3 ms·
Being a doctor and diagnosing a patient is really a statistical mapping operation. Symptoms are matched with known diseases/ailments in an effort to find the be
by jqpabc123 3y ago
Being a doctor and diagnosing a patient is really a statistical mapping operation. Symptoms are matched with known diseases/ailments in an effort to find the best statistical fit.
This survey showed an overlap of about 60 percent between doctors and ChatGPT.
Is this a good thing or a bad thing?
The obvious flip side is that they disagree 40 percent of the time. Who is liable for this 40 percent of cases? Are you as a patient willing to accept this?
Any system that disagrees 40 percent of the time probably needs to be scrutinized and verified 100 percent of the time if human life is at stake.
And a doctor could easily argue that this makes it not really all that useful after all ... unless patients are willing to sign away any right to sue for malpractice.
- weatherlite 3y ago> The obvious flip side is that they disagree 40 percent of the time. Who is liable for this 40 percent of cases? Are you as a patient willing to accept this? You're assuming here the doctors in emergency rooms are better than the machine 40% of the time , has anyone verified this? Maybe they're not.
- chrisjj 3y ago> Being a doctor and diagnosing a patient is really a statistical mapping operation. You forgot that the doctor makes observations. ChatGPT does not.
- jqpabc123 3y agoI didn't forget ... I simply gave ChatGPT the benefit of the doubt that it could request observations as desired. My overall conclusion is that ChatGPT would be no more effective as a doctor than it is as a search engine. Basically, every result would need to be independently verified by a human. In other words --- not really all that effective at all.
- chrisjj 3y ago> simply gave ChatGPT the benefit of the doubt that it could request observations Doctors /make/ observations.