Paul Kedrosky
speaker
1,259 appearances
4 recordings
2 series
first heard Apr 2026
last heard 6d ago
Paul Kedrosky’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsRecordings per month over the last 12 months — 4 in all, peaking in Sep 2026 with 1.
Appearances
Someone who shall remain unnamed but has a popular newsletter and used to work at a certain venture fund put out the radiologist paradox, which was the idea that back in 2016, Jeffrey Hinton, computer scientist and Nobelist, early pioneer in image models and deep neural networks, had
said in a talk that within five years, if not 10 years, large language, deep at the time, neural networks, learning models would be better than radiologists, and there's really no reason to continue training them.
Now, of course, he said 10 years on the outside.
Well, it's now 10 years later, and if you look at the data, we're continuing to produce more radiologists, and that...
Analyst then put out a note yesterday, and so did I think KOTU or someone else, and said like, well, checkmate.
Jeffrey hinted, look, we have a lot more radiologists.
And of course, this is a classic example of a profound misunderstanding of so many things at once, it's hard to keep track.
One is that, again...
It's not clear that being selectively better than radiologists at certain things like identifying, I don't know, prostate cancers or whatever else, obviously that's not good enough.
Radiologists do more than that.
But it also misunderstands the nature of the employment market because radiologists, like most of medicine, has created a very comfortable little cartel for themselves.
So even if there was gale force winds blowing at radiologists because of AI, the likelihood of you seeing it in such a short time, even if Hinton was right that they could in theory replace a significant slice of what radiologists do, it's a misunderstanding of the nature of the markets themselves.
So it misunderstands both the technology and the nature of cartelized employment markets, these kinds of arguments.
And yet,
It's used as an example of how the inexorable march of these things continues apace and it will always be augmenting.
I just think there's so many nested misunderstandings of what pressures AI is having on employment markets and how we might see it, where it might show up, then to take it up a level to then do these calculations and say, oh look, I can now come up with a defensible measure of how the augmenting function is working and then incorporate that in GDP.
I kind of have to say bullshit.
No, you can't.
We're failing at the simple stuff.
Oh, absolutely.
Showing 741–760 of 1,259 · page 38 of 63
← Previous
Next →