Nick Bostrom:7/7/98

Robin Hanson's paper gives an analysis of the conditions for an
economic singularity. In a few brief sections, important and
non-obvious relations between the key concepts related to explosive
economic growth are described. Everybody who is interested in "the
singularity" will benefit from understanding what economic theory can
say about the issue.

I have some doubt about the empirical part of the paper. Robin
writes: "The historical increase in the savings fraction has been
roughly constant with time for the last two centuries, suggesting a
near 100% savings fraction near the year 2150." Extrapolations of this
sort are of course very precarious, and Robin does not claim
otherwise. Yet, even as extrapolations go, this one seems especially
problematic. The historical data determining the model parameters
contains only three data points, if we bracket the one for 1750 which
Robin says is "very crude". It would be interesting to know whether
there is more data available that fits the curve. And what are the
error margins? How robust is the 2150 estimate to variations in the
parameter values?

I'm also wondering about whether there might be an alternative
explanations for the increasing savings rates. Rather than saying that
people save a greater fraction of their capital because the rate of
return on savings has gone up, is it possible that people save more
because the risk factor has decreased? A person considering saving some
of his money in 1750 might have faced a greater risk that his savings
would go up in smoke due to some political upheaval, be expropriated by
a monarch, or that he himself would get killed in a plague before he
got a chance to harvest the payoff. Even if the average rate of return
had been constant between then and now, might one not have conjectured
that risk-averse individuals would be less inclined to invest under
such circumstances, especially since institutions that could insure
agains unexpected losses were presumably less developed in those days?

An interesting question to consider is: What possible technologies
would have the properties that could make for an economic singularity?
My own view is that when we get both superintelligence and
nanotechnology (and I think one would quickly lead to the other) then
that will cause a singularity. This technological combination certainly
has the exremely wide range of applicability that Robin lists as a
precondition for a singularity. Whether it would benefit or harm
non-investors is more difficult to predict, since the social
ramifications could well be so dramatic that Robin's economic model
would no longer be valid. For example: It might not be possible to
enforce property rights; preferences and living conditions might change
so radically that comparisons between the amount of capital in the
world before and after the singularity will no longer be meaningful;
there might not even be a multitude of competing economical agents
after the singularity.

Nick Bostrom
Department of Philosophy, Logic and Scientific Method
London School of Economics
n.bostrom@lse.ac.uk
http://www.hedweb.com/nickb

Robin Hanson replies:7/7/98

--- I thank Nick Bostrom for his thoughtful review.

I do not dispute Nick's assessment that my empirical extrapolations
seem "especially problematic". I offered them mainly to give readers
some sense of where we have been within the model I describe. I don't
really expect finer data to continue to support a linear relation of
the savings fraction with time. Instead, I have recently been working
on a simple sum of exponentials model of long term economic growth,
using a data series by Brad DeLong (Estimating World GDP, OneMillion B.C. - Present). A rough draft is available.

Nick wonders whether "people save more because the risk factor has
decreased." This certainly makes sense, and is obviously one of many
issues I left out of my simple model. It turns out that there is a
long-standing puzzle in finance regarding pricing for risk; investors
act as if the risks of strong downturns are much larger than they
appear in the recent historical record. So you'd really need
historical data on percieved risks to examine this empirically.

I agree with Nick that superintelligence and nanotechnology are
technologies with potentially very wide applicability. But I'm more
skeptical about how fast early breakthroughs in either field will lead
to more advanced breakthroughs. I'm thus relatively confident that we
would retain "a multitude of competing economical agents," and I'm not
convinced that correcting for technical change in comparing capital
amounts will get much harder.

A final note: a dramatic loss in ability to enforce property rights
doesn't actually invalidate the model, though it might dramatically
change some parameters.

Robin Hanson
hanson@econ.berkeley.edu http://hanson.berkeley.edu/
RWJF Health Policy Scholar, Sch. of Public Health 510-643-1884 140
Warren Hall, UC Berkeley, CA 94720-7360 FAX: 510-643-2627

Kathryn Aegis:7/7/98

Once again, Robin Hanson has found a connection between the principles
of economics and the goals of transhumanism. It sparks thinking in a
new direction and provides a potential avenue of practical application
by an entrepeneur or investor.

A question that raised in my mind as I read Robin's paper relates to a
recent discussion with Kurt Schuler regarding future alternatives to
our present day parochial monetary currencies and the explosive growth
in investment that could result. Could the digital technologies
(encryption, ubiquitious exchange, instant transfer) of alternative
monetary regimes represent an example of the very technology that Robin
references?

