LogFAQs > #884490057

Transcendentia posted...

surgeons and doctors are so highly paid and hard to train that there's huge profit incentive for their roles to be automated. and technology is finally getting to the point where it's technically possible to automate those roles. so we'll see autonomous surgery and autonomous doctors much sooner than we all realize.

Doubt it. People always talk about deep learning AI with regards to things like radiology. Keeping up with some of the articles, even the simplest modality in radiology, a simple chest x-ray, has a sensitive and specificity that is pretty atrocious with these AI machines.

See, it's the combination of those two statistical variables, sensitivity AND specificity that's the really tricky part with automation. The more sensitive you make your machine, the less specific it becomes, and vice versa.

I could make a machine that has "100% accuracy in diagnosing all types of cancer"! It just gives everyone a cancer diagnosis and my specificity is atrocious (mostly false positives).

So when you factor in modalities that are even more complicated than a chest x-ray, we're still a ways away. I haven't even entered the whole liability aspect. The first nodule or pneumothorax that goes undiagnosed by the machine is gonna get sued like crazy.

So most likely, the technology will be used as an adjunct, but physicians will always double check the machine's work. EKG machines already give predictions and diagnoses, but they're wrong about half the time and can't factor in context (which is key). Would you trust a machine that had a coin flip chance at determining whether or not you're having a heart attack? How accurate does the machine need to be before it's acceptable if you were the patient?

In other industries, making mistakes is a bit more acceptable. When a machine's error is the difference between life and death, there is a Grand Canyon's worth of a distance the technology needs to cross before it's widely implemented like people say.

I don't doubt the technology is getting impressive. But TC's image rings true with what I've been following. Doctors and surgeons won't be replaced for a while. A lot of other industries will go by the wayside much sooner. Sure, the AI engineers and technicians will out live us. No arguments there.
---

"He has two neurons held together by a spirochete."