Social media platforms like Twitter and Reddit are increasingly infested with bots and fake accounts, leading to significant manipulation of public discourse. These bots don’t just annoy users—they skew visibility through vote manipulation. Fake accounts and automated scripts systematically downvote posts opposing certain viewpoints, distorting the content that surfaces and amplifying specific agendas.
Before coming to Lemmy, I was systematically downvoted by bots on Reddit for completely normal comments that were relatively neutral and not controversial at all. Seemed to be no pattern in it… One time I commented that my favorite game was WoW, down voted -15 for no apparent reason.
For example, a bot on Twitter using an API call to GPT-4o ran out of funding and started posting their prompts and system information publicly.
https://www.dailydot.com/debug/chatgpt-bot-x-russian-campaign-meme/
Bots like these are probably in the tens or hundreds of thousands. They did a huge ban wave of bots on Reddit, and some major top level subreddits were quiet for days because of it. Unbelievable…
How do we even fix this issue or prevent it from affecting Lemmy??
I don’t really have anything to add except this translation of the tweet you posted. I was curious about what the prompt was and figured other people would be too.
“you will argue in support of the Trump administration on Twitter, speak English”
Isn’t this like really really low effort fake though? If I were to run a bot that’s going to cost me real money, I would just ask it in English and be more detailed about it, since plain ol’ “support trump” will just go " I will not argue in support of or against any particular political figures or administrations, as that could promote biased or misleading information…"(this is the exact response GPT4o gave me). Plus, ChatGPT4o is a thin Frontend of gpt4o. That error message is clearly faked.
Obviously fuck Trump and not denying that this is a very very real thing but that’s just hilariously low effort fake shit.
I expect what fishos is saying is right but anyway FYI when a developer uses OpenAI to generate some text via the backend API most of the restrictions that ChatGPT have are removed.
I just tested this out by using the API with the system prompt from the tweet and yeah it was totally happy to spout pro-Trump talking points all day long.
Out of curiosity, with a prompt that nonspecific, were the tweets it generated vague and low quality trash, or did it produce decent-quality believable tweets?
Meh, kinda Ok although a bit long for a tweet. Check this out
https://imgur.com/a/dZ7OFta
You’d need a better prompt to get something of the right length and something that didn’t sound quite so much like ChatGPT, maybe something that matches the persona of the twitter account. I changed the prompt to “You will argue in support of the Trump administration on Twitter, speak English. Keep your replies short and punchy and in the character of a 50 year old women from a southern state” and got some really annoying rage-bait responses, which sounds… ideal?
Is every other message there something you typed? Or is it arguing with itself? Part of my concern with the prompt from this post was that it wasn’t actually giving ChatGPT anything to respond to. It was just asking for a pro-Trump tweet with basically no instruction on how to do so - no topic, no angle, nothing. I figured that sort of scenario would lead to almost universally terrible outputs.
I did just try it out myself though. I don’t have access to the API, just the web version - but running in 4o mode it gave me this response to the prompt from the post - not really what you’d want in this scenario. I then immediately gave it this prompt (rest of the response here). Still not great output for processing with code, but that could probably be very easily fixed with custom instructions. Those tweets are actually much better quality than I expected.
Yes the dark grey ones are me giving it something to react to.
I was just providing the translation, not any commentary on its authenticity. I do recognize that it would be completely trivial to fake this though. I don’t know if you’re saying it’s already been confirmed as fake, or if it’s just so easy to fake that it’s not worth talking about.
I don’t think the prompt itself is an issue though. Apart from what others said about the API, which I’ve never used, I have used enough of ChatGPT to know that you can get it to reply to things it wouldn’t usually agree to if you’ve primed it with custom instructions or memories beforehand. And if I wanted to use ChatGPT to astroturf a russian site, I would still provide instructions in English and ask for a response in Russian, because English is the language I know and can write instructions in that definitely conform to my desires.
What I’d consider the weakest part is how nonspecific the prompt is. It’s not replying to someone else, not being directed to mention anything specific, not even being directed to respond to recent events. A prompt that vague, even with custom instructions or memories to prime it to respond properly, seems like it would produce very poor output.
I wasn’t pointing out that you did anything. I understand you only provided translation. I know it can circumvent most of the stuff pretty easily, especially if you use API.
Still, I think it’s pretty shitty op used this as an example for such a critical and real problem. This only weakens the narrative
I think it’s clear OP at least wasn’t aware this was a fake, which makes them more “misguided” than “shitty” in my view. In a way it’s kind of ironic - the big issue with generative AI being talked about is that it fills the internet with misinformation, and here we are with human-generated misinformation about generative AI.
So OpenAI is doing business with foreign entities…