I did not read this particular article, I was explaining that from my own experience, all these articles telling how great ChatGPT is seem to be made up because my experience (and from what I read I'm not alone) is completely opposite. Maybe it's not able to solve the type of questions I ask? Fine. But it's not how ChatGPT is presented most of the times.

> I did not read this particular article

It costs you zero dollar to not post an irrelavant comment then. I really wonder how you justify this "I didn't read the article, but I have a very, very strong opinion (straight up calling it a made-up) on it" behavior. The internet is rotting people's brains I guess.

I've now read the article and I'm not changing my opinion. ChatGPT can do some things very well and people tend to hype those things and claim it's better than humans rather than recognising its limits. And again, you are still missing the point of my comment despite having read it, so I'm out of patience. Maybe try to ask ChatGPT to explain what I meant :)

And it costs zero to you to ignore my comment especially if you don't understand it. Other people seem to have understood what I meant and posted constructive responses, you didn't, but honestly it's not my problem.

Ah, I see. You just don't realize that your comment is not relavent to the original article... (of course not, because you didn't read it)

To be fair, the opinion is thoroughly justified by the article, which might have been more honestly titled Physicians' Reddit comments shorter than ChatGPT responses; relative accuracy unknown...

Not really. A team of licensed health care professional rated "the quality of information provided".

And unsurprisingly, the average 52 word Reddit comments [isolated from the context of other comments] didn't provide very much information compared with a much more verbose chatbot. The relevance of the ChatGPT response to the actual patient condition remains unknown.

This is relevant to the real world of primary care only if your sole access to a medical professional is Reddit...

Their first example of a good ChatGPT answer - about bleach in the eye - feels like copypasted SEOified liability-proof WebMD copy. Every medical site has that crap and it’s useless once you have a moderately difficult question.

N.B. as well: If someone thinks they have bleach in their _eye_ and can still open their eyes enough to write a Reddit post, much less read through ChatGPT’s extremely long answer, they’re almost certainly fine.

> Is there something about the premise of the study or its method that you feel are not good?

Yes. Down in the limitations section of the study (https://jamanetwork.com/journals/jamainternalmedicine/fullar...):

> evaluators did not assess the chatbot responses for accuracy or fabricated information

That is a... significant issue with the methodology of the study.