Originality.AI looked at 8,885 long Facebook posts made over the past six years.
Key Findings
- 41.18% of current Facebook long-form posts are Likely AI, as of November 2024.
- Between 2023 and November 2024, the average percentage of monthly AI posts on Facebook was 24.05%.
- This reflects a 4.3x increase in monthly AI Facebook content since the launch of ChatGPT. In comparison, the monthly average was 5.34% from 2018 to 2022.
The other 60% are old people re-sharing it.
and, is the jury already in on which ai is most fuckable?
In the last month it has become a barrage. The algorithms also seem to be in overdrive. If I like something I get bombarded with more stuff like that within a day. I’d say 90% of my feed is shit that has nothing to do with anyone I know.
If it wasn’t a way to stay in touch with family and friends I’d bail.
I was wondering who Facebook was for, good to know AI has low standards
Dead internet theory
FB has been junk for more than a decade now, AI or no.
I check mine every few weeks because I’m a sports announcer and it’s one way people get in contact with me, but it’s clear that FB designs its feed to piss me off and try to keep me doomscrolling, and I’m not a fan of having my day derailed.
I deleted facebook in like 2010 or so, because i hardly ever used it anyway, it wasn’t really bad back then, just not for me. 6 or so years later a friend of mine wanted to show me something on fb, but couldn’t find it, so he was just scrolling, i was blown away how bad it was, just ads and auto played videos and absolute garbage. And from what i understand, it just got worse and worse. Everyone i know now that uses facebook is for the market place.
My brother gave me his Facebook credentials so I could use marketplace without bothering him all the time. He’s been a liberal left-winger all his life but for the past few years he’s taken to ranting about how awful Democrats are (“Genocide Joe” etc.) while mocking people who believe that there’s a connection between Trump and Putin. Sure enough, his Facebook is filled with posts about how awful Democrats are and how there’s no connection between Trump and Putin - like, that’s literally all that’s on there. I’ve tried to get him to see that his worldview is entirely created by Facebook but he just won’t accept it. He thinks that FB is some sort of objective collator of news.
In my mind, this is really what sets social media apart from past mechanisms of social control. In the days of mass media, the propaganda was necessarily a one-size-fits-all sort of thing. Now, the pipeline of bullshit can be custom-tailored for each individual. So my brother, who would never support Trump and the Republicans, can nevertheless be fed a line of bullshit that he will accept and help Trump by not voting (he actually voted Green).
It’s such a cesspit.
I’m glad we have the Fediverse.
And 58.82% are likely generated by human junk then.
If you want to visit your old friends in the dying mall. Go to feeds then friends. Should filter everything else out.
Not my Annie! No! Not my Annie!
This kind of just looks like an add for that companies AI detection software NGL.
this whole concept relies on the idea that we can reliably detect AI, which is just not true. None of these “ai detector” apps or services actually works reliably. They have terribly low success rates. the whole point of LLMs is to be indistinguishable from human text, so if they’re working as intended then you can’t really “detect” them.
So all of these claims, especially the precision to which they write the claims (24.05% etc), are almost meaningless unless the “detector” can be proven to work reliably.
That’s an extremely low sample size for this
8,855 long-form Facebook posts from various users using a 3rd party. The dataset spans from 2018 to November 2024, with a minimum of 100 posts per month, each containing at least 100 words.
seems like thats a good baseline rule and that was about the total number that matched it
With apparently 3 billion active users
Only summing up 9k posts over a 6 year stretch with over 100 words feels like an outreach problem. Conclusion could be drawn that bots have better reach
each post has to be 100 words with at least 100 posts a month
how many actual users do that?
I have no idea because I don’t use the site
But to say less than 0.0001% just seems hard to believe
I don’t use the site either but 100 words is a lot for a facebook post
My number also assumes one post per person so it’s overestimating the %
If you could reliably detect “AI” using an “AI” you could also use an “AI” to make posts that the other “AI” couldn’t detect.
Sure, but then the generator AI is no longer optimised to generate whatever you wanted initially, but to generate text that fools the detector network, thus making the original generator worse at its intended job.
I see no reason why “post right wing propaganda” and "write so you don’t sound like “AI” " should be conflicting goals.
The actual argument why I don’t find such results credible is that the “creator” is trained to sound like humans, so the “detector” has to be trained to find stuff that does not sound like humans. This means, both basically have to solve the same task: Decide if something sounds like a human.
To be able to find the “AI” content, the “detector” would have to be better at deciding what sounds like a human than the “creator”. So for the results to have any kind of accuracy, you’re already banking on the “detector” company having more processing power / better training data / more money than, say, OpenAI or google.
But also, if the “detector” was better at the job, it could be used as a better “creator” itself. Then, how would we distinguish the content it created?
I’m not necessarily saying they’re conflicting goals, merely that they’re not the same goal.
The incentive for the generator becomes “generate propaganda that doesn’t have the language chatacteristics of typical LLMs”, so the incentive is split between those goals. As a simplified example, if the additional incentive were “include the word bamboo in every response”, I think we would both agree that it would do a worse job at its original goal, since the constraint means that outputs that would have been optimal previously are now considered poor responses.
