The radiology stuff is just jevons paradox combined with liability issues. No one said the amount of radiology work would decrease, the issue is that the ai boosters didnt realize that healthcare providers number one priority is not getting sued so they cant get rid of the radiologists until the government says so. Radiologists at this point are effectively just rent seekers taking advantage of slow government regulatory updates while they sign off on work done by AI. That doesnt mean the future is bright for radiologists, at some point the government will give the green light to do radiology work without human involvement.
>> "No one said the amount of radiology work would decrease"
Literally the Godfather of AI did [1]. The rest is pablum, Radiologists do far more than just protect against liability. I hope you never find out why they are so critical to vast number of patients getting the worst news of their lifetime. As OP lays out, life and its professions is just not as simple as these narratives would like us to think, and we keep backfilling reasons for why all dooming and glooming consistently comes up wrong. Just selective memory writ large to keep the literal infinite scroll of doom going.
> "No one said the amount of radiology work would decrease"
I was talking about number of scans analyzed, not time. In fact, the decrease in the amount of time it takes to do radiology work is exactly why they are now paid more. Everything in american medical care is about profit, hospitals love this new paradigm because they get to charge the same amount and do 5x as much work per radiologist. Since incentives between patients and providers arent aligned we've entered this hellscape where providers are constantly ordering radiology work that is completely unnecessary because the MBAs are pushing high margin stuff to the entire organization. I personally have many stories from my 2 doctor friends where the business forces them to overprescribe in the stupidest ways. It sucks that the whole thing is a racket and that Hinton couldnt see that but I dont think that makes him wrong about the long term future of the profession.
Hinton said 10 years, what do you say? Could you maybe take from his prediction being wrong that you should practice some humility? Of all the pro-AI arguments people make, the "eventually it will happen" one is one of the strangest. Yes, and eventually the sun will blow up, so what the fuck are we talking about here? Most smart people ten years ago thought truck drivers, cab drivers, radiologists, and customer support would be obsolete by now, hasn't happened.
Where Hinton was right: AI has and will continue to significantly change the field of radiology including making radiologists more productive.
Where Hinton was wrong: He figured if each radiologist could do twice as much work we'd only need half as many and therefore there would be an oversupply of them. Instead what happened is we now do three times as much radiology. I went over the economic reasons for this in the other comment.
I dont think its correct to say Hinton was wrong on his core claim that AI will radically change radiology, he was just wrong about what that looked like.
He said that "People should stop training radiologists now" because there were enough radiologists already and the profession would be obsolete in the future. He said nothing about productivity or getting more work done. He said radiologists would be replaced. He was arrogant and didn't understand the scope of what radiologists actually do, which is more than just look at pixels.
It's obvious that deep learning did/would radically change the part of any job that required analysis of patterns or images. The core point of the OP is exactly the claim you're making, which is that you seem to think that all a radiologist does is either look at images or "sign off" on what AI has done after looking at the pixels.
