AI and primary care: how to use it responsibly

AI in such mediums as Google, Chat GPT, and Claude is becoming a device through which doctors and patients can make diagnoses, prescribe treatments, and distill symptoms down to diseases.  If you are dizzy, what can be causing it, is it an emergency, how do you fix it?  What if you are having chest pain?  Are all your medicines necessary, can there be interactions?  In some ways AI can answer simple questions, but like most protocols and many doctors, it will typically suggest a conservative approach with more tests and visits, sometimes a trip to the hospital, because it cannot fully understand you as a person and the nuances of how your symptoms may not be diagnosable through the algorithms and the large studies it can access.  Primary care providers who know you well, who know the data well, who understand how to apply that data to you and how your symptoms may not fit in protocolized boxes, are far more likely to find the right diagnosis in conjunction with you.  That's something a computer or protocol or even physician-held calculator can't do.  This is why good primary care is not replaceable with AI.  We  run through several examples of this and show where AI is useful and where it can be lacking.