AI based mid level provider

[AR]

Well, then, when it gets widely distributed, I'll start listening to people about AI in medicine.

Look, I'm an academic doc. I do research. I'm NIH funded. And I know quite well that most research never produces anything beyond the "that looks promising" phase. (Mine included.) Getting things to work in medicine is hard. There is undoubtedly a role for machine learning in medicine -- but it's not there yet, and the people who keep talking to me about machine learning a) are mostly spouting hype; and b) have been talking about machine learning in medicine for a long time, and so far all of their predictions about how close we are have been inaccurate.

I think this is a large part of the disconnect. Stop listening to stuff people say. As someone who does research, I'd think it would be natural that you would look at actual research to draw your conclusions. Go look at the literature on AI and EKG interpretation. There are studies that are published that employ the latest technology. You can easily evaluate it and come to an informed opinion on where AI is at.

I agree we're not there yet, but if you look at the literature and think we're never getting there (on this limited problem of EKG interpretation), I think that would be an odd conclusion to draw.

[nastle]

AI and machines are productivity tools. Just like typewriters and copying machines were replaced by PC's. Not to mention things like spreadsheets replaced a ton of punch cards and adding machines. Now it is on your phone. No more operators and no more real secretaries taking short hand. Tools to make your judgments easier.

APN started as physician extenders too just tools to increase physician productivity

[Tim]

APN started as physician extenders too just tools to increase physician productivity

It's not the APN's. It is the lobbying organizations. Human's are the problem, not the APN's. I guess in a way, that is AI. Artificial Intelligence making silly decisions. Not to bright. The doc's are the last line or human intelligence. Used properly, they would improve productivity.
The same can be said for opening Med Schools with the pitch "We will recruit under served and rural candidates that will return to under served and rural locations." Residencies in profit hospitals using cheap labor and turning out new docs. No facilities and no groups or opportunities. Net result is limited options in training and increased competition in the crowded areas.

[Sigrid]

I think this is a large part of the disconnect. Stop listening to stuff people say. As someone who does research, I'd think it would be natural that you would look at actual research to draw your conclusions. Go look at the literature on AI and EKG interpretation. There are studies that are published that employ the latest technology. You can easily evaluate it and come to an informed opinion on where AI is at.

I agree we're not there yet, but if you look at the literature and think we're never getting there (on this limited problem of EKG interpretation), I think that would be an odd conclusion to draw.

I never said we won't get there, *when it comes to the limited problem of EKG interpretation*. I said we've been trying to get there for twenty-plus years and we're not there yet. I have read some of the EKG research, and my opinion was, "that looks promising". A lot of research looks promising, and then doesn't pan out. I do think we will eventually crack the EKG problem. I don't know when it's going to be. I'm also not holding my breath for it.

But this thread is not about EKG interpretation by machine learning, so I think there is a fair amount of talking at cross purposes between you and I. (I admit I'm the one that brought that subject up.) This thread is about machine learning, more broadly, replacing humans in medicine. That is what I am referring to when I say "people have been saying that AI in medicine is five years away for twenty years now," and I remain unimpressed. AI was supposed to have replaced radiologists by now, and it can't even read a mammogram.

As I said above, there will undoubtedly be a role for machine learning in medicine in the future. I don't know exactly what it will look like, and I certainly don't know when it's going to have widespread use. But, to me, it doesn't look like we're anywhere close to The Expanse's autodoc yet.

[AR]

I never said we won't get there, *when it comes to the limited problem of EKG interpretation*. I said we've been trying to get there for twenty-plus years.

I don't know how much stock to put into the last 20 years of "trying". Most of the cutting edge research on AI is focused on things other than EKG interpretation. If we had the AlphaGo/DeepMind folks focus all that talent and those resources on EKG interpretation, there is a good chance it would be solved by now. The problem is that even if AI could perfectly read an EKG it wouldn't actually replace humans in any meaningful way if that is the only thing it could do. So no one is going to be motivated to attack that problem the same way they would beating a go or chess expert.

That is what I am referring to when I say "people have been saying that AI in medicine is five years away for twenty years now,"

I think you must be talking to a weird subset of people. I'm obviously super optimistic about AI and not only have I never heard anyone say we're 5 years away, but I also don't even think we're 5 years away now.

[FIREshrink]

I think you all are severely underestimating the potential power of AI. AI replacing a lot of doc functions will happen. It's just a matter of time.

I don't think we're that close. But it will happen. It is inevitable.

And algorithm based diagnosis and treatment has been shown in some/many cases to cost less and have better outcomes than TAU.