The idea that "fancy autocomplete" is already better than the average doctor at taking a list of symptoms, family history, etc and turning them into a diagnosis is actually reasonable when you think about the fact that an LLM simply has a larger body of experience to draw on than any human.
This is an example of the sort of rote task that is 80% of white collar labor that a "fancy autocomplete" system like an LLM will naturally perform as well or outperform humans on over time.
Follow
@carnage4life It's also reasonable when you consider that LLMs are fancy pattern matchers and diagnosing symptoms is exactly pattern matching.
