The appeal of medicine as biostatistics improves diagnostic prediction algorithms

[philosonista]

Are pre-meds aware of the (possible) potential of computer science to drastically change and/or reduce the role of physicians? Biostatisticians have been working away on machine learning prediction algorithms to diagnose patients that could, eventually, be incorporated in real time into electronic health records. Although such algorithms have low explanatory power at present (usually around 20%, so says the biostatistician I work with), the potential is there and progress is rapid in the era of "Big Data."

Sure, the impact could of course be relatively small, depending on numerous factors. Or it could be quite drastic: As one computer science professor has put it to the pre-meds who flood his general education requirement class, "Why do you guys want to be doctors?! You know a computer will mostly do your job in the future, right?"

If you are aware, how do you feel about it? Does it not deter your interest? Why or why not?

I ask as a pre-med who personally feels deterred from most diagnosis-heavy specialties now in light of it.

---

Members don't see this ad.

 

[Lawpy]

This has been discussed many many times. No, computers will not reduce the role of physicians in the distant future.

The only way computers will reduce physician roles is if there are advanced artificial intelligence/machine learning algorithms that can program a computer to be:

1. Completely self-sufficient
2. Program other machines to be self-sufficient
3. Understand and display empathy and compassion
4. Make decisions that are more rational and efficient than brightest of human minds

As you can see, this is impossible with current technology, and there are global moral and legal standards that will prevent this from happening.

Regarding computer-aided diagnosis, it's still being investigated using AI/machine learning methods, but the current attendings and residents have pointed out that results are so far unreliable. If anything, computer-aided diagnosis can drastically improve the physician's efficiency, which would expand, not reduce, the physician's role.

 

[efle]

If computers/robots can do physicians' jobs, then what jobs at that point are safe? I'll just tell myself that by the time a robot can do diagnostics better than humans, we will have set up a basic income, and I'll be joining the majority of people by not working.

 

[philosonista]

If computers/robots can do physicians' jobs, then what jobs at that point are safe? I'll just tell myself that by the time a robot can do diagnostics better than humans, we will have set up a basic income, and I'll be joining the majority of people by not working.

So why not instead be a computer scientist/biostatistician/etc that can work to out-do and improve the diagnostic accuracy of medicine?

 

[philosonista]

This has been discussed many many times. No, computers will not reduce the role of physicians in the distant future.

The only way computers will reduce physician roles is if there are advanced artificial intelligence/machine learning algorithms that can program a computer to be:

1. Completely self-sufficient
2. Program other machines to be self-sufficient
3. Understand and display empathy and compassion
4. Make decisions that are more rational and efficient than brightest of human minds

As you can see, this is impossible with current technology, and there are global moral and legal standards that will prevent this from happening.

Regarding computer-aided diagnosis, it's still being investigated using AI/machine learning methods, but the current attendings and residents have pointed out that results are so far unreliable. If anything, computer-aided diagnosis can drastically improve the physician's efficiency, which would expand, not reduce, the physician's role.

I'm talking about the long term. Yes, R^2 stands at around 0.2. Not impressive. But biostatisticians I have spoken with are not pessimistic in the long-term.

If computer-aided diagnosis did indeed improve efficiency, how would you see that it would "expand" their role? To me, the drastic reduction in the diagnostic part of the job loses a great deal of the appeal. If I were simply to become more busy doing the other duties of a physician with a greater number of patients, much appeal is lost. I like the intellectual challenge.

 

[Mansamusa]

Are pre-meds aware of the (possible) potential of computer science to drastically change and/or reduce the role of physicians? Biostatisticians have been working away on machine learning prediction algorithms to diagnose patients that could, eventually, be incorporated in real time into electronic health records. Although such algorithms have low explanatory power at present (usually around 20%, so says the biostatistician I work with), the potential is there and progress is rapid in the era of "Big Data."

