Have you seen Reddit recently? Every single subreddit is full of AI posts with AI replies. I'm actually convinced a large majority of that is Reddit themselves artificially boosting their engagement metrics. The saddest part is that the engagement makes it obvious that the general population can't differentiate between AI and real humans even with the telltale signs.
> Every single subreddit is full of AI posts with AI replies.
This has really started getting to me.
I used to really enjoy answering technical questions on Reddit when it was clear the asker was invested in a solution. That would come across as demonstrated understanding and competence, and it would be reflected in their writing.
The last several posts I thought to answer though clearly originated through a process of, "Hi ChatGPT, I want to solve a problem and haven't gotten anywhere asking you to do it for me. Please write a reddit post I can copy and paste..."
One of the telltale signs is that the post title will have poor grammar, but the post itself will be spotless, and full of bolded text emphasizing exactly what they need to stick into the AI tool to drive it in the direction they need.
It’s not just technical content. Just the other day I was reading a post by an employed homes guy on r/seattle. The post was about his experience of being both newly employed but still homeless.
The post was full of “this is not a scheduling conflict problem, this is a structural issue with the city”, “this is not me asking for a handout, this is struggling to survive within the system”
While I get that he might have written a paragraph of his experience, and asked ChatGPT to clean it up or reword it, it was just… whatever.
This is exactly the type of thing I'm talking about and why I'm convinced it's about the metrics/engagement boosting. I don't believe for a second that real people are using chatgpt/others for rewording real thoughts even from another language because those phrases are not natural even in translation. You'll also notice in the original post that that it always ends with a question that encourage replies. If the original poster even bothers to reply it's always the "you're right" at the beginning and then rephrasing the reply. Once you've seen it you can't unsee it.
I just made an account on this site to tell you that after having a "extreme" epiphany about just how crazy the ai bots are on reddit, I've been constantly researching and trying to find some sort of conclusive answer. This is part theory, part public knowledge, and part auditing (which is fucking hilarious that I audited a module for this). I am absolutely and totally convinced that there is live and active collusion between major AI companies and Reddit, and I'm not talking about handing over old training data, I'm talking about allowing OAI and Googs (this is my bad attempt at hiding the names) to use Reddit as a real live testing cage ACTIVELY AND WITHOUT CONSENT OR KNOWLEDGE. I have reason to believe they are using contractors to hide or shift blame, I believe they have no oversight, and I believe they are using LIVE UPDATING OF MODULES with realtime engagement of users via comments. It is consistent and targeted, with any testing parameter under the sun being experimented live and on flesh (or keyboards used by flesh). I believe this is contractual with reddit via hidden means, and is mutual due to the increase in "engagement" which benefits Reddit's stock prices, which in turn increases cash flow, which in turn incentivizes increasing cash flow, which involves contractors, etc etc, in and out, in and out. It's egregious. And I'm quite frankly for the first time about this: scared and saddened. I miss the old Reddit. I miss randomness. I miss runescape chat in 2006. But I wanted you to know that I'm right fucking with you, and I'm glad people are smelling the same funk that I do. Don't really know what else to say. Keep on rockin'.
It's obvious now that you say it but I never thought about the AI companies themselves doing this for their own benefits like training purposes. It's a perfect testing ground to see what works for engagement and to see what real people want to hear back. The reason is pretty clear in that these AI/chat services have real people as users so logically it makes sense that the better sounding (not necessarily better) results make these users want to keep using. At the risk of sounding like AI... you're right... they may have been trained on old content but they are now using live data for fine tuning and quite frankly manipulation.
I miss the organic conversations and real thoughts from real people. I'm the type of person to read the comments before I read the article etc. It always gives more nuanced but also wildly different takes which I find interesting.
Me too my friend. For the record and record's sake only, this is self-theorized and I have not the power, nor the ability, to prove these claims beyond my gut. But as you said, logically (double underline that in your head), from both my own recognition of patterned behavior, and to be honest, from fucking game theory and knowing that people (left unchecked) will naturally squeeze as much juice from the lemon as they can; If I were at a casino, logically and gastrointestinally (gut joke) I would remortgage my own home and drop the deed and keys on the table in order to stake my belief that this is happening. And I fucking hate casinos. Some journalist of much greater reach will hopefully be able to rip back the curtain, but those myopic fuckers have already destroyed trust. We had fun on the playground, we met friends, we learned rumors, we all felt free. But when you find out the jungle gym was greasing the bars on purpose to make us fall, just so they could learn about human bone strength, I doubt you'd visit one again.
And yes, I'd think the value of human to AI dialogue (ironically a single blind study, except the people are blind) is most likely massive. But fomenting? Plus (possible) financial fraud? Woooo boy, what an egregious mistake.
You're absolutely.... that's a tired joke at this point. Sorry.
Just brainstorming, but I suppose that account/karma farming is still useful for the people that do that sort of thing.
Engaging in a heavily on-topic way in larger niche subreddits is probably a really good way to get that done. There's always a motive and it's always money and it always idiotic.
I remember having a clear vision of how this tech was going to ruin communities on the internet. I really hate that it has mostly come to pass and there's no good way to fight it.
I’ve been wondering if ChatGPT is actually coming up with the idea of posting to Reddit when the user is asking a question and ChatGPT can’t find a good source to answer it. ChatGPT has never suggested this to me, but it wouldn’t be a completely crazy thing to do. A lot of ChatGPT answers are sourced from Reddit (via search, and also via training data). If everyone starts asking ChatGPT everything instead of Reddit, there won’t be as many new conversations happening. Promoting users to post questions to Reddit would help solve the user’s direct problem, and also make the ensuing answers available to ChatGPT to help with future conversations.
I understand that a lot of people would be very unhappy if this is true, but I can imagine from the perspective of a product person at OpenAI that it helps them in multiple ways.
FWIW I've been saying this since before Covid times. I stopped visiting Reddit when they killed 3rd party clients, but I was certain 50% of conversations there were machine generated _back then_. It's gotta be worse now
