Re:You Will Lose Your Job to a Robot and Sooner Than You Think
by Piskvorrr on Monday November 06, 2017 @09:12AM
(#15635530)
Be wary of any hockey-stick graphs; as a rule of thumb, its producer is trying to sell you something. (Here, it seems the author is peddling a FUD package). You are quite correct, the text is straight out of a fin-de-siecle technooptimist pamphlet - except in the 1890s, the machines were supposed to liberate mankind, where now they're coming to enslave us.
Re:You Will Lose Your Job to a Robot and Sooner Than You Think
by jcadam on Monday November 06, 2017 @09:28AM
(#15635639)
At this point I almost think you could automate away my Primary care provider. All he does is take vitals, perform a cursory examination (may include a CBC, etc.) and then refer you to a specialist. I wonder why a nurse can't do the exam and plug all of my data into a system that spits out a diagnosis (You have strep throat - 98% confidence) and/or a referral to someone with some real expertise.
Re:You Will Lose Your Job to a Robot and Sooner Than You Think
by dragontamer on Monday November 06, 2017 @11:03AM
(#15636371)
That's actually a big debate right now. But the AI is completely not involved at all. A Nurse Practitioner is 4-years of undergrad + 2-years of Nurses school (which is roughly equivalent to a Masters Degree).
So there's a big belief that highly trained Nurses ought to be good enough for primary care.
-------------
With that said, I've traveled a bit across America and many rural areas only have small ~3 or 4 room "hospitals" that cover entire towns, or even entire counties (~2 hour drive area to the next 5-room "hospital").
In these cases, you definitely want a Doctor who has experience in way more fields.
Nurse Practitioners are probably good enough for primary care in cities, where referrals to specialists is possible. In rural areas where specialists may be a 4+ hour drive away, its simply not feasible to recommend a specialist. The doctor at the small county hospital is all you've got.
Re:You Will Lose Your Job to a Robot and Sooner Than You Think
by Piskvorrr on Tuesday November 07, 2017 @06:54AM
(#15642609)
And there's your problem. It Works for trivial issues, breaks badly for anything outside the happy path. But hey, we only care about the happy path, 80/20, that's good enough for a computer program! (Nowhere good enough for a diagnosis, but that's a minor issue we'll address in the next major version, or handwave away)
In the 1980s, "Expert Systems" were shown to outperform human experts (ie: Doctors) at diagnosing patients in certain situations. (ex: http://www.sciencedirect.com/science/article/pii/03787206859..., a 1985 paper on the subject of "Expert Systems"). Or this article, which has links to proven efficacy of Expert Systems in 1987: http://www.openclinical.org/aiinmedicine.html
30ish years later, we still have doctors primarily delivering our diagnosis. The march of progress may look exponential, but the hype is often overstated.
The terminology "expert system" has morphed into new terms over the years, but the problem has remained the same. Creating a system that can diagnose patients given the data that the patient supplies. Its the diagnosis problem plain and simple... but what has been called "Expert System" in the past will now morph into "AI" and "Machine Learning" under today's terms.
As such, I'm wary about graphs like this one: http://www.motherjones.com/wp-content/uploads/2017/10/201710...
The human brain is really, really, really bad at being a computer. It seems like computers are also really, really, really bad at trying to be a human brain.
The typical layman / non-computer programmer simply fails to understand the limitations of a typical computer. They see computers doing human-like tasks (like winning in Jeopardy), but then don't realize the enormous amount of programming effort that was required for that task to happen.