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

I often want good data, so I validate everything.

ChatGPT tends to only give a limited number of results in the response.



GPT-4 has made some substantial improvements recently in references. A few weeks ago using it would get things sort-of correct (the authors would be nearly the same, with one or two just added in or removed, maybe journal number a bit off, etc) and sometimes hallucinate more wholesale, or at least enough that searching couldn't find it as easily. Now, it gets things pretty much perfectly correct everytime as far as I can tell. It gives links that work, and exactly the right reference, every time, even without plugins. However, with plugins now, it doesn't even matter so much any more since it can browse anyway. Very impressive nonetheless.




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

Search: