Tyler Cowen argues that 20% annual AI growth is unrealistic. He expects AI to boost productivity growth by 0.2–0.5% instead, which would still exceed the gains from the internet era. Regulatory actions and other non-technical factors could significantly shape outcomes.
Chapters
People do say, yes, okay, AI could increase GDP, increase wealth so much that we'll be able to pay our debt without any problems. You and I, while we're positive, optimistic on AI, neither of us think that we're gonna get into the Boomer scenario of ending scarcity.
20% growth a year. No, not gonna happen. Yeah. Yeah.
I would be very happy if we increase the rate of productivity growth by 0.5%.
Right, that's been my prediction as a default. But that would not solve the problem. That would make it sustainable. That's right on the knife edge, knife edge number that would make it all work.
Yeah, exactly. Yeah.
So if that's your modal forecast, yeah, don't buy those puts. Now, we'd agree there's a lot of variance on any AI forecast, right? Yes. It could even be regulated. Even if you're sure on the tech side, a lot's just going to happen. Say it's 0.2% productivity growth in the growth rate. That's a huge help.
It gets you almost half the way there to a coherent adjustment. Yeah.
And that's again reason to not be too pessimistic.
Yeah. Now, to be clear, a 0.5% increase in the rate of productivity growth, that doesn't seem like a lot, but that would be historically a bigger increase than we got from, you know, anything. I mean, bigger increase in the internet, right? Sure, yeah.
I mean, it is the internet in a way, but yes.
It was founded on the internet, yes. The internet was the agar culture for the growth of the AI.
That's why the internet's important. We're just beginning to realize this, right? Exactly, yeah, yeah. Yeah, so— It's why a lot of people can't admit AI might be a good thing, because then they'd have to admit the internet was a good thing. And they're so committed to never saying that.
