The prediction of the godfather of artificial intelligence turned out to be wrong

Jeffrey Hinton, the 76-year-old godfather of artificial intelligence who recently received the Nobel Prize; We know more about the warnings it issues.

After spending more than a decade at Google, he resigned last year to instead express his concerns about the rapid development of the technology and warn people about it.

Eight years ago, Hinton announced that radiologist training should be stopped immediately; The late 2010s were full of articles predicting the end of radiology, but today, eight years later, Hinton’s prediction is clearly unfulfilled.

Deep learning can’t do what a radiologist does, and we’re now facing the biggest radiologist shortage in history, with imaging at some centers being delayed for months.

This isn’t to say that Hinton was completely wrong about AI, but that the reality is much more nuanced.

It takes a lot of time to become a radiologist; After college, four years of medical school, one year of introductory general medicine, four more years of radiology residency, and then one to two years of subspecialty fellowship training are on the way. Such extensive training is due to the complexity of the work, which the expert must eventually complete without making any mistakes, and the prospect of being replaced by an algorithm is very uncomfortable; Radiologists have spent more than a decade studying to lose their jobs to computers?

Given the widespread interest in using machine learning in radiology, the concern is understandable; Of the approximately 1,000 medical devices approved by the US Food and Drug Administration that have artificial intelligence, at least 76% are designed for use in radiology.

Meanwhile, some envision a future in which artificial intelligence simplifies our jobs, maximizes efficiency, reduces errors, and reinforces the need for diagnostic radiologists. There are others who confirm Hinton’s view that radiologists are on the edge of a cliff; But they haven’t looked down yet.

However, as much as the absolute failure of artificial intelligence is stupid, replacing them with humans is also far from the mind; It looks like we’ll eventually chart a middle ground, and it’s not that AI will replace radiologists, but that radiologists who use AI will replace those who don’t.