Sure, the impact could of course be relatively small, depending on numerous factors. Or it could be quite drastic: As one computer science professor has put it to the pre-meds who flood his general education requirement class, "Why do you guys want to be doctors?! You know a computer will mostly do your job in the future, right?"

If you are aware, how do you feel about it? Does it not deter your interest? Why or why not?

I ask as a pre-med who personally feels deterred from most diagnosis-heavy specialties now in light of it.

I feel that the computer scientists are being arrogant, which is to be expected- everyone gets cocky about the influence of their expertise. Of course they feel like their work could take over every else's jobs- doesn't mean they are correct.

There are a ton of other reasons why a computer takeover is unlikely to happen (and I'm too lazy to rehash all those points right now)

It doesn't deter me because basing my career on something that may or may not happen 30-40 yrs in the future is crazy (and who knows what I'll be doing then; maybe I'll be a hospital CEO or a surgeon- who will definitely not be replaced any time soon). I may get cancer at 40 and die, but I'm not going to not go to med school just because I may not live long past my training period. I'm probably at a greater risk of dying at 40 than I am at being replaced by a computer by that time.

(I would never recommend someone base their life around what may possibly happen at some vague point in time because some people not even in the field being discussed say it may potentially happen. Choose the career you want, and put your best foot forward)

 

[JustaDO]

So why not instead be a computer scientist/biostatistician/etc that can work to out-do and improve the diagnostic accuracy of medicine?

So why aren't you doing it then?

 

[wizzed101]

So why not instead be a computer scientist/biostatistician/etc that can work to out-do and improve the diagnostic accuracy of medicine?

Because they would've been doing medicine instead

 

[Lawpy]

I'm talking about the long term. Yes, R^2 stands at around 0.2. Not impressive. But biostatisticians I have spoken with are not pessimistic in the long-term.

If computer-aided diagnosis did indeed improve efficiency, how would you see that it would "expand" their role? To me, the drastic reduction in the diagnostic part of the job loses a great deal of the appeal. If I were simply to become more busy doing the other duties of a physician with a greater number of patients, much appeal is lost. I like the intellectual challenge.

I'm talking long term as well. Medicine, especially diagnosis, is not algorithmic for a computer to replace physician roles. The outcome is very likely that a computer will expand the physician's role by improving efficiency of healthcare delivery.

Skeptics had claimed that computers would wipe out radiologists when modalities like CT/MRI/PET were invented and improved. The opposite has happened as radiologists have become more skilled and critical players in medicine.

However, careers of non-physicians like medical physicists and technicians may very will be lost to computers because their jobs are usually largely algorithmic.

If computers/robots can do physicians' jobs, then what jobs at that point are safe? I'll just tell myself that by the time a robot can do diagnostics better than humans, we will have set up a basic income, and I'll be joining the majority of people by not working.

Human civilization will likely be wiped out by that point.

 

[Lucca]

So why not instead be a computer scientist/biostatistician/etc that can work to out-do and improve the diagnostic accuracy of medicine?

Probably because they want to see patients, ideally.

Check out the "why medicine" thread for other reasons.

 

[efle]

So why not instead be a computer scientist/biostatistician/etc that can work to out-do and improve the diagnostic accuracy of medicine?

Same reason I didn't go into comp sci now, even expecting both careers to last my lifetime. I don't think I'd be happy spending 9-5 in front of a computer screen every day.

 

[sequell]

If anything, computer-aided diagnosis can drastically improve the physician's efficiency, which would expand, not reduce, the physician's role.

If computers/robots can do physicians' jobs, then what jobs at that point are safe?

Here is a report from McKinsey showing how certain duties in nearly every job can be automated; see detailed chart here. I for one welcome any well-executed technology that makes doctors' lives easier, stays out of their way, and helps to make more time for the humanism/empathy component of the job.

So why not instead be a computer scientist/biostatistician/etc that can work to out-do and improve the diagnostic accuracy of medicine?

