[ExI] von Neuman machines

Keith Henson hkeithhenson at gmail.com
Fri Aug 26 13:28:18 UTC 2016

On Thu, Aug 25, 2016 at 11:05 PM, Adrian Tymes <atymes at gmail.com> wrote:
> On Thu, Aug 25, 2016 at 9:27 PM, Keith Henson <hkeithhenson at gmail.com>
> wrote:
>> Question I have is:  Do any of you know of projects beyond what the
>> Wikipedia article mentions?
>> Alternately, do any of you have insight into solving the problems?  I
>> am particularly interested in proposals to take an asteroid, say a
>> carbonaceous chondrite and make it into useful materials to feed to
>> the replicator.
> One way to get from here to there is not to try to make the first iteration
> a universal replicator.  Think instead, self-replicating machine shop.

This obviously works.  Herman Kahn wrote extensively about this in the
context of recovery from thermonuclear war.  He pointed out that the
"B" country (industrial facilities that were not blasted) could
recreate the "A" country in around 20 years, because that had already

The trouble is, if you talk about a 20 year replication time, a lot of
the problems may go away, but so does the motivation.

> What tools can make more of themselves, if assisted by suitable macro-scale
> manipulators (as in "not a way to sneak in nanotech") and fed smelted
> regolith?  I hear a plasma welder could make another plasma welder.

Starting from what materials?

> Once you have that set, what can make more of themselves and said
> manipulators?
> ...and the control circuitry and power systems?
> ...and machinery to gather and smelt regolith?

Regolith is not like iron ore.  Even if it was, you don't have the
carbon and free oxygen to power the smelting.  Nor can you scale down
processes that involve a lot of heat.

> Rather than focusing on one tool to do it all, think of it as an industrial,
> mostly automated city expanding from a seed.  There will be many things it
> can not fabricate, at least at first.  But if you have a limited-capability
> factory build another such factory, then you have two factories, and I'm
> sure you're familiar with the math from there.

The critical parameter is replication time.  A lot of the steps like
chemical sorting out elements are just slow.

> By the time you're churning out square kilometers of solar panels per day,
> you're probably generating enough power that you can beam some of it back to
> Earth.

Beaming energy from the moon is not very practical.  If you put the
power satellites in GEO, then you have the problem of getting the
parts off the moon and down to GEO.

Not trying to be a wet blanket on the subject, but it is way harder
than it looks.


> If the only cost is that seed (which, if you can get to no more than
> 5 kg, there are ways to get to the Moon for under $1M), time, and a
> rectenna, you can sell the electricity incredibly cheaply.
> Of course, that time factor is critical: you'll need something that can
> expand quickly.  Being automated so as to go all day every day helps; so
> does starting out heavier (if you need X kg, starting at 10 kg instead of 5
> is one less doubling), but that runs into diminishing returns (cost
> increases more or less linearly, while doublings needed to get to a certain
> point decrease with the log).
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat

More information about the extropy-chat mailing list