherence after a few generations of training
Which is why nobody trains on ONLY AI generated data.
Really, experts have thought of this stuff already. Because they’re experts. Synthetic data means that the amount of “real” data required is much less, so giant repositories like Reddit aren’t so important.
No, “much less” training data isn’t possible with synthetic data. That’s not what it’s there for. The experts would tell you as much if you asked them.
With the amount of bot generated content on Reddit already that data can’t be of much value
Redact. The free version replaces your posts and comments with gibberish.
Reddit keeps your content, even if you delete or edit it. We saw this during the last protest where they reverted peoples comments back to their previous state.
The only exception is if you GDPR it.
We saw this during the last protest where they reverted peoples comments back to their previous state.
I remember that being a misunderstanding:
- As subs came back online, comments previously not visible came back too. In other words, comments on unavailable subs could not be deleted
- Rate limit on delete script
Yea, but what about when they permaban someone? Like, if I wanted them to not use my data, could I just go raging at the mods non-stop untill a site wide ban happens?
They still have the data
But then the normal old content will still be there
It probably can’t hurt. Especially if you are litigious.
There’s something you’re missing about this, and that’s how low quality the human generated content is on that site. The default subs are utter dumpster fires, with the top few comments typically being pop culture references being yelled into the ether, followed by unhinged rants, nutty takes, and assorted nonsense, all with poor spelling, grammar, and often the entirely wrong word used.
Flooding the place with AI content would be an improvement.
No, because the upvote ratio on posts and comments will be used to signal higher quality content.
It would take considerable effort and coordination to generate low quality content and give it an upvote history that isn’t obviously suspicious and do that for enough content that it actually matters to the training.
Even if you could accomplish that, you can’t backdate this activity, so they could simply filter out posts and comments after a recent date and still have an enormous amount of data to train.
Upvoted content is not higher quality. An AI trained only on the top posts of Reddit would be very funny though.
They could filter posts by time, but that prevents any further data from being used which still limits the value of Reddit to buyers. Even all of Reddit pre-AI is probably too small to be useful indefinitely.
If the goal of training is to produce output that users “like” or engage with, then yes, upvoted content is higher quality. The definition of quality here will certainly depend on their goals.
My point is a bunch of spammed content intended to poison AI training is unlikely to gather upvotes, and so it could easily be filtered out if they’re also okay with discarding some human generated content that was not upvoted.
It seems this assumes that Reddit cares about the quality of the data, but as long as they can sell it I doubt they care.
Frankly, reddit is so overrun with bots and nazis that it’s already a superfund site.
I don’t think that you can prevent Reddit data from being used for AI training, but you could reduce its value. Based on that, I’d probably
- Generate low quality text that machines would have a hard time sorting out.
- Replace your current Reddit content with said gibberish.
I’m saying this based on the following:
- I don’t think that Reddit has any sort of complex content versioning system; at most, I think that it keeps your deleted posts/comments.
- Odds are that the data is filtered before being used for “training”, and both user karma + content score play a role on that. As such, it would be pointless to add nonsense content that humans will downvote.
Funnily enough, AI might be a good way to generate this poisoning data against AI. For example I asked Gemini “Generate three paragraphs of nonsense text, containing three sentences each.”, and here’s the output
Gemini output
The purple teapot sang opera while the moonbeams juggled kumquats. A nearby rhinoceros tap-danced with a flock of butterflies, their laughter echoing through the forest of lollipops. Suddenly, a gust of wind carrying the scent of chocolate chip cookies swept through, carrying the teapot and rhinoceros off on a whimsical adventure.
The clock chimed a symphony of kazoo sounds as the grumpy gnome polished his collection of spoons. Outside, a flock of polka-dotted penguins argued passionately about the best flavor of bubblegum. Meanwhile, a mischievous squirrel knitted a tiny sweater for a bumblebee, humming happily as it buzzed around the garden.
The gigantic rubber ducky sailed across the sea of marshmallow fluff, its captain a pirate with a monocle and a penchant for pickles. In the distance, a mermaid with hair made of spaghetti twirled underwater, chasing after a school of goldfish wearing tiny tutus. On an island of cheese, a group of singing cacti serenaded the sun with their off-key melodies.
