I tested 11 AI detectors on my pre-ChatGPT writing and I'm as little as 5% human

A developer tested 11 AI detectors on their own pre-ChatGPT writing from 2019 and found that the detectors classified them as only 5% human, suggesting that the tools may produce false positives for human-written text.

Article URL: https://originalseparation.substack.com/p/i-have-been-95-robotic-since-2019 Comments URL: https://news.ycombinator.com/item?id=48922160 Points: 1 Comments: 0