I agree that the nature of the algorithm allows it to generate false results which would not be acceptable in many domains. But I think there is tremendous potential to improve what you might call the "baseline" of what information and diagnosis is available. There are great doctors out there for sure but in many places there are close to zero doctors, or worse, bad doctors. While it's a very complex domain there are much simpler parts of it. A child with a cancer tumor in their brain is a very different case than someone who has a rash and a headache. There's a great deal of regulation as well that will come into play here so it's going to take a while. I know Google has a whole LLM that is using only healthcare data that they have been working on for a while.