I don't know the specifics but my general opinion is that people should use neural networks to crunch data around probability of side-effects of medication or to diagnose tumors, classify sleep cycles and so on.
Asking an AI to diagnose a patient with an LLM is very risky. The LLM will predict the most probably cause but will it pick up on nuances in patients appearance to rule out a less common but more serious disease?
I'm very skeptical about integrating LLMs into the medical field but AlphaFold has moved medicine forward in massive way for example. I want cancer diagnosis and body scans to be filled to the brim with AI but keep it away from my doctor.