That's a really good question. I think CS/stats/AI will tempt many young people that are a) good at math and b) interested in healthcare challenges (e.g. working on IBM Watson). I think many docs wanted to go into medicine for the human element (with exceptions in radiology/pathology/public health/etc.), and the science part is important, but...almost ancillary, a means to an end. If that's the case, technology should certainly be ancillary, too.

If I were simply to become more busy doing the other duties of a physician with a greater number of patients, much appeal is lost. I like the intellectual challenge.

I'm pretty sure (human) physicians, along with patients, will always make the final call. However, I can imagine that it will become harder and harder to justify going against the recommendation of a computer algorithm with 10, 20, 30+ years of real-world outcomes and scientific papers programmed into it. That's fine; it just means conceding that the collective knowledge from millions of actual patient outcomes is probably more than the individual memory of a single doctor can reasonably hold, so physicians employ a computer to do it for them and go back to taking care of the patient. How many residents, even now, look up obscure facts on their phones while on rounds? I'm comfortable with my doctor doing that if it helps keep him/her up to date with the latest R&D.

arrogant

Some might say optimistic

maybe I'll be a hospital CEO or a surgeon- who will definitely not be replaced any time soon

You bring up an interesting case study. Surgery as a specialty is very tech-heavy and welcomes new devices, tools, etc. So it's interesting that surgeons love new technologies that help with a specific element of the job (suturing, cutting, visualization, etc.), but as soon as someone tries to help with (i.e. automate) every aspect of the job, a lot of FUD emerges.

 

[ndafife]

Planes effectively fly themselves at this point. Heck, moving to a drone based commercial flight system might even save tons of money for airlines.

But there will always be pilots because people trust humans over machines. They need to see the pilot when they step on the plane.

Same will be true for doctors.

 

[pulmoblast]

If current EMR software is any indication of things to come...


Sent from my iPhone using SDN mobile app

 

[philosonista]

If current EMR software is any indication of things to come...


Sent from my iPhone using SDN mobile app

Right, it's made for billing, not data, as it stands.

 

[philosonista]

Here is a report from McKinsey showing how certain duties in nearly every job can be automated; see detailed chart here. I for one welcome any well-executed technology that makes doctors' lives easier, stays out of their way, and helps to make more time for the humanism/empathy component of the job.


That's a really good question. I think CS/stats/AI will tempt many young people that are a) good at math and b) interested in healthcare challenges (e.g. working on IBM Watson). I think many docs wanted to go into medicine for the human element (with exceptions in radiology/pathology/public health/etc.), and the science part is important, but...almost ancillary, a means to an end. If that's the case, technology should certainly be ancillary, too.


I'm pretty sure (human) physicians, along with patients, will always make the final call. However, I can imagine that it will become harder and harder to justify going against the recommendation of a computer algorithm with 10, 20, 30+ years of real-world outcomes and scientific papers programmed into it. That's fine; it just means conceding that the collective knowledge from millions of actual patient outcomes is probably more than the individual memory of a single doctor can reasonably hold, so physicians employ a computer to do it for them and go back to taking care of the patient. How many residents, even now, look up obscure facts on their phones while on rounds? I'm comfortable with my doctor doing that if it helps keep him/her up to date with the latest R&D.


Some might say optimistic


You bring up an interesting case study. Surgery as a specialty is very tech-heavy and welcomes new devices, tools, etc. So it's interesting that surgeons love new technologies that help with a specific element of the job (suturing, cutting, visualization, etc.), but as soon as someone tries to help with (i.e. automate) every aspect of the job, a lot of FUD emerges.

Thanks a bunch for the comment, whoever you are.

I'm really holding onto medicine at this point out of an interest in surgery. Surgery, however, seems far less automatable. Any telling the degree to which surgery might be automated?

I mean, really, if that is going to go to, I'll just run off to CS grad school or some such thing.

 

[NimbleNavigator]

Planes effectively fly themselves at this point. Heck, moving to a drone based commercial flight system might even save tons of money for airlines.

But there will always be pilots because people trust humans over machines. They need to see the pilot when they step on the plane.

Same will be true for doctors.

