A Response to Nature's "Google AI has better bedside manner than human doctors — and makes better diagnoses"
A Response to Nature's "Google AI has better bedside manner than human doctors — and makes better diagnoses"
theluddite.org
The Luddite
12 comments
-
-
It’s refreshing to read something by an author with functional critical thinking skills.
-
I agree
-
-
Well human doctors have a history of coming into the appointment, ignoring all the things I written and the nurse/assistant wrote, ask me to repeat everything, then cut me off every sentence to give me a answer.
Many are also old boomer doctors who were taught black people have higher pain tolerance, and other inaccuracies about non-caucasians. I have some friends who did drug tests before a casual visit because they "fit a profile".
-
We’re actually trained not to just use what the roomer wrote. The reason is that what the receptionist writes and roomer writes can be inaccurate, and inaccuracies can multiply each time they’re transcribed.
For example, the call center might write “pain in testicle,” and then the roomer might write “lump in left testicle for 2 weeks” and then the patient tells me the lump has been in the right testicle for 3-4 weeks. If we just all copied the original note, we might be working with the wrong symptoms or wrong location. And asking questions assuming the notes are 100% accurate can lead a patient into giving us inaccurate answers, which is a much lower risk if we ask open-ended questions and let you fill them in. We do read the roomer’s notes, but our documentation is much better if we are getting the information directly from you rather than playing telephone.
As for cutting people off, I can’t speak for your individual doctors, other than to say there is a certain personality type who will answer every question (even yes/no questions) with a 1-2 minute meandering answer. And if we have 20 questions to get through, we simply can’t ask every patient for the rest of the day to wait an extra 20-40 minutes just to avoid cutting people off. If your doctor is doing that even when you’re giving a 1 sentence answer, though, you may need to look for a new one.
-
I'm a second year medical student, and those boomer doctors are what I will avoid being at any cost. I currently spend a fair amount of time and effort challenging some of my professors and classmates on their attitudes and assumptions regarding health needs and healthcare discrepancies for racial minorities, low SES patients, and LGBTQ+ patients.
-
Doctors are also terrible at keeping up to date. You can judge how out of date a doctor's cholesterol knowledge is:
- Cares about total cholesterol — 40 years out of date
- Tells you HDL is good, LDL is bad — up to 20 years out of date
- Tells you cholesterol isn't a problem unless free fatty acids are high — current
- Tells you LDL is protective, and higher is better (without high FFA) — current to this month
-
-
This paper does convincingly show that the LLM has a better botside manner, because of course it does
ayooo
-
It's also a puff piece, though it does contain the common and increasingly-comical self-flagellation at the end about how all these models are probably racist. Like all corporate DEI initiatives, everyone involved is only interested in admitting the problems in such a way as to avoid ever challenging the power structures that replicate them.
Ask me how much time I'm spending on (and how much they're paying me to sit through) DEI training at work. I mean, Im not against it, but if the CEO isn't gonna share the internal power structure, wtf is the point?
In functioning countries, healthcare is publicly owned, publicly operated, publicly accountable, and guaranteed to all residents. In other words, it's democratized. Compare that to living in a country where millions of people do not have access to medical care and where Google is actively working on a proprietary chatbot alternative to doctors,
(While calling it democratization of healthcare)
which gets a puff piece in the most prestigious scientific journal, and ponder what working in or writing about tech does to a motherfucker.
-
Wow, even Nature is becoming a corporate shill monkey. Seriously sad times we live in.
-
Every time I read AI is better than humans, I will reply: you are a retarded moron
12 comments
Slaps forehead ...aaand I think I've read enough to form a conclusion lol.
Repackaging corpo press releases is not journalism nor science. They should be ashamed to publish it in a journal next to actual science and discovery