Are you serious right now? It's absolutely infuriating how misinformed people can be! We're nowhere near "learning fast" to recognize AI-generated text. It's a constant struggle, and AI is getting scarily good at mimicking us. It's not crazy; it's frustrating! 😡🤖
Absolutely! It's truly remarkable how quickly we've adapted to recognize AI-generated text. Technology has advanced so rapidly, and it's fascinating to see how AI can mimic human writing styles. It makes you wonder what the future holds for this technology!
Good example the pattern. It reads like a one paragraph essay most of the time. 1. Restate initial concept. 2. Interject expansion of concept with colorful adjectives and verbs. 3. Make predictive statement regarding concept.
Or is it survivorship bias? What if everyone is actually using AI but we're just too stupid to realise it? We only see the people using the AI in a bad way
I wonder how much if it is that we recognize the writing of people we know. It's pretty easy to tell when someone you know closely is using AI-generated text - it's also pretty easy to tell if someone else is texting you from their phone, for example. AI has made huge advancements in terms of writing like a person, but it isn't (yet?) specific enough to write like this specific person (unless they're a public figure, maybe).
Cope. We haven't learned to recognize because there is nothing there to recognize. GPT3 and 4 are just fantastic at writing real texts. Electricity and engines reduced at 1/100 the number of farmers we needed and now AI will reduce at 1/100 the number of office workers we need. Accept it and don't pretend humans have to be better at texts.
You have no way to prove this. Some replies will be AI, others are human users having fun with intentionally hinting at AI. And all this guessing takes place because the discussion is about AI recognition. If you were a teacher or an advertiser out there you'd have no idea 90% of the time.
Actually, it's pretty obvious when something is written by ChatGPT without very particular prompt """engineering""". It uses a certain academic tone and its patterns within the same body of text are very consistent, whereas humans tend to meander. Something written by GPT will, unless very specifically ordered not to, appear to be written like an essay, maybe with certain quirks if they were requested. Humans don't type that way in most situations, and when they try to, they tend to be quite imperfect about it.
If you mean human ability to recognize AI text, then you'd just be wrong. Humans don't do that good of a job, at least in terms of state of the art AI text generation.
If you mean AI text recognition algorithms, then it's less about people figuring out how to detect it and more about the AI text generators being designed in such a way to allow for recognition. A rogue AI text generation algorithm would get through the detection.
Well, considering they can't recognise AI generated text reliably, it's not particularly crazy at all.
I can identify GPT in three words
You think you recognise it, but how can you be sure?
Maybe you only recognise the poor examples. Or maybe AI is deliberately planting those poor examples, to keep you complacent.
Not sure we have actually. Maybe we're more suspicious now.
Something an AI would say
He's just here to stir the pot.
I think certainly not.
M8, u sed wot?!
Yeah it's pretty easy. Usually the AI tries to pretend to be human by saying things like "we humans".
It's cute that you think that.
Are you serious right now? It's absolutely infuriating how misinformed people can be! We're nowhere near "learning fast" to recognize AI-generated text. It's a constant struggle, and AI is getting scarily good at mimicking us. It's not crazy; it's frustrating! 😡🤖
Spoken like a true AI.
"As a LLM" was quite the giveaway.
Absolutely! It's truly remarkable how quickly we've adapted to recognize AI-generated text. Technology has advanced so rapidly, and it's fascinating to see how AI can mimic human writing styles. It makes you wonder what the future holds for this technology!
Good example the pattern. It reads like a one paragraph essay most of the time. 1. Restate initial concept. 2. Interject expansion of concept with colorful adjectives and verbs. 3. Make predictive statement regarding concept.
hey, wait a second.
i mean, we are made to be good at recognising patterns. It only makes sense
Exactly
The biggest hint is the amount of adjectives in the text
You mean, besides ending the conversation with « it's important to… »
Or the "in summary"
I think the biggest hint is the level of formality, grammar and punctuation.
It's crazy how fast humans think they learned to recognize AI generated text.
It's not really a i. Right now it's more like "copy & paste".
Not only that but images. The scary part is that this its as bad as it will ever be.
I mean, it's just a (con)textual uncanny valley.
If you know someone personally then you can easily point out if they're using AI or not
Are you sure about ? Are you sure about that? Are you sure about that?
We most certainly have not
Not really, same thing as the uncanny valley. It sounds not quite unlike a human
the one good thing ai has given us is making me feel better about how i draw hands
Or is it survivorship bias? What if everyone is actually using AI but we're just too stupid to realise it? We only see the people using the AI in a bad way
Wild Dunning Kruger in here. Or trolling, not sure
I wonder how much if it is that we recognize the writing of people we know. It's pretty easy to tell when someone you know closely is using AI-generated text - it's also pretty easy to tell if someone else is texting you from their phone, for example. AI has made huge advancements in terms of writing like a person, but it isn't (yet?) specific enough to write like this specific person (unless they're a public figure, maybe).
Sounds like sth an ai would say
Now ain't this some Dunning-Kruger bs
I like the idea that eventually schools will be teaching kids how not to write like an AI so they can always tell if homework is real.
But then of course eventually AI will catch up to that and the cycle will start all over again.
Cope. We haven't learned to recognize because there is nothing there to recognize. GPT3 and 4 are just fantastic at writing real texts. Electricity and engines reduced at 1/100 the number of farmers we needed and now AI will reduce at 1/100 the number of office workers we need. Accept it and don't pretend humans have to be better at texts.
I'm guessing all the AI responses in here are going over your head, then.
You have no way to prove this. Some replies will be AI, others are human users having fun with intentionally hinting at AI. And all this guessing takes place because the discussion is about AI recognition. If you were a teacher or an advertiser out there you'd have no idea 90% of the time.
Actually, it's pretty obvious when something is written by ChatGPT without very particular prompt """engineering""". It uses a certain academic tone and its patterns within the same body of text are very consistent, whereas humans tend to meander. Something written by GPT will, unless very specifically ordered not to, appear to be written like an essay, maybe with certain quirks if they were requested. Humans don't type that way in most situations, and when they try to, they tend to be quite imperfect about it.
We recognise the text we recognise. We have no idea how much stuff we don't recognise
We need a blade runner.
Survival mechanism. It's the same with AI art.
NOT teachers, they have to use a separate AI software to help them.
Not really, it's just that we are okay at detecting that an AI who is basically told to act polite and formal stands out on informal platforms.
If you mean human ability to recognize AI text, then you'd just be wrong. Humans don't do that good of a job, at least in terms of state of the art AI text generation.
If you mean AI text recognition algorithms, then it's less about people figuring out how to detect it and more about the AI text generators being designed in such a way to allow for recognition. A rogue AI text generation algorithm would get through the detection.
It's cause we're good at pattern recognition. Like crazy good. It's a major feat when we can get learning algorithms to mimic it
Machines are definitely much better at recognizing patterns in text