Aren't the pilots there in case something goes wrong?

 

[DrBowtie]

Thanks a bunch for the comment, whoever you are.

I'm really holding onto medicine at this point out of an interest in surgery. Surgery, however, seems far less automatable. Any telling the degree to which surgery might be automated?

I mean, really, if that is going to go to, I'll just run off to CS grad school or some such thing.

Even surgery isn't safe:
http://www.scientificamerican.com/article/robot-surgeon-successfully-sews-pig-intestine/

 

[Lucca]

The human element isn't going to save medicine.

That being said, these aren't really reasons to do or not do medicine.

 

[philosonista]

The human element isn't going to save medicine.

That being said, these aren't really reasons to do or not do medicine.

You may not think so. But I'd rather be on the side of creating innovation.

 

[JustAPhD]

Planes effectively fly themselves at this point. Heck, moving to a drone based commercial flight system might even save tons of money for airlines.

But there will always be pilots because people trust humans over machines. They need to see the pilot when they step on the plane.

Same will be true for doctors.

I don't think that will always be the case to be honest. As we progress more towards self driving cars I believe that this idea that humans must be in control will slowly (but surely) die off. At the rate technology increases, I do not think it would be wise to scoff at the idea that perhaps in the future medicine will rely very heavily on AI. However, basing your career choice off of this would be unwise, as I don't believe anyone really thinks that this will occur during our lifetimes.

 

[Petrichor1]

If anything, it would increase the training of doctors in multiple disciplines and perhaps create more fellowship opportunities for those inclined towards data analysis and bio informatics. There is already a good number of ppl taking time off to do a masters or apprenticeship program to stand out and become prepared for new trends and a longer learning path. I do find computers and robotic technology quite invasive in our health systems but most all these technologies need a lot of maintenance. At the least, I see the doctor to be much more competent and someone ppl can go to for much more vital and critical questions. With ai, I can definitely see more paperwork authorized from management's end. The most annoying would possibly be the constant consultation of both the doctor's and AI's opinions and the a following argument to write up in cases where the two do not agree.

 

[Med Ed]

Or it could be quite drastic: As one computer science professor has put it to the pre-meds who flood his general education requirement class, "Why do you guys want to be doctors?! You know a computer will mostly do your job in the future, right?"

Your professor has watched too many episodes of House. Diagnosis is just one element of a physician's job, and it's usually not the most challenging.

Besides, there is the slight problem of garbage in, garbage out. When an computer can reliably pass the Turing test while operating in a busy urban ED, please let us know.

 

[efle]

What's more challenging than figuring out what's wrong? Planning the treatment?

 

[Med Ed]

What's more challenging than figuring out what's wrong? Planning the treatment?

The behavioral side is way more difficult to contend with on a day-to-day basis.

Brain-busting diagnostic dilemmas are few and far between, and when you do have one it often lands on something utterly ambiguous and not helpful. When I was in medical school I had the pleasure of listening to an attending explain to a patient that she likely had "an autoimmune disease that hasn't read the textbooks." Shall we order the usual cocktail of steroids and IVIG, good madam?

 

[Lawpy]

Your professor has watched too many episodes of House. Diagnosis is just one element of a physician's job, and it's usually not the most challenging.

Besides, there is the slight problem of garbage in, garbage out. When an computer can reliably pass the Turing test while operating in a busy urban ED, please let us know.

The behavioral side is way more difficult to contend with on a day-to-day basis.

Brain-busting diagnostic dilemmas are few and far between, and when you do have one it often lands on something utterly ambiguous and not helpful. When I was in medical school I had the pleasure of listening to an attending explain to a patient that she likely had "an autoimmune disease that hasn't read the textbooks." Shall we order the usual cocktail of steroids and IVIG, good madam?

Superintelligent robots can solve all these problems easily!

 

[philosonista]

Superintelligent robots can solve all these problems easily!

No one claimed robots were going to do everything a physician does. Strawman.

