Tyler Cowen — The #1 bottleneck to AI progress is humans

Economist Tyler Cowen told Dwarkesh Patel in this interview that AI is unlikely to push growth to 20 percent a year, because an economy grows only as fast as its constraints allow. Cowen's starting point is cost disease: some sectors speed up while others that use AI poorly, such as government, healthcare, education and nonprofits, hold the total back. Cowen guesses these make up about half the US economy and would take around 30 years to be displaced if they failed to adapt. When intelligence becomes abundant, other inputs such as regulation and energy supply become the binding limits.

Cowen's own forecast is about half a percentage point of extra growth per year. That would transform the world over 30 to 40 years yet be hard to notice in any single year, like a drug that reaches patients in 10 years instead of 20 but still faces clinical trials. Cowen cites ordinary-looking market data and growth economists such as Chad Jones, who is not predicting super rapid growth. These figures are Cowen's guesses. A separate Patel episode covers what stays scarce as AI automates work.