Medical software companies have been working on things like this for years now. Still terrible. Don't let the flashy headlines fool you. They're all awful and rife with errors. If your hospital tells you your medical imaging is being read by AI, demand a board-certified radiologist read it before they release you or make any lasting treatment decisions. It's one thing for a radiology department to use it for triage and assistance (auto-measure lung nodules, aortic diameters, etc.), but actual interpretations are nowhere close to ready for prime time. It's one thing in outpatient settings where nearly everything is normal or near-normal. It's another thing when put in place at high-volume emergency rooms or cancer hospitals.
AI can't draw a human with 5 fingers consistently. Do you really want it making life and death decisions for you?
I did not say I want AI in healthcare today.
I did not say AI was ready to be in healthcare today.
Of course I get what you are saying. In the near term, you are right.
I feel the same way about an AI on a car driving itself. I say to myself that anyone who trusts FSD cars has a deathwish. In 10 years, it might fine. But today, I would not trust it. Today Tesla is rightly fighting FSD lawsuits, defining the standard and scope for FSD liability.
In time, tech moves forward… it advances… improves… is iterated… is perfected.
Implicit in my comment is when I said >> when the AI is actually good, useful & secure << meaning when it works as intended and has 99% success rate. Meaning it is not good today. Meaning goodness will come in the future. Not today.
That will not happen with Apple ios27 in a few months…. Not with GPT-10… Not in 5 years. But maybe in 20. That speaks to AI general functionality becoming acceptable/ predictable / reliable/ repeatable /insurable /affordable. In healthcare it may take longer.
But when the AI is good and proven, I look forward to seeing it.
I am older. I worked in the auto industry before moving to tech. There was a time when early anti-lock brake systems on cars were publicly decried as unsafe. The way they activated seriously frightened drivers and caused other unintended and sometimes deadly accidents, creating new liabilities for automakers. And their early fail rates were not anything to be proud of. But in time they were perfected and now they are unquestioned.
I don’t want today’s AI to serve me not even a cup of coffee. I don’t want AI making a life and death decision for me today.
But when the tech is perfected, I look forward to seeing it. When the tech can safely turn a 10 hour wait in the ER into a 3 hour wait because it makes an overworked MDs job easier… or because it improved back-of-house admin operations…. I look forward to seeing it.