[extropy-chat] Re: Nano-assembler feasibility - Bradbury

[Chris Phoenix]

Robert J. Bradbury wrote (in two posts):
>Chris writes:
>>In a recent technology post I wrote: One of two events will happen.
>>Either we will built a mechanochemical fabricator, or we will discover a
>>significant error in the theory.
> 
> This is a very unproductive approach to the problem.  It assumes that
> nanotech requires a nanomechanical fabricator.  And that is clearly wrong.

Note I didn't say that nanotech requires a nanomechanical fabricator.  I 
said that the theory predicts a mechanochemical fabricator can be built. 
  I'm not claiming the fabricator is necessary, just that it's sufficient.

And I'm not sure which "nanotech" you mean.  I mean molecular 
manufacturing.  MM doesn't require diamondoid mechanochemistry, but that 
may be the easiest path.  MM does require programmability: the same 
machinery able to make multiple diverse products under computer control 
and deliberate design.  That's the main thing that distinguishes it from 
chemistry.  And that makes it fairly easy to implement 
self-manufacture--about which, more later.

> It assumes that diamondoid or sapphire cannot be fabricated by enzymes.
> Or more importantly that *any* material with a high covalent bond
> density cannot be fabricated by enzymes.

Diamond would be difficult or impossible.  Graphene looks quite doable. 
  (There's an enzyme, cytochrome P450-2K4, that breaks down buckyballs.) 
  And again, I'm saying fabricators are sufficient but not necessary.

> You have to get *off* the friggen fabricator wagon.  Look at DNA polymerase and
> the ribosome -- been there, done that. It is at the nanoscale and it *ain't*
> impossible.  So if you are going to have objections they have to be in the
> realm of "We can't fabricate that" (which is already a questionable claim
> because it is difficult to differentiate the difficult from the impossible.

The other objection is "How easily can you control it?"  Which is an 
engineering question.  I like all-digital control channels.  That's not 
to say analog is impossible.  But we understand it less.  So again, I 
think diamondoid machines may be the easiest to actually develop.  And 
may be the only thing that can do vacuum chemistry, which has various 
advantages for fabricating covalent lattices.

> Once one is in that land it becomes an economic question -- "can one
> manufacture it and make a profit?"  In the nanotech world its "oh yes
> we can make that but its going to cost you megabucks".  So the bottom line
> is whether you can afford megabucks and still make a profit?

In your nanotech world, it costs megabucks.  In mine, it costs 
millibucks.  Scientifically uninteresting, economically crucial.

> ASSERTION: It has been demonstrably proven that it is possible to manufacture
> complex chemical structures which contain a significant degree of covalent
> bonding at the nanoscale level to be strong (hell one has everything from
> tooth enamel to abalone shell to know this) are indeed possible.
> 
> CONCLUSION: There will not be any "significant" errors in the theory.

To me, the question of "Can you manufacture such structures with 
machines in vacuum" is very interesting.  I think that's what Brett has 
been trying to debunk, not the bigger question of whether covalent 
structures can be made by any chemistry.  And some of the theory 
asserting that possibility is untested.

> BTW: I have proposed the "impossible" in molecular fabrication --
> molecular chain mail.  It is likely to be significantly more difficult
> to manufacture than Fine Motion Controllers.  If someone could get
> it done by self-assembly then my hat would really be off to them
> (and I don't take my hat off lightly).

Seems doable by self-assembly, if you add a surface with a grid on it. 
You build up the rings in sections, from strings that will join 
end-to-end but will also match/attach to specific spots on surface and 
each other.  For two interlocked rings, you lay down string A, then 
string B that likes to cross string A, then string C that likes to cross 
string B and joins to the ends of A, then string D to close string B.

For chain mail, see the picture at 
http://www.mailleartisans.org/articles/pics/61994in109.jpg
There are four levels of crossings and two types of rings.  So you lay 
down chain A for the lowest fraction of the gold rings (toward the 
bottom of the picture) and P for the lowest fraction of the silver rings 
(toward the top of the picture).  Then B binding A (one end) and 
crossing P, and Q binding P (one end) and crossing A.  Then C binding B 
(both ends) and crossing Q to complete the gold ring, and R binding Q 
(both ends) and crossing B to complete the silver.

You could form the grid with DNA, attaching strands of varying length 
for binding and attracting side chains of the various lettered strings. 
  Some of the strings' side chains would bind to appropriate spots on 
the grid, and some to appropriate spots on each other.  You might have 
to anneal carefully and then cross-link after each step.  But I think 
it's doable.

