Hacker News Reader

I tried this on my recent AI satire article. While ChatGPT detected my story as AI (because I wrote it like a machine), GPTZero gladly says it's human written.

ChatGPT's response: https://imgur.com/a/u7iBBaX

GPTZero's response: https://imgur.com/a/89icz2X

To be clear, my story was ChatGPT-assisted. But I wonder why ChatGPT couldn't detect it correctly like GPTZero?

View on HN

If your story was ChatGPT assisted one can argue that it was GPT0 which misclassified it.

Correction: I used ChatGPT to discover new words and rephrase the story myself.

Here's the link to my non-paywall story for references:

https://medium.com/humor-bytes/i-your-nba-highlights-broadca...

On the topic of assistance, where do detectors draw the line between AI-augmented, aided, and generated writing?

I don’t know which is why this is completely up for debate, Office also offers grammatical corrections and you have websites like Grammarly that do the same often without any “AI” as far as I know.

Like with anything else this is something that would require society to come to a consensus.

And there are various issues with AI assisted or generated content depending on the context. For example educational institutions may see this as cheating whilst corporate entities might fear legal challenges in relation to copyright and IP.

Reply on news.ycombinator.com