Meanwhile, the detector network has a far simpler task - given some input string, give back a value representing the confidence it was output by a system rather than a person.
I think it’s also worth considering that LLMs don’t “think” in the same way people do - where people construct an abstract thought, then find the best combinations of words to express that thought, an LLM generates words that are likely to follow the preceding ones (including prompts). This does leave some space for detecting these different approaches better than at random, even though it’s impossible to do so reliably.
But I guess really the important thing is that people running these bots don’t really care if it’s possible to find that the content is likely generated, just so long as it’s not so obvious that the content gets removed. This means they’re not really incentivised to spend money training models to avoid detection.
Keep in mind this is for AI generated TEXT, not the images everyone is talking about in this thread.
Also they used an automated tool, all of which have very high error rates, because detecting AI text is a fundamentally impossible task
AI does give itself away over “longer” posts, and if the tool makes about an equal number of false positives to false negatives then it should even itself out in the long run. (I’d have liked more than 9K “tests” for it to average out, but even so.) If they had the edit history for the post, which they didn’t, then it’s more obvious. AI will either copy-paste the whole thing in in one go, or will generate a word at a time at a fairly constant rate. Humans will stop and think, go back and edit things, all of that.
I was asked to do some job interviews recently; the tech test had such an “animated playback”, and the difference between a human doing it legitimately and someone using AI to copy-paste the answer was surprisingly obvious. The tech test questions were nothing to do with the job role at hand and were causing us to select for the wrong candidates completely, but that’s more a problem with our HR being blindly in love with AI and “technical solutions to human problems”.
“Absolute certainty” is impossible, but balance of probabilities will do if you’re just wanting an estimate like they have here.
I have no idea whether the probabilities are balanced. They claim 5% was AI even before chatgpt was released, which seems pretty off. No one was using LLMs before chatgpt went viral except for researchers.
chat bots have been a thing, for a long time. I mean, a half decently trained Markov can handle social media postings and replies
Chatbots doesn’t mean that they have a real conversation. Some just spammed links from a list of canned responses, or just upvoted the other chat bots to get more visibility, or the just reposted a comment from another user.
Im pretty sure chatbots were a thing before AI. They certainly werent as smart but they did exists.
Title says 40% of posts but the article says 40% of long-form posts yet doesn’t in any way specify what counts as a long-form post. My understanding is that the vast majority of Facebook posts are about the lenght of a tweet so I doubt that the title is even remotely accurate.
Yeah, the company that made the article is plugging their own AI-detection service, which I’m sure needs a couple of paragraphs to be at all accurate. For something in the range of just a sentence or two it’s usually not going to be possible to detect an LLM.
> uses ai slop to illustrate it
That laptop lol.
The most annoying part of that is the shitty render. I actually have an account on one of those AI image generating sites, and I enjoy using it. If you’re not satisfied with the image, just roll a few more times, maybe tweak the prompt or the starter image, and try again. You can get some very cool-looking renders if you give a damn. Case in point:
😍this is awesome!
A friend of mine has made this with your described method:
PS: 😆the laptop on the illustration in the article! Someone did not want pay for high end model and did not want to to take any extra time neither…
Seems like an appropriate use of the tech
It’s incredible, for months now I see some suggested groups, with an AI generated picture of a pet/animal, and the text is always “Great photography”. I block them, but still see new groups every day with things like this, incredible…
I have a hard time understanding facebook’s end game plan here - if they just have a bunch of AI readers reading AI posts, how do they monetize that? Why on earth is the stock market so bullish on them?
As long as they can convince advertisers that the enough of the activity is real or enough of the manipulation of public opinion via bots is in facebook’s interest, bots aren’t a problem at all in the short-term.
surely at some point advertisers will put 2 and 2 together when they stop seeing results from targeted advertising.
I think you give them too much credit. As long as it doesn’t actively hurt their numbers, like x, it’s just part of the budget.
Engagement.
It’s all they measure, what makes people reply to and react to posts.
People in general are stupid and can’t see or don’t care if something is AI generated
they measure engagement, but they sell human eyeballs for ads.
But if half of the engagement is from AI, isnt that a grift on advertisers? Why should I pay for an ad on Facebook that is going to be “seen” by AI agents? AI don’t buy products (yet?)
yes, exactly.
Engagement is eyeballs looking at ads
… unless it’s AI masquerading as eyeballs looking at ads.
AI can put together all that personal data and create very detailed profiles on everyone, automatically. From that data, an Ai can add a bunch of attributes that are very likely to be true as well, based on what the person is doing every day, working, education, gender, social life, mobile data location, bills etc etc.
This is like having a person follow every user around 24 hours per day, combined with a psychologist to interpret and predict the future.
It’s worth a lot of money to advertisers of course.
For me it’s some kind of cartoon with the caption “Best comic funny 🤣” and sometimes “funny short film” (even though it’s a picture)
Like, Meta has to know this is happening. Do they really think this is what will keep their userbase? And nobody would think it’s just a little weird?
Engagement is engagement, sustainability be damned.
Well, maybe it is the taste of people still being there… I mean, you have to be at least a little bit strange, if you are still on facebook…