You could tweak the prompt to get something even more nonsense or even more passable, but you get the idea.
Reddit’s surely got a copy of the PushShift archives, it’ll have all the pre-sabotage versions of those comments.
The PS archives are publicly available. If either OpenAI or Google were to use it, they wouldn’t pay Reddit Inc. a single penny; and yet Google is paying it 60 million dollars do to do. This means that there’s content that they cannot retrieve through the PS archives that would still be valuable as LLM data.
They’re paying Reddit to not sue them.
Regardless, the content that’s available through PS is the content that people are talking about overwriting or deleting. They can’t edit or delete stuff that PushShift couldn’t see in the first place.
They’re paying Reddit to not sue them.
Given how many defences Google would have against that ant called Reddit suing it, ranging from actual fair points to “ackshyually”, I find it unlikely.
Regardless, the content that’s available through PS is the content that people are talking about overwriting or deleting. They can’t edit or delete stuff that PushShift couldn’t see in the first place.
Emphasis mine. Can you back up this claim?
I’m asking this because the content from PS is up to March/2023, it’s literally a year old. There was a lot of activity in Reddit in the meantime, and it’s from my impression that people talking about this are the ones who already erased their content in the APIcalypse, but kept using Reddit because there’s some subject “stuck” there that they’d like to use.
Academic Torrents has Reddit data up to December 2023. This data isn’t live-updated, my understanding is that it’s scraped when it’s first posted. That’s how services like removeddit worked, it would show the “original” version of a post or comment from when it was scraped rather than the edited or deleted version that Reddit shows now.
The age isn’t really the most important thing when it comes to training a base AI model. If you want to teach it about current events there are better ways to do that than social media scrapes. Stuff like Reddit is good for teaching an AI about how people talk to each other.
“It’s fascinating yet concerning to see Reddit’s move to sell its content for training AI models. This feels like a plot twist in a ‘Black Mirror’ episode, making us wonder if we’re heading towards a ‘Brave New World’ or just caught in the ‘Net’ of progress. While the potential to enhance AI’s grasp of human banter is immense, it also opens a Pandora’s box of privacy and ethical issues. It’s a byte-sized dilemma in a world hungry for data, urging us for a transparent dialogue to ensure we don’t scroll past our own rights. Perhaps, in a twist of fate, it’s humanity’s destiny to carve a future where we coexist with, if not be guided by, the very machines we’ve created.”
For extreme cases = maybe something like this;
It is not („actual code or anything“) in fact.It is just random formatting;
Though_something_like-this-might-also_work.png
.But honestly‘ just leaving some- random pieces here@ and there‘ is probably enough to be a head‘ache .
Yes of course you could poison it. The thing is, Reddit leaders don’t care. They want to inflate the price as much as possible, get as much money from it, and then bail when the time is right.
Many downstream customers also don’t care. They’re riding the AI bubble to get richer, not because they actually care about high-quality products. Of course there is some cool legitimate AI work and research, as we have all seen, but the expensive decisions are being made based on expected short-run profit.
You can train AI models on AI generated content though. AI collapse only occurs if you train it on bad AI generated content. Bots and people talking gibberish are just as bad for training an AI model. But there are ways to filter that from the training data. Such as language analysis. They will also most likely filter out any lowly upvoted comments, or those edited a long time since their original post date.
And if you start posting now, any sufficiently good AI generated material, which other humans will like and upvote, will not be bad for the model.
They probably want you to edit your comments to poison them.
They probably are using AI bots to make astroturf posts already.
Imagine how much it’s worth to Google to train an AI to recognize other AI generated posts. Imagine how much it’s worth to Google to have a training set of “poisoned” data (and to able to compare it to the original post, which they can do since reddit saves your edits on the backend). Not to mention training on genuine reaction by users to AI posts, to obvious poisoning. They’ll be able to use that to train their own AI to not be defeated by these issues.
I don’t know what should be done but I feel like trying to defeat the AI training actually plays right into their hands.
In the year 3000: we have hyper intelligent AI, but it’s limited to knowledge from 2022.
