I have some insight into this as I’ve worked a decade in the public sector of Denmark, and I don’t think the general role of doctors will change that much.<p>I don’t see ML have much, if any, impact on a doctors role in diagnosis. I know this is a little strange to hear to a lot of techies, but we once had IBMs Watson work on our data to see what IBM and Watson could come up with, and while the output wasn’t bad, it was sort of useless because we already had 50 years with of analytical models on virtually everything. So in essence what Watson did with our data, was to generate a lot of BI that was inferior to the BI we already had. Diagnosis is sort of the same story because what doctors do is basically to refer your symptoms directly with the medical lexicon and your history. ML is likely going to help collect the data and tie it together, to make diagnosis both better and faster, but you’re likely still going to have a doctor review the results, just like you would your “data scientists” on any other important ML data. Where ML will change things is in early detection, where your medical data from every source will be crossed and checked much better because that’s the sort of things ML is good at. So maybe you’re going in for a blood test for something completely unrelated to cancer, but because the tech has improved, it’ll also screen you for cancer and alert your doctor if your numbers are off.<p>Where we’ll see the biggest impact will be in areas that can be turned robotic. Things like the lab work on your blood tests, which is currently still a very manual process. Or in surgery where precision machinery will slowly take over a lot of the cutting. You’re likely still going to need surgeons to monitor the process, not so much the actual surgery but the planning and the recovery, because these things are impacted by so many real world factors that we may never get ML models that are good enough to handle them on their own.<p>It’s often in the places you may least expect it that IT makes the biggest difference. The biggest impact I saw in the medical system was automated medicine distribution and in wound cleaning. The medicine distribution “used” to be handled by nurses putting the pills for patients into boxes and then healthcare workers administering it the right time. The automated way was having the pharmacy distribute smart-dispensers that would alert citizens to take their medicines and then sort of “punch” the right pills into a tray when the patient clicked on it. I put “used” in “ because the automated smart system is more expensive than the old way, and this resulted in many places still opting for the old way, despite it being more error prone. The wound cleaning is an AR/ML success story. Basically wounds can be really nasty, and contain nasty things that even the best nurses in our system won’t spot as well as ML. So what we did was hack a pair of Google glass to never send data to Google (I believe Google was very helpful in this process by the way) but instead feed images of the wound to ML recognition and then alert the nurse to areas in the wound where the nurse had missed a spot of nasty. Really awesome stuff.<p>In general I expect that medical, along with farming, tech will be some of the most interesting tech areas in the next few decades, but I don’t expect either to replace doctors or nurses. I expect to see it make doctors and nurses better at their jobs. It will free up some tasks, and change which doctor professions get the highest pay because a neurosurgeon won’t be the Hollywood rockstar, but really, they kind of aren’t outside of Hollywood anyway, at least no in Denmark.