Hacker News Reader

Comment by DrRobinson | Hacker News Reader

I think it's worth pointing out that this is an answer generated by ChatGPT. It seems a bit dishonest to hide that fact since OP asked what helped _you_.

View on HN

The answers seemed generic but I didn't even notice. This is just going to get worse over time, isn't it?

https://i.imgur.com/FpZ167W.png

Interesting, what do you use to generate this?

I wrote a userscript (with help from ChatGPT) that identifies if comments on HN are written by AI or a human. I based it on https://huggingface.co/openai-detector. Its still a little shabby and only works on HN, but I imagine this is going to be required for general non-specific Internet browsing going forward. It could even be expanded to hide AI generated comments. Looks like this: https://i.imgur.com/BTt1DTh.png

Hmm, the HuggingFace detector seems to identify notes from my personal notebook as being about 80% fake. Passages from Wikipedia, between 72% and 96% fake. I think this is actually detecting "grammatically correct well-punctuated prose" rather than "from GPT-2".

Yes. The point is that it is intended to detect comments obviously written completely by GPT, not detecting that a human a wrote it. You have to discern between the two, as the scale is not really utilized here. What I've found over the last few days of playing with this is that most AI generated text will be 99.97% accurately identified, while human text ranges somewhere between 70-98. So I could move the threshold for my script quite a lot. I just wanted to see how it works with these thresholds for a while.

Reply on news.ycombinator.com