@zackwhittaker @epu it never occurred to me that my family doc may be recording our meetings without consent. Or with consent buried somwwhere in the intake process.

I'm definitely going to ask - and decline.

@zackwhittaker The privacy issues are real, and a much stronger argument, as dictation services frequently did make mistakes. I have memories of my dad complaining about their quality, or how long it took, both of which impacted patient care. It was also harder to fix a mistake when there was a longer delay between recording and discussing something with a patient and the actual charting being done a couple days later.

Pre-1996, the privacy of those was probably also bad, but weren’t networked.

@zackwhittaker There is a simple way to test the reliability of these AI transcriptions: keep a recording as well and then compare that to the physician's notes as a QA test using statistical sampling. It probably works best if the physician reviews the notes immediately after seeing a patient.

Also, HIPAA rules apply. If they leak patient data, there could be fines or criminal penalties.

@zackwhittaker my spouse opted out and the doctor did it anyway 😡

This is something very specialized so switching is not much of an option

@zackwhittaker I was watching this presentation about using AI to maximise doctor-patient time and it seems like such a good idea, I hope more models are being trained to higher precision: https://youtu.be/mYGp8EbRSFo

And I also encountered this collection of specialized models for identifying medical terms (diseases, medicine, chemical compound, etc.): https://openmed.life/

The focus should be on ensuring accuracy and privacy. If AI can increase doctor time, it will be a net benefit to all.