If a robot can outdo the diagnostic error of the average physician, I fail to understand why doctors are so against the idea. Though I'm not surprised this is how many of you are reacting. Some biostatisticians I spoke with this summer have said conversations get quite heated with physicians, especially since genomics is now taking off and getting better at producing probabilities of disease based on an individual's genetic makeup.

The current error rate is not okay, really:

http://archinte.jamanetwork.com/article.aspx?articleid=1108559

When it comes down to it, I would rather create technologies and innovation and have a long-term impact than practice medicine on the basis of what technologies and knowledge are available in this lifetime. And if ya'll want to mock that by doubting the potential of AI, well, none of us may be around to see the eventual impact of AI, but I want to be around to push it forward to however far it can go while I am around. And I find it very hard to picture that a computer far more technologically advanced than we have now, pulling from millions of data points and crafting better algorithms in real time ("machine learning") will be worse off than the average doctor.

Biostatisticians are the math nerds of the doctors. Can't we all just be friends?

Maybe if I said "changing" the role of physicians instead of "reducing," ya'll's egos can take a rest.

 

[Lawpy]

No one claimed robots were going to do everything a physician does. Strawman.

If a robot can outdo the diagnostic error of the average physician, I fail to understand why doctors are so against the idea. Though I'm not surprised this is how many of you are reacting. Some biostatisticians I spoke with this summer have said conversations get quite heated with physicians, especially since genomics is now taking off and getting better at producing probabilities of disease based on an individual's genetic makeup.

The current error rate is not okay, really:

http://archinte.jamanetwork.com/article.aspx?articleid=1108559

When it comes down to it, I would rather create technologies and innovation and have a long-term impact than practice medicine on the basis of what technologies and knowledge are available in this lifetime. And if ya'll want to mock that by doubting the potential of AI, well, none of us may be around to see the eventual impact of AI, but I want to be around to push it forward to however far it can go while I am around. And I find it very hard to picture that a computer far more technologically advanced than we have now, pulling from millions of data points and crafting better algorithms in real time ("machine learning") will be worse off than the average doctor.

Biostatisticians are the math nerds of the doctors. Can't we all just be friends?

Maybe if I said "changing" the role of physicians instead of "reducing," ya'll's egos can take a rest.

Actually, superintelligence can in fact solve all medical problems fairly easily. This isn't a surprise since these robots are smarter than brightest human minds. The only downside is it would lead to a cascade effect that would most likely wipe out human civilization.

And this has nothing to do with egos lol. Technology advances medicine yes, but it doesn't make any sense to say that AI/machine learning reduces physician roles. People said the same thing about radiologists/radiation oncologists when new radiological modalities were invented and perfected, and the opposite has happened, with radiology and its subspecialties/related fields becoming increasingly important.

It's simply misleading and inaccurate to present antagonistic relationships between technology and physicians.

 

[wizzed101]

I don't know of a single profession/industry that has robots/machines/industrialization displacing the human workforce without suffering a heavy decline in quality. Consumers don't know how good they are having it until it is taken away from them.

Edit: for example, agriculture. We have so much food now, and GMO will further help us solve the hunger problem. BUT we have a severe shortage of non-GMO, home grown, pesticide-tree food. You can argue that there is nothing wrong with GMO and pesticide had helped solving blah blah blah. That is not the point. People want particular things. The higher the living standard, which is brought about by technology, ironically the more exposed these shortages.

It wouldn't surprise me when there are demands for software hand-coded by virgins in the future

 

[Lucca]

I don't know of a single profession/industry that has robots/machines/industrialization displacing the human workforce without suffering a heavy decline in quality. Consumers don't know how good they are having it until it is taken away from them.

Edit: for example, agriculture. We have so much food now, and GMO will further help us solve the hunger problem. BUT we have a severe shortage of non-GMO, home grown, pesticide-tree food. You can argue that there is nothing wrong with GMO and pesticide had helped solving blah blah blah. That is not the point. People want particular things. The higher the living standard, which is brought about by technology, ironically the more exposed these shortages.

