The research is fine if you actually read it and understand what they’re researching.

It’s almost always the headlines and PR pieces that exaggerate it.

“ChatGPT is more empathetic than Reddit doctors” isn’t interesting. Strip the “Reddit” out and then everyone can substitute their own displeasures with doctors and now it’s assumed true.

Not to mention knock-on-effects of teens trying to decide what to be when they grow up. Some would-be-docs are going to read this and go, huh, maybe I should pick a professional field that AI won’t take over. And bam, in 15 years, there are slightly less docs.

At least in the US there are more people who want to be doctors than positions available. The bottleneck is medical school acceptance rates. To practice medicine you need a medical license, which you can only get from an accredited university.

I knew students who were rejected from medical school, but I also knew far more students who were at one time pre-med, saw how much effort and debt that would-be doctors would need to take on, and saw the risk of pursuing an e.g. biology degree where their entire future hinges on getting into med school, and they chose a different field.

> It’s almost always the headlines and PR pieces that exaggerate it.

That is definitely happening here.

https://jamanetwork.com/journals/jamainternalmedicine/fullar...

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

Yikes. (I do fault the researchers quite a bit for quietly slipping that little detail into a page-long "limitations" section.)