Singularity: The Trendline Argument

Robin Hanson (hanson@econ.berkeley.edu)
Sun, 27 Sep 1998 12:39:08 -0700 (PDT)

John Clark writes:
>I don't agree with Robin, I think the singularity will happen, however
>if I was arguing Robin's case I know what I'd say, change is slowing
>down. Even in technologically advanced parts of the world the change in
>daily life between 1899 and 1949 was greater than between 1949 and 1999.

To clarify, I haven't said the singularity won't happen. I've said its proponents haven't clarified what it is or offered a coherent argument in favor of it. Hal mentions my pro-singularity argument:

Hal Finney writes:
>Actually if we just look at the empirical data, Robin has done some curve
>fitting which does suggest a Singularity some time in the next century.
>At http://hanson.berkeley.edu/longgrow.html he shows how historically the
>time it takes for the world economy to double has decreased in several
>relatively discrete steps over the course of history. Extrapolating
>forward suggests that we might transition to a doubling time of two years
>sometime in the next few decades, with a potentially even more dramatic
>transformation a few years later.
>Of course without a model for why such changes should occur, it is hard to
>give them much credence. Robin noted elsewhere that the two year doubling
>time is in rough accord with Moore's law, suggesting that perhaps it could
>be interpreted as a transition to an economy dominated by computing power.

I consider this trendline argument to be the strongest case I've heard for something like a "singularity." I give it substantial credence, but it leaves a lot of uncertainty. A theoretical model of this next big change would add a lot more confidence, were it available.

My best shot so far is to say that growth is now limited by the rate at which labor populations can grow, a limit that will be eliminated with uploads or AIs. Growth would then be limited by the rate at which labor can be made more productive, which would be the computer hardware improvement rate if that was the main way labor then became more productive. It might be worth trying to formalize this argument within the usual economic growth models.