This wsj article (I Finally Have a Physician Who’s Available and Who Gets Me. Meet Dr. Grok) got me thinking

I think AI seems to be a good application to first line, or even post visit medical advice. The problem (as I see it) is a question of privacy and trust.

Medical doctors are regulated either by the industry, or by the government. When you go to a doctor, there is a certain level of trust that he is qualified. When you go to a generic AI, there is no trusted agency that puts their reputation on the line saying that the AI is qualified.

Privacy is the other issue with AI. I do not know what the privacy guarantees are for current AI systems. I do know that the mindset of AI companies is to scrape as much quality data into their training data to make the best AI possible. That is not the mindset that takes privacy seriously. Also we are at the early stages of this revolution so it is still the wild west approach to exploring the territory while we discover what we really want.

The last concern is how to create competition between independent AI systems to ensure the quality of the systems continues to improve. The medical term for this is we need to continue to have the ability to have a second opinion.

One possible solutions is that primary care facilities provide medical AI chat bots that ensure privacy the same way that a doctor visit ensures privacy. The conversations could be part of your medical record at the clinic to help the doctors and nurses understand you better for the clinic visits. Access to the AI could be annual subscription based, possibly with the fee waived when there is a clinic visit in that year. The office could ensure the quality of the service they are providing by doing some review of interactions from their patients.

There would have to be multiple service providers for medical practices, and the provider would have to be known to the patients. That way they can have a real choice for second opinions.