I once had to spend quite a lot ot time editing a fully AI-generated book, so I believe I have a trained eye on spotting LLMs, without learning specific rules like em dashes. But I will try highlighting what ringed the bell for me.
The very first paragraph:
A large pig gas chamber in the UK has been demolished after decades of operation, becoming a powerful symbol of what animal activists can achieve by exposing practices the meat industry largely keeps out of public view.
The part in bold is suspicious, because it’s so bland and vague. It could apply to basically any animal rights win ever. If humans wrote it, they would probably show more personality here and wrote something punchier.
Also, I feel like AI overuses “powerful symbol of what X can Y by Z” quite a lot, and it’s in this paragraph too. It genuinely used it in probably every paragraph of that sloppy book.
Moreover I tried running this article through GPTZero AI scan tool. Despite it being quite unreliable at times, GPTZero scan believes it’s human written but polished by AI.
Also tried Pangram AI detection tool which claims to have lowest rate of false AI accusations, it said only 68% of the article is fully human written. Everything else is AI-assisted or fully AI.
i read it and dont think it looks written by AI, what makes you think it is?
Hard to say, it’s rather vibes based.
I once had to spend quite a lot ot time editing a fully AI-generated book, so I believe I have a trained eye on spotting LLMs, without learning specific rules like em dashes. But I will try highlighting what ringed the bell for me.
The very first paragraph:
The part in bold is suspicious, because it’s so bland and vague. It could apply to basically any animal rights win ever. If humans wrote it, they would probably show more personality here and wrote something punchier.
Also, I feel like AI overuses “powerful symbol of what X can Y by Z” quite a lot, and it’s in this paragraph too. It genuinely used it in probably every paragraph of that sloppy book.
Moreover I tried running this article through GPTZero AI scan tool. Despite it being quite unreliable at times, GPTZero scan believes it’s human written but polished by AI.
Also tried Pangram AI detection tool which claims to have lowest rate of false AI accusations, it said only 68% of the article is fully human written. Everything else is AI-assisted or fully AI.