The editors of the Washington Post tell us that we should be thankful for artificial intelligence in medicine:
If you’re struggling to come up with something you’re grateful for this Thanksgiving, here’s a development all feastgoers can celebrate regardless of their political leanings: Artificial intelligence is revolutionizing medicine, making health care more accurate and less expensive for everyone.
They provide a number of example. I found this one particularly interesting:
A recent study out of Boston comparing the performance of chatbot-assisted physicians in diagnosing patients with that of chatbots alone found that the bots performed considerably better. Given a patient’s case history and symptoms, the chatbot alone scored an average of 90 percent in correctly diagnosing their condition. Physicians using the technology scored only 76 percent on average — just marginally better than the 74 percent average for humans with no AI help at all.
This process will accelerate and advance faster than can be imagined. All of the incentives point that way.
There are many implications for that which I don’t believe have been fully appreciated, particularly in medical education. Who is selected to become a doctor and how they are trained will inevitably change. The traits that will make a physician effective in 2030 will be much different than those that made a physician effective in 2000.
Chatbots still cant take a history or do a physical. So if you give them a history with all of the pertinent facts they do well but if the history is not complete then not so much. Real live people dont give you textbook symptoms very often. Anyway, pretty sure I read the Boston study but dont remember the details. Do have to wonder if we are seeing the effects of online access and AI already. It’s been pretty clear for a few years that you dont really need to memorize everything like you had to do in the past so on tests where memorization is key people wont do as well as in the past. You can look it up if you know the general area you should be looking at. AI is already helping in some well circumscribed areas but I think people remain overly optimistic about the speed it is advancing.
Steve
AI is simply a glorified search engine that amalgamates the results instead of listing links. It has no way of knowing if the data is correct or not.
For an AI “doctor” to be reliable, it would need to use curated info, and it would still be a glorified search engine, but it would be faster than a human doctor.
I agree with @steve that human doctors interact with real human patients, and I can fully imagine that the average patient is less than ideal. A curated AI might be able to assist a human doctor, and it might be able to pre-screen patients.
I saw where an AI recipe finder has to be coded with keywords that are exempt from use in recipes. If the AI is not smart enough to know not to recommend using glue in a recipe, it ain’t very intelligent. Will the AI “doctor” have a similar requirement – nails are not an iron supplement.
AI does not think. It is limited to human knowledge. To actually think would mean that it could assume it’s knowledge is wrong. An AI trained on Newtonian physics could never develop Einsteinian physics.
“Chatbots still cant take a history or do a physical. So if you give them a history with all of the pertinent facts they do well but if the history is not complete then not so much.”
I don’t know that much about AI, although I have a geek stepson in law who reflects Tasty’s views.
But I have an awful lot of sympathy for steves comment. My father told me there is absolutely nothing like a full and accurate history, and laying your eyes and hands on the patient. Nothing. I’m not saying that AI can’t evaluate all kinds of probabilities and data bases at warp speed. But steves comment is really just “garbage in/garbage out.”
I guess time will tell.
“How long have you been having these symptoms”
How many times docs have asked me.
I honestly don’t know.
Patients sometimes lie, often don’t remember, have difficulty articulating the details, AI looks like it is OKAY but just another tool in the toolbox.
Electronic records are tough for me, they enable hospitals to disconnect the doctors from their patients.
Every day a different doctor quickly scans your records and is assumed then to be able to access your condition as well as the doctor who has known you for 30 years.
But on the other hand with a thousand other patients maybe the doctor you have known so long really has barely known you and scans your chart to remember even your name.