Should we build a database of those who abuse food and install facial recognition software at every fast food joint?
The more interesting question is how much money would it cost to develop an AI system whose diagnostic performance relative to GP's is similar to IBM's chess computers relative to good chess players? Leave specialists out for now, they come down the road.
I actually think it is possible to eventually get to the unlimited consumption of medical care, but not if the care is provided by one-one interactions with well compensated people.
There are already significant computer diagnostic assistances for MD's, and indeed my last visit was solved by that, leading me to suspect I could have done the same thing myself by just interacting with the computer and skipping both the Doc and the lab (the lab was chasing an incorrect hunch anyway). The Doc kept leaving the examining room and returning, and finally he took me to his office, and we stared at the search results and read them together. He then prescribed a drug, and never bothered to test for the actual condition (African tick fever, whereas we had been chasing and testing for suspected Malaria).
I suspect the diagnostic side is easily within grasp at the GP level anyway. The Big O aught to throw some of his stimulus wasting at a competition to see who can deliver, in one year or less, an AI system which surpasses the 95 percentile level of USA GP's at diagnostic accuracy, with for now, a NP or DA feeding the data to the AI system (the poking and prodding front end so to speak). The next step is to have the local community college offer an AS degree in Poking and Prodding so we can crank out humans doing what humans do better than computers (tactile sensing) and leave the intelligence portion to the computers (since they have an edge over humans there).
Surgery is a little harder, but robots are already doing that. At some point, the robots will do that with better quality as well, and at much lower prices. Some related fields should have died already, like Anesthesiology. Its hard to believe that an AI system and smart instrumentation can't exceed the 95'th percentile in that field, and it is one with high liability, so bringing consistently high performance to that arena would be a good development. |
