Re: Economic Growth Assuming Machine Intelligence

Robin Hanson (
Mon, 05 Oct 1998 09:27:17 -0700

Thank you Doug Bailey for some thoughtful comments!

Doug Bailey writes:
>I noticed you used a neo-classical growth model and exogenous growth
>functions as the basis for your analysis. Might endogenous growth
>theory be a better fit for the AI scenario? Or at least wouldn't it
>be interesting to see what predictions an endogenous model might
>make? From what I remember of endogenous growth theory, the way it
>models growth seems to match better an economy where AI was

It is easier to get more variation in growth rates with endogenous growth models, and there is also less consensus on how to model endogenous growth. So to avoid the accusation that I cooked my big growth speedup implications, I choose to use exogenous growth. This may provide a conservative estimate of growth implications. But yes, it would be interesting to try the simplest endogenous growth model, which I think is the AK model.

>Another area that might warrant further review is the neo-classical
>idea of diminishing returns. I know Paul Romer alleged that the
>diminishing returns problem might not apply if capital was expanded
>to include human capital.

This isn't another area. The essense of endogenous growth models is that they eliminate diminishing returns.

>What impact might a drastically increased
>efficiency in capital utilization have on growth models? How would this
>potential effect of AI-development be factored in to growth models?

Unless you forsee particular reasons to expect a change in the *rate* at which such utilization becomes more efficient, this effect is already included in the standard models of exogenous growth.

Robin Hanson RWJF Health Policy Scholar, Sch. of Public Health 510-643-1884 140 Warren Hall, UC Berkeley, CA 94720-7360 FAX: 510-643-8614