← Back to context

Comment by mgraczyk

16 hours ago

I'd like to brag that I got in trouble for saying this to somebody in 2021, before ChatGPT

I put a chapter of a paper I wrote in 2016 into GPTZero and got the probability breakdown 90% AI, 10% human. I am 100% human, and I wrote it myself, so I guess I'm lucky that I didn't hand it in this year, or I could have gotten accused of cheating?

  • That's more an indictment of the accuracy of such tools. Writing in a very 'standard' style like found in papers is going to match well with the LLM predictions, regardless of origin.