Re: What is Intelligence?

From: Hal Finney (hal@rain.org)
Date: Wed Oct 09 1996 - 10:04:33 MDT


I have enjoyed the debate about rates of growth in a situation
where computers play some part in augmenting the (effective?)
intelligence of researchers.

I had missed the first part, so I'm not sure whether the computers
help by allowing uploading and/or AI's, with possibly million-fold
speedups of thought compared to human brains? Or perhaps the change
is less dramatic, things like hyperlinks and search engines to
quickly find and disseminate research results, improved modelling
engines to decrease research costs, better CAD/CAM facilities to
speed up design cycles, etc.

Speaking very broadly, most growth cycles have had either an exponential
or "yeast curve" pattern. The yeast curve looks exponential at first, but
then slows down and hits a ceiling as some limiting factor is reached.
This raises the (unextropian!) possibility that growth curves which
today look exponential will also someday reach limits.

Exponential growth can be characterized simply by the percent increase
per year. I believe economic growth has run around a few percent per
year for many decades or even centuries. Computer performance has
increased more like 30 or 40 percent per year for the past few decades.

So a couple of questions which arise are, could general economic growth
rates move up to the rate of growth of computer performance, in case
the economy somehow became tied much more closely to computers?

And, will computer performance growth rates greatly increase because of
some feedback mechanisms related to the enhancements computers themselves
make possible?

If the first were true, this would in itself be utterly revolutionary
in its impact. It's difficult to even conceive what life would be like
in a world in which we would all be twice as rich, in real terms, every
couple of years. Even a decade of this kind of growth would completely
transform the world, like centuries of progress today. So it would be
interesting to hear opinions on this question.

The second question is more difficult to answer. Realize that the
dramatic rates of growth of computer technology already are made possible
in large part due to feedback mechanisms. Certainly the high performance
chips of today could not be designed or simulated without the computers
of a few years ago. So feedback is already present, and in fact it is
necessary to maintain these rapid doubling rates.

The problem with trying to forecast this into the future is that
there is also the factor of diminishing returns. All these forms of
economic growth benefit from feedback, but they face the problem that
it may become harder and harder to extract improvements from nature.
We pick the low hanging fruit first, and then the higher fruit takes
more effort. In fact there is no technology "just around the corner"
which will guarantee to maintain current computer growth rates into the
2010 time frame. Many people expect to see a leveling off of computer
performance before then as the fundamental technological limitations of
silicon chips are reached.

So, feedback by itself will not guarantee an increase in growth rates.
Rather, you have to look at details of how much the feedback helps, in
the face of whatever difficulties can be anticipated. If the world
economic growth rate did increase to a doubling time of two or three
years, then presumably computer performance growth rates would benefit,
as would virtually all other technologies. But translating this into
a specific increase in growth rates is going to be very difficult to
do.

Hal Finney
hal@rain.org



This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:35:47 MST