[extropy-chat] Re: Nano-assembler feasibility

[Hal Finney]

I agree with much of Brett's skepticism on this issue.  Given the fact
that so many prominent scientists continue to maintain that assemblers
are impossible, supporters of the concept need to take on the burden
of proof.  It's not enough any more to say, tell us why it won't work.

In the case of assemblers, we're missing the most important pieces:
the parts which do assembly.  We can invent bearings and struts,
cables and even works of art like the fine motion controller.  But I
have yet to see any proposal for how these pieces can be built.  I mean
specifically what tool-tips will be used, how they will be swapped out,
and what construction sequences will be used to build the components.

Now, there are two parts to this question.  The first is, how can an
assembler self-replicate; how can it build its own parts.  (Downplaying
self-rep is not relevant here; obviously the system composed of the
assembler plus supporting technology has to be able to replicate
in some way, otherwise we'll only have one assembler.)  The second,
even more difficult, is how can we bootstrap into an assembler from
today's technology; how can we build its parts without already having
an assembler.

I'm really just asking the first question, the easier one.  If you could
show an assembler design that was complete enough to give confidence that
the answer to the first question was positive, it would eliminate many of
the objections of the skeptics.  No longer would you have to use shaky
biological analogies or vague and confusing jargon to address Smalley's
fat-finger and sticky-finger issues.  You could just point to the design,
to the construction steps, and show that each part could be made, that
the fingers were not too fat or too sticky, and that it would work.

I think it's going to be important to answer that first question, as a
first step to motivate the efforts to answer the second (bootstrapping).

Now, I am somewhat familiar with work towards answering this question in
two areas.  The first goes towards constructing bulk diamond.  This looks
promising and, while the details need to be fleshed out, it looks to me
that tool tips have been designed which could plausibly work to produce
large pieces of diamond.

The problem is that this is not enough to produce all possible parts,
even among those that are composed solely of carbon and hydrogen.  One of
the issues in producing bulk diamond is which surface you are building on:
the different crystaline surfaces of diamond have different spacings of
the carbon atoms and the layers, as well as any terminating hydrogens.
Different strategies for construction are necessary for different
surfaces.  But these techniques will not be applicable to a small part,
because it does not have consistent surfaces with uniform atomic spacing.
Each atom in the small part is going to be a unique challenge and require
a customized assembly procedure.

Then there are the problems in gripping the parts, and in putting the
first two pieces together at the beginning of constructing a part.
These are where I think Smalley's finger problems are most likely to be
an issue.  I'd like to see a synthesis for diamond that didn't already
start with a big slab of diamond.  How will you start from nothing and
produce bulk diamond?  That would be a good first step towards addressing
problems involving gripping and releasing parts.

The other direction towards answering my question comes from one paper
by Merkle from 1997(!),
http://www.zyvex.com/nanotech/hydroCarbonMetabolism.html.  This is the
only paper I am familiar with that makes a real attempt to close the
loop and show how an assembler can construct its own parts.  This is
the kind of work I'd like to see expanded, but as far as I know no one
has developed it any further.

The reasons why this does not fully answer my questions are, first, that
the assembler being described is a very simplified and schematic one; and
second, that the reactions seem to have potential fat-finger problems.
They require four assembler arms (or equivalent) to come together in a
tiny space.  Yet you've never seen a diagram of an assembler from Drexler
which required four arms, have you?  If assemblers are going to require
this many arms, wouldn't it be nice to find out ahead of time?

I appreciate that these are all very hard problems.  Realistically, it
may be impossible to produce an assembler design sufficiently detailed to
answer them, without an enormous and well-funded effort.  So we are back
to the chicken and the egg, that we can't get started on a large project
until we know we will succeed, and we can't know we will succeed until
we get a detailed design, and we can't get a detailed design without
getting started on a large project.

This is why the alternative strategy is being tried, of claims that we
already have enough data to know that we will succeed.  But I think the
political reality today is that this strategy is not working.  The only
alternative is to take many years of time and effort to tackle these
problems one at a time, until we do get to the point where we have a
design that answers all the questions and eliminates the objections,
by construction.

Hal