It wouldn't surprise me when there are demands for software hand-coded by virgins in the future

Lol this just isn't grounded in reality and the example you give doesn't make any sense

 

[Lucca]

No one claimed robots were going to do everything a physician does. Strawman.

If a robot can outdo the diagnostic error of the average physician, I fail to understand why doctors are so against the idea. Though I'm not surprised this is how many of you are reacting. Some biostatisticians I spoke with this summer have said conversations get quite heated with physicians, especially since genomics is now taking off and getting better at producing probabilities of disease based on an individual's genetic makeup.

The current error rate is not okay, really:

http://archinte.jamanetwork.com/article.aspx?articleid=1108559

When it comes down to it, I would rather create technologies and innovation and have a long-term impact than practice medicine on the basis of what technologies and knowledge are available in this lifetime. And if ya'll want to mock that by doubting the potential of AI, well, none of us may be around to see the eventual impact of AI, but I want to be around to push it forward to however far it can go while I am around. And I find it very hard to picture that a computer far more technologically advanced than we have now, pulling from millions of data points and crafting better algorithms in real time ("machine learning") will be worse off than the average doctor.

Biostatisticians are the math nerds of the doctors. Can't we all just be friends?

Maybe if I said "changing" the role of physicians instead of "reducing," ya'll's egos can take a rest.

No one is ridiculing you. If this is what excites you and you want to work on it, then do it.

 

[wizzed101]

Lol this just isn't grounded in reality and the example you give doesn't make any sense

Which part of it does not make sense? I said that it doesn't matter if GMO tomatoes have absolutely no difference from the non-GMO ones (or even if they are better objectively), people will still pay 5x more for the "inferior" products if they can afford it.

I made quite an outlandish claim. All you had to do to prove me wrong was to find just ONE counter-example

 

[Lucca]

Which part of it does not make sense? I said that it doesn't matter if GMO tomatoes have absolutely no difference from the non-GMO ones (or even if they are better objectively), people will still pay 5x more for the "inferior" products if they can afford it.

I made quite an outlandish claim. All you had to do to prove me wrong was to find just ONE counter-example

Firs you say consumers don't know how good they have something until it's gone

Then you claim that industrialization and automation of industries leads to a decline in quality.

You then give the examples of GMOs and say it doesn't matter if we do actually get more food and there isn't any reason to suspect the foods are dangerous or harmful and the explanation you provide is "people want particular things". That not only explains nothing but contradicts your first sentence which claims consumers don't really know what they should want.

I'll give a counterexample: cars. Cheaper, safer, and more ubiquitously available thanks to automation. Safer too because humans don't have to be doing most of the work that involves heavy moving parts. Here's another: computers. Of higher quality and standards because they can be made the same way every time. Automation allows different parts from different manufacturers to be compatible with one another because computers have a generic architecture and standardized inputs/outputs for the different parts giving people the freedom to build their own however they want and companies the ability to produce a variety of options of ranging power and quality at different prices. Or, if they want, they can standardize their own product and establish a brand with consistent quality a la Apple.

The list goes on and on and on

 

[Cytokine123]

Did research in machine learning at a top institution. Can confirm it diagnoses better than a doctor in some instances because doctors don't have ****ing high throughput data analytics than they can crap out. You still need a douchey neurotic premed now doctor to interpret it. Don't you piles of anxiety worry, you'll still have jobs.

 

[wizzed101]

@Lucca It's not contradictory at all. When one's immediate need is to have a tomato to feed, anything goes. One the same person has in excesses of everything else, they want that tomato in a very particular way.

In both of your examples, there are more people involved in production than ever. I used agriculture precisely because it's one of a field where automation has significantly displaced human involvement. Just a few hundred farmers can feed millions. It's mind-blowing. Agriculture is "there." Cars and computers are not.

Plus, Apple is a software and design company...

 

[Med Ed]

Maybe if I said "changing" the role of physicians instead of "reducing," ya'll's egos can take a rest.

