Let's assume the last person that entered their radiologist training started then and the training lasts 5 years. At the end of their training the year is 2021 and they are around 31. So that means they will practice medicine for cca 30 years which would put the calendar at around 2051. I'd wager in 25 years we'd get there so I think his opinion still has a large percentage of being correct.
People can't tell what they'll eat next sunday but they'll predict AGI and singualrity in 25 years. It's comfy because 25 years seems like a lot of time, it isn't.
Different, but nobody could predict what would happen next. We know now how different it was, but we didn’t know back then how it would be different now. There were people/companies who were right, and more people who weren’t. I had good predictions, and bad predictions. I didn’t understand why people didn’t use their phone already like how they use smartphones now. You could do everything what you can do now (except things which were discovered since then, mainly ML stuffs). Browse the internet (it was always interesting how people didn’t know what was WAP), listen to music, read books, play games, run random apps (there was waaaay more freedom regarding this back then by default, people just didn’t know how). But still, we needed smartphones. That was the thing which crossed the line for normies, and for most of them only more than 5 years after iPhone was released. My prediction of convergence would have failed without the modern smartphone, which I couldn’t foresee. It was pure luck. We needed a breakthrough.
That doesn’t mean that you can’t predict anything with high certainty. You just don’t know whether the status quo will be disturbed. And when you need a status quo disturbance for your prediction, you’re in pure luck category. When your prediction requires lack of status quo changes, then your prediction is safer. And of course sorter the term the better. When ChatGPT came out, Cursor and Claude Code could be predicted, I predicted them. Because no changes in status quo was required and it was a short term prediction. But if there would have been a new breakthrough, then those wouldn’t have been created. When they predicted fully self driving cars, or less people checking X-rays, you needed a status quo change: legal first, but in case of general, fully self driving cars, even technical breakthroughs. Good luck with that.
You cannot predict whether you'll be alive tomorrow, you can be 99.99999% sure, but never 100%. The point is that based on certain information/data you can form a personal opinion, and someone can form a different one on the same data. My opinion is that AI is going to keep booming due to the enormous amount of private and government cash being injected into it with both pacifistic and militaristic application of said AI resulting in it being capable of successful interpretation of radiology imaging in 20 years.
Let’s say we do manage to develop a model that can replace radiologists in 20 years. But we stop training them today. What happens 15 years from now when we don’t have nearly enough radiologists.
Why do we assume that radiologists would have literally 0% involvement in the radiology workflow?
I could see the assumption that one radiologist supervises a group of automated radiology machines (like a worker in an automated factory). Maybe assume that they'd be delegated to an auditing role. But that they'd go completely extinct? There's no evidence of, even historically, a service being consumed that has zero human intervention.
> the training lasts 5 years. At the end of their training the year is 2021
The training lasts 5 years, 2021 - 5 = 2016 If they stopped accepting people into the radiologist program but let people already in to finish, then you would stop having new radiologist in 2021.
Training is a lot longer than that in Québec, radiology is a specialty, so they must first do their 5 years in medicine, followed by a 5 year diagnostic radiology residency program. And it's frequently followed by a 2 years fellowship.
