[extropy-chat] Why no assembler design?

[Hal Finney]

Eugen is probably right that we don't presently know enough about the
physics of chemical reactions to be able to design a working assembler.
Most chemical reactions happen very quickly, with an incoming molecule
moving fast enough to overcome the potential barriers; while the nanotech
proposals I have seen rely instead on relatively slow motion and firm
pressure to induce the reaction.  It's something like the difference
between hitting something with a hammer versus squeezing it in a press.

When I posed my question I didn't mean to ask why we don't have
an implementable plan that leads from today's technology to a
self reproducing assembler; but rather, why there are not even
any solutions to the simpler (because less constrained) problem of
designing a self-reproducing assembler without concern for how it would
be bootstrapped.  Nanosystems was published in 1992, and it seemingly
goes at least 50% of the way towards such a design.  Why, in the ten
years since, hasn't anyone taken us the rest of the way?

Nanosystems has plans for component parts, like bearings, rods, cables,
and structural supports.  It includes some proposals for chemical tool
tips sufficient to build bulk diamondoid.  And it has a sketch of a
design for a robotic arm with an internal channel to allow for changeable
tool tips.

So what is missing?  I see two main areas.  The first is that the
discussion of tool tips and synthesis is sketchy and incomplete.  Most of
the parts shown in the book could not be built using the described
methods, which are adapted to bulk diamond.  Many of the bearings,
cables, etc. are small and relatively self contained structures where
the bulk techniques would not seem to work.

Glancing over recent Foresight Conference papers, it appears that
simulations of potential tool tips and mechanosynthesis reactions are
still in their infancy.  http://www.rfreitas.com/Nano/DimerTool.htm is a
paper by Merkle and Freitas published just a couple of months ago that
is a very basic analysis of a possible tool for depositing a pair of
carbon atoms.  It looks at how stable the tool is when it's just sitting
there; not at how the tool is built, or whether it would work to build
some diamond structures.  This is, on the surface, a step backwards from
the level of performance analyzed in Nanosystems.  IMO this demonstrates
that Nanosystems was, to some extent, overconfident in its presentation
of how these synthetic tools could work.  (Frietas and Merkle have a
book coming out next year analyzing diamond synthesis in more detail.)

Second, there are a number of architectural issues that have to be
resolved.  How many manipulators does the assembler have?  How do they
grip and hold the work piece?  How do you arrange that you can always
get to the reaction sites with the needed angles, even as the work piece
grows to be the size of the assembler itself, or larger?

Are the components, the cables and bearings and such, manufactured
separately and then attached to the work piece somehow, or are they
built in place as the output product grows?  What is the assembler's
environment, how are the raw materials provided to it, as well as power
and control signals?  Is the assembler free floating or attached to a
large solid surface?

These questions can be multiplied ad infinitum.  And some of them look
pretty tough.  Holding onto a growing and complexly shaped object while
continuing to perform reactions at a variety of angles and pressures
won't be easy.

Overall it looks to me like there are a number of significant gray
areas where we don't know enough to create a full design for a self-
replicating assembler.  In many ways it seems that the Nanosystems is
still state of the art, with most of the succeeding work being "backfill"
to verify and better establish the concepts laid out in that 1992 book.

BTW in looking around on the web I found a reference to another book
being published next year by Freitas and Merkle, Kinematic Self-
Replicating Machines, http://www.molecularassembler.com/KSRM.htm.
(What the heck is that a picture of on the cover?)  The book appears
to cover the entire field of designs for physical self-replicators, not
just nanotech ones.

Hal