[extropy-chat] Re: Self replication again (was MARS: Because

[Chris Phoenix]

Jeff Davis wrote:
 > Nano--as in MNT-- is even more primitive.  Mostly
 > theory--good theory, I'll grant, fun theory-- but
 > nevertheless theory of the most primitive sort.  As to
 > engineering capabilities,... so close to zero at this
 > point that well, speaking honestly, it's non-existent.
 >  Don't get me wrong.  I'm not puttin' it down.  GO
 > NANO!!

Seems to me that MNT is about at the same point as molecular 
electronics.  They've built a few transistors; we've joined a few 
molecules.  They have theory of how to design whole circuits once they 
learn how to synthesize them.  We have theory of how to design machines, 
likewise.

I wouldn't call either one non-existent.

 > For a long time I was hooked on the idea of self-rep
 > as the holy grail of ultra-productivity.  But recently
 > the Foresight people have stepped back from
 > self-rep--I think as a way of stuffing the grey goo
 > evil genie back in the bottle.  They've declared that
 > MNT can still deliver sans self-rep.

Just to be clear: self-rep is when a self-contained, autonomous device 
can build a complete functional copy of itself without external input. 
A system is "autoproductive" when it can be used by under direct 
control/intervention to build copies on demand.  A blacksmith's shop is 
autoproductive.  So, for example, the Merkle "assembler" is 
autoproductive but not self-replicating: it needs a continuous stream of 
sonic control/power pulses.  A tabletop nanofactory could build a 
physically complete copy, and in theory could build in the blueprints as 
well.  (The Merkle gizmo couldn't contain blueprints; it doesn't even 
have a computer.)  But the nanofactory copy wouldn't be functional 
because it wouldn't have any ability to connect up its feedstock pipes, 
much less manufacture its feedstock.

BTW, the idea that mechanochemical disassembly (turning complex 
molecules into feedstock) is a natural correlate of assembly (turning 
feedstock into complex molecules) is one of the most pervasive errors in 
popular understanding of MNT.

 > I thought it
 > over, and I agree.  Which also helped me to step back
 > from my own somewhat-rigid devotion to self-rep, and
 > to see self-rep as a "best case", an "upper bound", a
 > "theoretical ideal" in the overall range of values for
 > industrial productivity: caveman with rock at one end,
 > robot n builds robot n+1 at the other.  Everything in
 > between--all "real world" industrial
 > production--involves some mix of human and machine
 > input.

Actually, for human purposes, the sweet spot is in the middle.  At one 
extreme is a stick and a carving knife: you have to do everything 
yourself.  Then comes a CNC milling machine that you have to load and 
program, then handle the parts.

The sweet spot is a nanofactory that you download blueprints into--to 
build finished products, and (occasionally) more nanofactories.

Going beyond that toward the other extreme, you design robotics on the 
nanofactories so that they can move away from where they're manufactured 
and grab a waiting feedstock hose and connect it up.  This is already 
more trouble than it's worth.  Anything further, such as nanofactories 
with built-in chemical plants, nanofactories that can harvest 
agricultural material, and so on, is even more wasteful.

In short, the best molecular manufacturing system is a machine, not a 
component, and not a system.  Elegant, efficient, easy-to-design 
machines don't self-replicate--they don't self-anything.  This is why 
I'm not too worried about gray goo arising as a result of molecular 
manufacturing efforts: because it's hard to do, and in that context it's 
completely pointless.  (In other contexts, such as proving you're a l33t 
haxxor, it's not pointless--so when it gets easy enough for script 
kiddies or even MIT students to do, we'll still have to worry.)

Chris

-- 
Chris Phoenix                                  cphoenix at CRNano.org
Director of Research
Center for Responsible Nanotechnology          http://CRNano.org