> First and most important "self-replication is not an essential component
> of nanotech" (Robert Freitas may disagree with my opinion on this.)  It
> is helpful in getting the scales required for useful nanoscale production
> up to those levels required by humans -- but it is *NOT* essential.
> For example humans derive great benefits from manufacturing
> at the nanoscale level of 130-70nm chips that are produced by the millions.
> These chips are *NOT* self-replicating.  Humans also derive great benefit
> from the production of beer and wine that are based on self-replicating
> nanoscale machines that we have been using for thousands of years.
> So the only thing one can argue here is costs of production
> efficiencies and I don't notice anyone doing that.

Production efficiencies are crucial.  They're the difference between 
costly production (what's the cost per gram of silicon transistors?) and 
essentially unlimited production, where factories can sit idle most of 
the time and products cost the same as raw materials.  If we could build 
spacecraft for a few dollars a kilogram, would NASA have waited for 
years after the first scramjet blew up?  They could've run a test a week 
till they got it right, and we could have cargo-carrying scramjets today.

I agree we can build nanoscale structures without self-rep, and we can 
do self-rep biochemistry without programmability.  But put together 
self-rep and programmability, and you get a manufacturing revolution 
that makes the industrial revolution look trivial.

Drexler has been arguing costs of production all along.  After writing 
the NASA example above, I ran a Google on ["cost of" site:foresight.org] 
and found this in Unbounding the Future (1991): "At this point, the cost 
of materials and equipment for experiments will be trivial. No one today 
can afford to build Moon rockets on a hobby budget..."

> I agree with Chris that we do not (in any significant way) understand
> what the limits are on the phase space of parts construction or parts
> assembly.  We are therefore driving blind.  We only have some simple
> hints as to what might be possible from known biology.  

We don't know where the limits are, but we know or can guess where the 
limits aren't.  We know anything demonstrated by biology or biochemistry 
is possible.  In another area, a rich source of hints is the work on 
diamondoid structures and mechanochemistry.  Sorry, I know it annoys you 
that I keep bringing this up.  But I think it's more than a VHS/Beta 
choice; it's an analog/digital computer choice.  For some purposes, 
including much engineering, digital is fundamentally better.

> I would like however to see CRN produce a hardcore estimate of precisely
> how gray goo might be developed and the damage it might cause.  And
> more importantly how it might be defended against.  (Gray goo is *not*
> indefeatable -- but one does need however to be prepared for it.)

How it might be developed?  You mean a blueprint?  Or just a scenario? 
The scenario is simple: you give a bunch of engineering students a CAD 
program and a nanofactory, and let them compete with each other.  Gray 
goo is an obvious target.

Goo capabilities and defenses can barely be handwaved about at this 
point.  The chemical processing function might be roughly as complex as 
an oil refinery--or maybe as simple as a TDP plant.  The robotics might 
be approximated by the DARPA road race--or simple blundering about might 
be sufficient.  We can't know without a lot more information than we 
have today.

Defending against gray goo will depend on the gray goo design.  We know 
even less about that, because we don't even know the design space and 
the parameters.  Some versions of goo will be sterilizable with a mild 
dose of X rays or UV.  Some will be big enough to be easy to hunt down. 
  Small, redundantly-designed stuff would worry me quite a lot.  But how 
small can it be built?  We don't know.

Another question is detection, both to identify the problem and to clean 
it up.  What will we be able to do with non-proximal sub-wavelength 
imaging?  The ability to take noninvasive nanoscale census of 
cubic-centimeter volumes could make a huge difference, but it's too 
early to tell if that will be possible.

One thing I'm pretty sure of: If we get to the point where script 
kiddies are releasing modified versions for fun, we're in trouble.  It 
would take a lot of energy and disruption to scan for thousands of goo 
variants.  We do a lousy job with computer security, after two decades 
of practice.

But we don't have to worry about gray goo unless we manage to survive 
the unstable arms race.  That's what really worries me.  Weapons are a 
lot simpler than goo, and we already know enough about the performance 
of simple diamondoid products to say that cheap diamondoid manufacturing 
would completely (and rapidly!) revolutionize weapons and surveillance.

BTW, if bio-nano can't revolutionize weapons, that's a strong argument 
for treating diamondoid as a special case.

Chris

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