It's not an ego issue. It's the fact that you have folks with a seemingly limited understanding of medicine making grandiose proclamations about the impact of [insert technology] on its future. Those of us who have been in the biz for any length of time have seen this play out more than once, and to similar effect:

 

[Med Ed]

No one claimed robots were going to do everything a physician does. Strawman.

Perhaps you should develop an AI capable of detecting sarcasm.

Some biostatisticians I spoke with this summer have said conversations get quite heated with physicians, especially since genomics is now taking off and getting better at producing probabilities of disease based on an individual's genetic makeup.

Reference?

 

[Cytokine123]

It's not an ego issue. It's the fact that you have folks with a seemingly limited understanding of medicine making grandiose proclamations about the impact of [insert technology] on its future. Those of us who have been in the biz for any length of time have seen this play out more than once, and to similar effect:

But on the converse, would it be appropriate to say there's a lot of folks with a limited understanding of technology making grandiose proclamations about the its impact (or lack thereof) on [insert medical specialty] on its future?

 

[7331poas]

Because they would've been doing medicine instead

CS is not a downgrade from medicine as a career path lol

@Med Ed Maybe you should look up some of the things CS has done and the growth in those fields has been faster than any in human history. The size of a transistor etc.

The point is that in the future (not in mine I think), there is no doubt that a computer can do a human's job better. That goes for every job that ever existed. We are all just playing the game until that happens though.


I am a CS student moving into medicine because that is what I enjoy. But im not naive.

 

[DrBowtie]

But on the converse, would it be appropriate to say there's a lot of folks with a limited understanding of technology making grandiose proclamations about the its impact (or lack thereof) on [insert medical specialty] on its future?

I'd trust the people doing the day to day job, not those trying to create technology to fit jobs they know little about.

 

[Cytokine123]

I'd trust the people doing the day to day job, not those trying to create technology to fit jobs they know little about.

So no consulting at all goes into collaborating with this technology? I don't know what medical center you are basing your experiences on, but I have seen physicians sitting down with engineers, computer scientists, etc. There are also physicians with backgrounds in computational sciences (think MD PhD, or even MD with an undergrad in computer engineering) who do their day to day job but have a vision for innovating it.

 

[DrBowtie]

So no consulting at all goes into collaborating with this technology? I don't know what medical center you are basing your experiences on, but I have seen physicians sitting down with engineers, computer scientists, etc. There are also physicians with backgrounds in computational sciences (think MD PhD, or even MD with an undergrad in computer engineering) who do their day to day job but have a vision for innovating it.

My experience is that I've seen many new and trending products, software etc. get sold to administrators by salesmen that then attempt to force its usage even though it makes zero sense in day to day usage. Would a collaborator who is getting a consulting fee really tell a company their product is redundant, useless, should be scrapped?

Also these collaborators are often academics which aren't always practicing the reality of day to day clinical practice.

 

[Lucca]

@Lucca It's not contradictory at all. When one's immediate need is to have a tomato to feed, anything goes. One the same person has in excesses of everything else, they want that tomato in a very particular way.

In both of your examples, there are more people involved in production than ever. I used agriculture precisely because it's one of a field where automation has significantly displaced human involvement. Just a few hundred farmers can feed millions. It's mind-blowing. Agriculture is "there." Cars and computers are not.

Plus, Apple is a software and design company...

The hundreds of thousands employed by the factory towns producing apple products in China would be very surprised to hear that Apple is not involved in manufacturing.

I think it's just a fact that less humans are involved in the production of a single car today than were when Ford made his model T so I don't know where you are pulling that from.

Also, that is not what you said. Your example was GMOs and their proposed effect, namely abundance of food. You then said people "want things in a particular way". How can they want things in a particular way if they don't know how good they have it until it is gone? The statement you just made, that when presented with options people generate additional preferences, is exactly the opposite of your original statement, that people do not understand their own preferences until options are absent.