Kathryn Aegis, aegis@igc.apc.org

Hanson replies:8/7/98

Kathryn Aegis asks if "digital technologies ... of alternative monetary
regimes represent an example" of technologies which could induce
explosive economic growth. My intuition would be that by themselves
such technologies are far from sufficient. You might have a better
case if you added in lots of new kinds of markets that such digital
money might be used in. But most new markets are blocked due to
regulatory reasons, not because of poor digital money. I'm not very
confident of my intuitions here, however, and could be persuaded to
change my mind by someone like Lawrence H. White.

Billy Brown:26/2/99

Robin Hanson paints an interesting picture of the relationships between
the factors that underlie economic growth, and I certainly would not
argue his conclusions on economic grounds. I would, however, suggest
that his results could easily be misinterpreted when one attempts to
apply them to the real world.

The problem is that in the kind of future many transhumans expect to
see, economic growth is a poor proxy for human benefit. Consider, for
example, the parallels between electronic computers and molecular
manufacturing:

Computers have been undergoing an exponential improvement in
price/performance ratios for some decades now. Enthusiasts like to
point out that if cars had improved at the same rate over the last
twenty years, the average vehicle would cost a few pennies, travel at
supersonic speeds, and be capable of running for years without
refueling. Nevertheless, as Hanson points out, the economic effects
have been relatively modest. Computers have made some companies rich
and others poor, and have on the whole been beneficial, but they have
not turned us all into millionaires.

Now, molecular manufacturing promises to bring about the same kind of
change in most manufacturing industries. This implies that most
material goods will undergo a period of extremely rapid innovation,
with costs collapsing as capabilities rapidly improve. Hanson's model
predicts, probably correctly, that the net effect on economic growth
will again be much smaller than we expect. But there is a crucial
difference: after a few decades of such progress our car really will
cost only a few pennies, and so will all other manufactured goods.

What this means is that economic growth will become much less relevant
as a means of measuring human prosperity, at least by our current
standards. If prices for most material goods collapse while measures
such as GNP show modest growth, the practical result is a vast
improvement in the human condition.

Billy Brown, MCSE+I
bbrown@conemsco.com

Hanson replies: 26/2/99

While Billy Brown "would not argue [with my] conclusions on economic
grounds," he cautions against applying them to the real world because
"economic growth is a poor proxy for human benefit." Why? Currently
the "economic effects" of rapidly falling computer hardware prices
have been "relatively modest." By analogy, Mr. Brown presumes that
with molecular manufacturing (i.e., nanotech) the "prices for most
material goods [would] collapse while measures such as GNP show[ed]
modest growth." Since "the practical result [would be] a vast improvement
in the human condition," he concludes "economic growth will become
much less relevant as a means of measuring human prosperity."

Mr. Brown, there is no such thing as a non-economic human benefit.
If the participants in some social process perceive a type of benefit,
economists consider that benefit type fair game for economic analysis.

Now published statistics like GNP do neglect many types of human
benefit. They would not, however, fail to notice a rapid fall in the
price of most material goods such as cars. Such a fall would certainly
show up as rapid GDP growth. How fast would car prices fall with
nanotech? That is exactly the question at issue here, isn't it? The
arguments of fast growth skeptics are not refuted by the fast growth
claims of nanotech optimists, though of course skeptics might be refuted
by more detailed economic analyses suggesting fast nanotech growth.

Robin Hanson, hanson@econ.berkeley.edu

Billy Brown:11/3/99

After re-reading Robin Hanson's paper, exchanging a bit of private e-mail,
and thinking about the issue, I've come to conclude that our disagreement is
a result of one of those definition problems that always crop up when you
have people from different fields debating a complex point. It seems to me
that what he means by "economic growth" would by definition include any sort
of human benefit, which of course invalidates my claim. What I meant by
"economic growth" was something more like "the Federal government's official
GNP figures", which is another matter entirely.

I must therefore concede that his model is not subject to the sorts of
problems I suggested. Anything that has any real effect on human welfare
should show up as economic growth, and I don't see any reason to contest his
conclusion about the conditions a technology must meet to cause economic
growth. Whether nanotechnology can actually meet those conditions is a
complex question best addressed elsewhere.

So, that reduces most of my comments to a complaint about the inaccuracy of
current measurement methods, which isn't especially relevant to the paper.
That being the case, I think that it is time for me to retire from the
field.

Billy Brown, MCSE+I
bbrown@conemsco.com

Hanson replies: 12/3/99

I'm happy that Mr. Brown and I were able to work out our differences via
a private conversation, and I'm sorry that I didn't take the discussion
private from the very start.

Economists are well aware of the problems with government statistics.
The problem is that allowing government agencies more leeway in
including "squishy" harder to measure factors in their estimates also
allows more opportunities for corruption in constructing estimates.
Privately produced estimates can and do include squishy estimates, but
their quality is limited by the fact that much less money goes into
producing private estimates.