I have two minds about this.<p>GPT talks with strong confidence about things, and often in ways where it’s not immediately obvious what is incorrect. That can be a bad thing if the user is unaware of it. Certainly don’t mainline advice from a statistical model into your health practices.<p>On the other hand, this could be an extremely powerful tool for exploration. If it is used to come up with possible issues for an expert to disprove, it could be helpful, depending on the signal to noise ratio.<p>Complicating this is the obscene costs of the American healthcare system, which I worry makes the human part of a potential human-in-the-loop system out of reach for many people.
IBM tried that, too, with Watson. Eventually they had to stop saying it was of any use in medical diagnosis and they quit selling it.<p>ChatGPT answers all take the form of "sources say..." and are full of conjecture. They lack any specific understanding of the patient they're given to, and they're full of contradictions. When pressed for details, ChatGPT will often backpedal on what it just told you.<p>Now some might say, "How like doctors, indeed!" But I still wouldn't recommend seeking out advice from someone or something where you know this to be the case.