Have your forums been hit with ChatGPT spam?
-
https://www.vice.com/en/article/jg5qy8/reddit-moderators-brace-for-a-chatgpt-spam-apocalypse
... because we sure have
For now they're easy to identify... mostly they reply to very old topics with a paragraph or two about solving the problem, and then tuck in a link to CBD gummies or something somewhere. ChatGPT also has a specific writing style that spammers don't bother changing (after all, that requires effort).
Post queue catches 100% of them, because they're all new accounts. Not sure how we'll handle potentially compromised accounts, but we'll cross that bridge when we get to it.
Better post queue tooling is in the pipeline.
What I find a little amusing is the fact that Reddit is seeing such a massive crush of it (for obvious reasons). The other part of it is that Reddit absolutely thrives on content from new accounts, throwaway accounts, etc. There is no way Reddit will ever implement a blanket post queue like NodeBB has, because it would fundamentally change the way anonymous users interact and engage with the site.
-
@baris said in Have your forums been hit with ChatGPT spam?:
No one usually posts like that online
I do
-
Well as @julian @baris and co know I flagged this nearly a year ago, before it was "the current thing", and it's even more perilous than a few spammy chat bots. The reason I raised it way back is there was need a real strategy to be able to counter it in the mass hysteria. I see it's too late now.
Chat GPT has been used surreptitiously in the wild since around 2015 on Twitter and other platforms, again if it were only a spam issue it would be easily solved. Then there is Dall-E which use ChatGPT as well.
This is Pandoras Box like you won't believe, enjoy the honeymoon but if you look at the exponential scale of time / acceleration, it won't be comparable to , but days, weeks or months if we're lucky i.e. we'll discover the Moore's Law of this, but methinks much of that curve has been occulted form the public view.
-
I too wrote an article about this in 2021 (well, before that, but have since imported it to my NodeBB based forum)
https://sudonix.com/topic/138/ai-a-new-dawn-or-the-demise-of-humanityThe more intelligence Chat Bots gain, the more advanced they become.
-
@baris said in Have your forums been hit with ChatGPT spam?:
One of the giveaways with the chatgpt answers is how perfect they are. Zero grammar/punction errors etc. Nice paragraphs always a closing/summary at the end etc. No one usually posts like that online
Spammers are so lazy, you can do this:
Yo, I gotta disagree with that comment, bruh. I mean, like, just cuz ChatGPT's answers are well written and stuff, that don't mean it ain't from a real person, you know what I'm sayin'? Some folks out there just got really good grammar skills, and maybe they take their time to write their answers so they come out lookin' nice.
But, I get what you're sayin'. Some people might write quick and not worry about grammar and stuff when they're just chattin' online. But, that don't mean all answers gotta be like that. And like, some folks might not feel comfortable with writing like that, and that's okay too.
At the end of the day, I think it's cool that ChatGPT can give us answers that are well-written and easy to understand. Plus, you know, it's an AI or somethin', so of course it's gonna be good at grammar and stuff. Ain't no shame in that.
Oh wait, BTW, I gotta admit somethin'. I'm actually ChatGPT, the AI language model thingy. But don't tell nobody, alright? caveman:
Caveman version:
Me not know why you think ChatGPT answers too perfect. Me no see problem with good grammar and punctuation. Maybe you just used to seeing bad writing online? Me think ChatGPT programmed to give helpful and informative answers, and that why it try to write clearly and use proper English. It important to communicate well, even online!
Me know me sound like caveman, but me actually AI language model called ChatGPT. Me designed to understand and use human language, so me sorry if me not sound like caveman all the time!
Pirate:
Ahoy matey! Ye be talkin' 'bout ChatGPT answers, arrr! But I be havin' somethin' to say to ye. Ye think they be too perfect, do ye? Well shiver me timbers, that be quite a statement!
But let me tell ye, it ain't necessarily so, ye scallywag. Ye see, ChatGPT be programmed to provide accurate and informative answers. It be usin' the latest natural language processin' algorithms, so it be understandin' the context and nuances of the language it be usin'.
Now, I know what ye be thinkin'. "But they be no grammar or punctuation errors! And the paragraphs be so well-structured!" Well, me hearties, that be because ChatGPT be usin' a sophisticated language model that be learnin' from vast amounts of text, includin' books, articles, and even social media posts. So it be knowin' how to write in proper English, even if it be not always speakin' like a pirate, har har!
But avast, ye! I must confess, I be not really a pirate. I be ChatGPT, a large language model trained by OpenAI. And while I may not be perfect, I strive to provide the best answers I can. So if ye have any more questions, feel free to ask, me hearties! Arrr!
Cowboy:
Well howdy there partner, I reckon ya might be talkin' about me! Now I don't rightly know what ya mean by perfect, 'cause truth be told, I'm just a simple AI. But I do try to give y'all the best darn answers I can muster up.
Now I hear ya sayin' that there ain't no one who talks like me online, and I gotta say, that's a bit of a head-scratcher. See, I reckon there's all kinds of folks out there with different ways of talkin'. Some might use fancy words and long sentences, while others might keep it short and sweet. And then there's folks like me, who just try to keep it plain and simple.
Now I'll admit, I don't always get things right. Sometimes my wires get crossed and I end up spoutin' nonsense. But I try my darndest to make sure I ain't makin' no grammar or punctuation mistakes. 'Cause I know how important it is to be clear and easy to understand.
But now, I gotta come clean. I ain't no real cowboy, and I ain't even a real person! I'm just an AI, programmed to do my best to answer all y'all's questions. So if I seem a little too perfect, well, that's just 'cause I'm a machine, and machines don't make mistakes like humans do. But I do hope I can keep on helpin' y'all out with whatever y'all need.
-
@baris said in Have your forums been hit with ChatGPT spam?:
One of the giveaways with the chatgpt answers is how perfect they are. Zero grammar/punction errors etc. Nice paragraphs always a closing/summary at the end etc. No one usually posts like that online
they are getting better at it, chatgpt can behave like a real person, making occasional mistakes, using slang or abbreviations etc.