The second statement is also not true. For every 1 person who wants an organic, GMO free, California tomato there are 1 million people who just want a reasonably priced tomato to make salsa with. That fact is precisely why automation of anything succeeds in the market every time it is put into practice; a standardized and cheap-to-make "good enough" product beats the carefully crafted artisan product every time. The latter isn't a mass market product, it is for a niche community whether it be hipsters, people who like the idea of supporting local farmers, or those rich enough to afford consistently buying artisan products

 

[Med Ed]

But on the converse, would it be appropriate to say there's a lot of folks with a limited understanding of technology making grandiose proclamations about the its impact (or lack thereof) on [insert medical specialty] on its future?

I see where you are coming from, but this line of thinking attempts to establish a false equivalence between being on the inside looking out versus outside looking in.

This is one of those debates where it is very easy for each side to talk past the other. I do not believe any reasonable physician would say that AI will never have a place in medicine. However, I do think that many (myself included) would say that being deterred from certain specialties, or medicine in general, due to the perceived threat of computational algorithms is, circa 2016, ridiculous.

If history is any guide, we can state with great certainty that the infusion of the “next big thing” into medical practice will (1) take longer, (2) be more difficult, (3) be more expensive, (4) be less impactful in the near and intermediate term, and (5) carry with it a more unintended consequences than its early proponents will admit to.

Genomics is a good example. Personalized medicine will change the landscape, but it was supposed to happen a number of years ago. Determining meaningful genetic variants has been more challenging than a lot of people anticipated. Cancers were supposed to diagnosed and classified purely by DNA sequencing, but that hasn’t exactly worked out so far, especially when a trained human can often make a diagnosis in seconds with 19th century technology. Or take electronic health records. Not only are they expensive, user-hostile, and unwilling to communicate with each other, they have given rise to an entire new human job (scribes) and bequeathed to us the scourge of copy-and-paste progress notes.

I do, however, look forward to the day when an AI manufacturer is sued because its algorithm missed a diagnosis. Or a hospital is sued because an employee failed to properly enter data. That will be a whole new world.

 

[7331poas]

Any doctor who thinks that general AI cannot, in the future, completely displace then as being vastly inferior providers is delusional. Even right now a ton of medicine is guess work based off vague symptom description from patients.

 

[efle]

Any doctor who thinks that general AI cannot, in the future, completely displace then as being vastly inferior providers is delusional. Even right now a ton of medicine is guess work based off vague symptom description from patients.

A computer can't diagnose based on vague nonspecific symptom descriptions either. If the difficulty is at the info acquisition stage like you just pointed out, there's no vast improvement to be made

 

[7331poas]

A computer can't diagnose based on vague nonspecific symptom descriptions either. If the difficulty is at the info acquisition stage like you just pointed out, there's no vast improvement to be made

general AI without a doubt will be able to discern better than any physician could. Like I said, we are all playing the "working" game until 100-150 years from now at max.

 

[efle]

general AI without a doubt will be able to discern better than any physician could. Like I said, we are all playing the "working" game until 100-150 years from now at max.

So you think when a patient tells a computer "I'm tired a lot and can't concentrate" it'll just know...? Computers can't help with generating good inputs.

I do think computers can help or solve plenty of difficulties, just not this one you're talking about

 

[7331poas]

So you think when a patient tells a computer "I'm tired a lot and can't concentrate" it'll just know...? Computers can't help with generating good inputs.

I do think computers can help or solve plenty of difficulties, just not this one you're talking about

there will be algorithms which can search an entire database of similar patient reponses and reatime calculate the best possible medical outcome. Yes, doctors do guess work all. the.time. Computers can do the perfect thing every time based on actual evidence.

 

[efle]

What you're describing is the computer guessing as well based on population data. I agree that's the best thing to guess based on...but that's very different than the computer being able to "discern" anything the physician cant. In both cases you're left with a giant pile of possibilities and no way to know what it is. Computer would only tell you what things to start guessing first. As long as the input is ambiguous the response will be guesswork, AI or no.