A word of caution on ChatGPT, MD
"However, the use of generative AI in biomedicine also presents challenges and threats, especially in the context of accuracy. Consumer tools like ChatGPT were not trained on the entire corpus of biomedical data, such as full-text articles behind the paywall at major publishers, nor was it tested or trained by experts in biomedicine. The output of these AI systems in response to prompts is therefore often erroneous."
@cyrilpedia
I've been using ChatGPT to research medical topics lately. I use it as a minimally biased observer rather than a tutor. ChatGPT knows how to stay on topic; ChatGPT knows how to weigh things fairly.
It's very good for gaining depth and clarity on subjects you're already reasonably comfortable with, but not so much for stuff you're just learning about. Catching it in a mistake makes it perform much better, for whatever reason. Resource escalation, perhaps.
