00:00:00
Sach got a contrarian belief for 2026.
00:00:02
>> Yes, AI will increase demand for
00:00:05
knowledge workers, not decrease it. I
00:00:07
would refer you to Erin Levy's post
00:00:09
called Jevans paradox for knowledge
00:00:11
workers. And the point of Jevans paradox
00:00:14
is that as the cost of a resource goes
00:00:17
down, the aggregate demand for it
00:00:19
actually increases because you discover
00:00:21
more and more use cases. So I think this
00:00:23
will certainly happen with code. In the
00:00:25
past, it's been very expensive to
00:00:27
generate code. You have to hire
00:00:28
engineers. there's not enough of them.
00:00:29
It's an expensive resource. So the
00:00:32
amount of software generated in the
00:00:33
economy was limited by that. I think
00:00:35
it's going to increase massively now
00:00:36
because the cost of generating code is
00:00:38
coming down so much. But there's other
00:00:40
examples too. You take a field like
00:00:41
radiology that's frequently cited as a
00:00:44
profession that AI is going to put out
00:00:45
of business. That's not what the data
00:00:47
shows. The data shows that the number of
00:00:48
radiologists is increasing. Why? Because
00:00:50
the number of scans that people want to
00:00:52
make is increasing. And it's true that
00:00:54
AI can do some of the work, but you
00:00:55
still need a doctor to prompt the AI, to
00:00:58
interpret the AI, to validate it. So you
00:01:00
get more efficient. The cost of scans
00:01:02
goes down and instead of it being a
00:01:03
super specialtity that happens very
00:01:05
rarely that you need a referral on top
00:01:07
of a referral to get, it becomes
00:01:09
something that's normalized and everyone
00:01:10
starts doing it and you start getting
00:01:12
more and more scams that leads to better
00:01:13
and better outcomes. I think there's
00:01:15
going to be a lot of those examples
00:01:16
through the economy and we're going to
00:01:17
look back and see that the job loss
00:01:20
narrative was not only wrong but we
00:01:22
actually got job gains.
