Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Eh I think it's untrue prose wise, and I can buy a non-anthropomorphizing argument that ChatGPT can be creative with language use. However I don't think it's untrue when it comes to being truly creative with ideas.

If you ask it about less well established research (but still well known about within whatever scientific community circa 2021) it utterly fails to reason about it or be creative about it at all. It often says wrong things but forgetting that it just doesn't coherently put together different ideas that weren't in its training set IME. Like well below the bar for a conversation with an average lay person about whatever biology topic.

Tbf I haven't tried this much with ChatGPT4, but initial impressions weren't great and I played around with the original release quite a bit. I'd say I was impressed with the language but very rarely with the content.



I talked to a kid about research he hadn’t seen before and I wasn’t impressed by the language or the content.

TBF, he was a kid and he didn’t know why I was talking about. But I think it’s ok to generalize to all humans.


I have soo many counter examples… maybe just keep trying?


Care to share any?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: