OpenAI’s new Epic integration is read-only.
That may be the most interesting part of the announcement.
Clinical AI is becoming good enough to understand the chart.
The harder question is whether we trust it to change the chart.
I think healthcare AI will move through a permission ladder:
Read → Recommend → Draft → Act with approval → Act autonomously
The companies that win clinical AI may not be the ones with the smartest models.
They may be the ones physicians trust with the most permission.
Tesla’s Cybercab launch and GPT-6 Astra made me think about the same question in healthcare:
How much autonomy should we actually give AI?
Clinical AI probably won’t jump from scribes to autonomous doctors overnight.
It may evolve in levels:
L1: document
L2: recommend
L3: act with physician confirmation
L4: own bounded clinical workflows
L5: fully autonomous care
The next race in clinical AI may not be about who has the smartest model.
It may be about who gets the autonomy boundary right.
Most clinical AI starts with:
“How do we automate the note?”
I think that’s too narrow.
The bigger question is: what should happen after the clinical encounter?
Notes are one output.
Tasks, follow-ups, orders and next steps are the workflow.
I’ve spent the past few months speaking with physicians and watching how they actually use AI in clinical workflows.
One thing is becoming increasingly clear to me:
Clinical AI is not mainly an accuracy problem. It’s a workflow problem.
A model can be impressive, fast, and technically strong, but if it adds friction to a doctor’s day, it still fails.
That idea has shaped how I’m building Asternoos.
Excited to share our latest systematic review with Professor J. Fernando Arevalo from the Wilmer Eye Institute on the "Risk of Stroke Development Following Retinal Vein Occlusion."
https://t.co/oZnTqv7ku3
#MedicalResearch#Ophthalmology#StrokePrevention#WilmerEyeInstitute
@_Sunandra_ و فکر میکنم ادمایی مث تو باعث میشه که این چرخه کمک به دیگران،بزرگ و بزرگتر بشه،هرچند گاها چند مورد نمک نشناس هم اون وسط پیدا میشه،ولی خب در مقیاس بزرگ میتونه ی چرخه خوبی تشکیل بده
@_Sunandra_ بزار یه چیزی بت بگم.من بعد اینکه برا اولین بار دوسال پیش باهات صحبت کردم و تو بدون هیچ چشم داشتی وبدون اینکه منو اصلا بشناسی،کلی منو راهنمایی کردی،با این کارت خیلی حال کردم و تصمیم گرفتم که من هم به بقیه بدون هیچ انتظاری کمک کنم واین حس خیلی خوبی به ادم میده
Friends in Iran who are MDs or med students are eager to collaborate with labs in Western countries for research remotely. If you need researchers or can offer help, drop a comment below with preferred skills or requirements! Let's work together to advance science
#RetweeetPlease
To students on #MedTwitter - if you're interested in #Ophthalmology, check out our summer research program @UCSDMedSchool funded by @NatEyeInstitute! Applications close on Jan 15, 2024. More info available here: https://t.co/hQ0aiBr4An