Draft Paper: Santa's Workshop

From: Dani Eder (danielravennest@yahoo.com)
Date: Tue Mar 26 2002 - 10:40:31 MST

Below is a draft of a paper I am developing.
Comments are welcome.


Santa's Workshop

by Dani Eder
v0.2, 26 March 2002


An automated factory that can make copies of itself in
a short (~1 year) time and can switch to making useful
products would have major implications for mankind's
future. From a finite input to produce the first
factory, a growing and infinite stream of products
could be produced. NASA studied the feasibility of
such a factory in 1980, but the computer power to run
it was not available then. It is now. A start is
made at re-examining the concept and the issues it


NASA studied the idea of a self-reproducing factory
over 20 years ago (1) in the context of future space
missions. In that context, transportation is
expensive. Therefore it makes sense to send a 'seed
factory' to a destination such as the Moon or an
asteroid. The seed makes copies of itself, which then
continue to reproduce exponentially until you have a
desired number of factories. They then switch to
producing whatever product you wanted large quantities
of in the first place. The estimated computer power
to run the factory, ~2GB memory and ~30GB storage, was
way out of reach then, and the concept received no
further attention. Such computer power is well
within the reach of desktop PCs today, so it is time
to reconsider the concept.

The NASA study was limited in scope in several areas.
Only factories in space were considered due to NASA's
area of interest. If such a factory can be built in
space, it should also be possible to build one on
Earth. In fact, it should be easier since we already
know a lot more about designing machines to work here
on Earth. In addition to ignoring Earthbound
factories, the factory design in the study merely
produced copies of itself on the Moon to show the
feasibility of reproduction. The study did not
address what useful products the factory would
eventually make, nor did it address the idea of
starting with a simplified seed factory that uses a
subset of the machines and processes that later
generations of factories could use. The study assumed
for the sake of analysis production of 100% of it's
own parts and 100% automated operation. In a real
world system less than 100% values may make economic
sense, especially for a factory on Earth. The
limitations in the original study should be addressed
by future work.

The reproduction time of the factory on the Moon was
estimated at about 1 year. If a reproducing factory
can be built on Earth, it should have approximately
the same reproduction time: Less raw sunlight is
available on Earth for the solar panels producing
power, but the computer components will likely be
bought rather made, eliminating the whole electronics
manufacturing sector of the lunar factory. From an
economic standpoint, 100% compound growth is a
phenomenal return rate. After a factory has
reproduced enough times, mass quantities of goods can
be produced cheaply, hence the title of this article.
For example, a seed factory that runs on 100 KW of
solar panels is allowed to reproduce for 30 years.
You now have 1 billion factories with a power output
100,000 GW, roughly 50 times the total electric output
of the planet today.

There are a number of additional questions and issues
that should be addressed in future analysis of this

(A) Does a self-replicating factory have inherent
advantages over the industrial complex we have on
Earth with many specialized but physically distributed
factories? By co-locating the factory components you
can cut transportation costs and automate the transfer
betweeen components. On the other hand you lose the
benefits of resources (sunlight, ores) that are more
concentrated in certain locations on Earth than

(B) Is there a simpler subset of machines and
processes for the 'seed factory' that can then build
the rest of the machines in the later generation
factories? For example, could you start with a
milling machine, a mobile robot with an assortment of
hand-tool attachments, a solar panel, and a pile of
raw stock and proceed to make additional types robots,
machine tools, etc. What would be the optimal growth
strategy for such a seed factory?

(C) What is the optimum fraction of self-replication?
It may make sense to buy hard-to-make components (like
computer processors and hard disks) rather than making
them internally. Even a fully automated factory can
order parts over the internet and have them delivered
as needed, and sell and ship products as needed to pay
for it's purchases. The UPS guy may wonder, however,
why there are no cars in the parking lot. A related
issue is how reliable the equipment needs to be vs.
maintenance overhead and the need to complete >1 set
of new equipment before the original equipment wears
out to sustain reproduction and growth.

(D) What are the economic and social implications of
fast factory reproduction? If you own the first
factory, you could end up owning a major fraction of
the world's economic output. Of course, automation is
being introduced progressively into non-replicating
factories too.

(E) How do you analyze, design, and simulate the
production process? Like any mass-produced item, the
factories would benefit in cost from designing it
once, then producing many copies. Determining the
optimum amount of effort to spend on design may be a
challenge. If the factories grow in number faster
than the discount rate you are using for economic
analysis, then the present value of the future
factories is very large.

(F) Does it make sense start with a partially
reproducing and/or partially automated factory and
progressively improve it? How will such an evolving
system interact with the rest of the world’s economic
system, which is also evolving towards greater

(1) Freitas, Robert A. & Gilbreath, William P., eds.
"Advanced Automation for Space Missions: Proceedings
of the 1980 NASA/ASEE Summer Study", NASA Conference
Publication CP-2255, 1982. Available online at:

Do You Yahoo!?
Yahoo! Movies - coverage of the 74th Academy Awards®

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT