3 ms·
The original article referred to by the blog post is here from last week-- https://qz.com/2016153/ai-promised-to-revolutionize-radiology-but-so-far-its-failing/
by throwthere 5y ago
The original article referred to by the blog post is here from last week-- https://qz.com/2016153/ai-promised-to-revolutionize-radiology-but-so-far-its-failing/ https://qz.com/2016153/ai-promised-to-revolutionize-radiolog... .
The conclusion is that AI will revolutionize radiology. It's just that nobody knows when. And it's not like there's some socioeconomic or whatever barrier preventing AI from being used (as an aside, there are barriers of course)-- it's simply that AI isn't good enough yet.
It's not a surprise to anyone who relies on radiologists and has reviewed the current AI state of the art. Yes, with machine X on patients meeting criteria Y, you can rule out specific disease Z. But the algorithms don't generalize very well. It's like declaring you'll have self-driving cars in 5 years because you can drive one straight on a highway in sunny Arizona that only occasionally causes fatal crashes.