Hi I have noticed on here that sometimes someone will reply with what looks like a very AI response to someone’s question. Initially the message looks kind and understanding, but after a while it seems obvious to me that it’s AI. (As someone who has tried Chat GPT a few times). I am wondering if people might use it to put a ‘good’ reply to a thread on here? And genuinely mean well, or if it’s just weird? It makes me feel uncomfortable that might just be me though.
A concerning thing is that some original posts may have been generated by AI. GPTZero estimates that between the five original posts I checked, 61% of the text is AI generated. Some of these are phrased as leading questions with a common pattern between the posts. Some have spelling mistakes edited into the AI text.
The issue for me isn’t about denying people who are in genuine need of help with phrasing something.
This is interesting and a bit depressing. Thank you for conducting some research. GPTZero checks if the post was AI created and how much of the post is AI?
Perhaps, in the future, on forums like this there could be an inbuilt thingy that each post goes through that tells the reader what percentage is AI and to what degree it’s the regurgitation kind of AI or the or original human thought with help with articulation kind?And what the likely hood is there is no real human involved at all? But again as it likely relies on on AI can it be trusted? I’m not well read on the subject (if everything you can read has been produced by AI what’s the point?) but it’s easy to imagine several incredibly depressing scenarios regarding the future of AI.
Could there be posts here that have no flesh and blood human involved? Just AI posing as a human?
I write academic stuff regularly. I wouldn’t touch AI with a bargepole because I think it comes out with quite bland stuff, i wouldn’t even trust it to write an essay plan because again the plans are depressingly bulk standard (I know this because I know someone who uses it for everything whose work I read).
I worry that my completely non-AI generated content can be flagged as AI - although this has never happened. But my main worry is that academic institutions use an AI checker when grading my work to check that I haven’t written it using AI (plagiarism), and that means my work is then useable for AI because it has been uploaded [to the cloud? Not sure that’s the right term anymore?/to the AI database?]….. and as far as i can see there is no way to opt out of this and continue to study. That means that AI can use my ideas - and even my writing style, which is far far far less desirable but maybe when people ask it to write something that is “non-AI sounding” or more human sounding or erroneous? So I worry about both my good ideas and my specific style of poor grammar being stolen from me and becoming part of the blob. I feel strangely protective of my poor sentence construction and awful grammar.
I suppose the point I’m making is those of us who don’t want to merge indiscriminately with a big blob of data seem to have little choice. And here in this thread people are expressing a want to know to what degree they are corresponding with Copy&Paste, AI, or human… but it doesn’t seem entirely possible to know because we use AI to check and it’s reading all our posts anyway presumably? The mind boggles.
I’d better stop or my brain (which is not the size of a planet or even a large server and is made of real biological matter) will explode and my nervous system will become very frayed.
GPTZero checks if the post was AI created and how much of the post is AI?
It gives a percentage of ‘chance’ how much of the text is AI generated.
I think it would be an excellent idea to have an inbuilt AI detector that automatically scans posts and gave them a percentage rating of AI generated text. Yes, I agree that it wouldn’t be full proof, but it would give people the opportunity to make their own minds up and raise awareness of the issue.If they are AI initial first posts they may be identified as spam because new posts are held in a moderation queue. That is more likely with AI generated text that is selling something or has links to something dodgy. I expect it is possible that humans may make an initial post followed by solely AI generated second and subsequent posts, although I don’t know if it’s common practice.
I worry about both my good ideas and my specific style of poor grammar being stolen from me and becoming part of the blob.
I agree it is concerning but it doesn’t directly copy. It can produce only something that resembles your writing style (unlike yours, it would be bland) and it doesn’t claim ownership. In the traditional sense, anyone who has access to your writing can copy and paste without attributing anyway.
And here in this thread people are expressing a want to know to what degree they are corresponding with Copy&Paste, AI, or human… but it doesn’t seem entirely possible to know because we use AI to check and it’s reading all our posts anyway presumably?
The human element of textual analysis in detecting AI is paramount. Human scrutiny cannot be replicated.
It won't hug you - but it won't leave either.
Sometimes people can really struggle because AI does leave them though…. Some research suggests people really struggle when their trusty AI “friend” receives an update instigated by whatever company actually owns the AI and then the AI seems very different because the update changed the tone and quality of the responses and then human feels abandoned and duped. Vulnerable people who may form emotional attachments to AI can be left bereft.
It won't hug you - but it won't leave either.
Sometimes people can really struggle because AI does leave them though…. Some research suggests people really struggle when their trusty AI “friend” receives an update instigated by whatever company actually owns the AI and then the AI seems very different because the update changed the tone and quality of the responses and then human feels abandoned and duped. Vulnerable people who may form emotional attachments to AI can be left bereft.
