[extropy-chat] Smalley, Drexler and the monster in Lake Michigan

[Rafal Smigrodzki]

----- Original Message ----- 
From: "Hal Finney" <hal at finney.org>

> He can't just wave his hands and say, if one thing doesn't work, we'll
> try something else.  He can't point to living things as an existence
> proof (because Drexlerian nanotech's revolutionary properties go far
> beyond anything possible with biology).

###  I think it's useful to consider the conditions that led to the
formation of biological mechanosynthesis, conditions which do not pertain in
the case of artificial mechanosynthesis, and try to imagine how changed
conditions could change the possible outcomes.

Consider, for example the need for self-assembly of every protein: The vast
majority of proteins are produced by a standard system (the ribosome) out of
a very limited number of building blocks, in such a way that the simple
one-dimensional ordering (protein sequence) determines the three-dimensional
outcome. This introduces severe limits on the kind of chemical interactions
between parts of a protein - they cannot be cross-linked for higher
strength, except by the use of other proteins or the limited expedient of
S-S bridges, so only select proteins are cross-braced (see collagen
formation), or else you'd need to have a bunch of specialized
post-translational modificators for each protein, and a bunch of
modificators for each modificator, and ...

The supporting structure of an enzyme has to not only support the active
moieities, but also code for its own folding, necessitating many loops which
are there only because of the folding requirement (see the difference in
size between the catalysts based on cage-compounds and proteins of similar
catalytic ability). Since there is no internal cross-bracing except for
non-covalent bonds (with few exceptions), the structure is very
thermosensitive.

Or consider the need for evolutionary malleability coupled limited
information storage and low computational capabilities: you need to change
enzyme structure very quickly to adapt to new conditions (so you have to
reuse existing devices and kludge them together to make new functions), but
the only way you can compute new structures is to make prototypes out of a
small number of building blocks (a cell couldn't store enzyme information
for making e.g. 20 000 building blocks instead of 21) and see if they
survive. This is a severe limitation on the amount of searching in the
design space you can do, even in millions of years.

The situation for the nanotechnologist is different. Let's imagine you have
enzymes (garden-variety biological ones) to make 20 000 distinct building
blocks - rods, braces, steppers, thousands of prosthetic groups for various
catalytic steps. Instead of making a protein catalyst out of 500 aminoacids,
you could use five or six structural blocks and attach two-three prosthetic
groups to have the same catalytic ability- but in a much smaller and more
durable package, as in a cage compound. You would need an assembler able to
grasp many more building blocks than a ribosome does, maybe even many
separate ribosome-like entities, yes, but without the computational limits
inherent in being a single cell that needs to reproduce *and* make most of
its components almost from scratch, this should be doable.

Also, if you are not limited to reusing existing enzymes, you can calculate
the optimal structure, and choose the optimal blocks out of your library of
20 000, you will make much more efficient devices. You can search the design
space far away from the list of available components and if you find
something worthwhile, you can make the components to get you there. With a
library of 20 000 specialized blocks you should be able to design organisms
with orders of magnitude higher concentrations of active elements, therefore
faster reaction speeds, smaller size, higher complexity - or simply, you'd
have MNT. It's true that the "assembler" might be a bank of 20 000
specialized building-block producers with a hierarchy of assembling devices,
each accepting the feedstocks of dozens of others to finally churn out the
products (such as a specialized producer, or fur for a teddy bear) - but
this makes it all the more believable, at least for me.

I do think that Drexlerian nanotech is possible, and pointing to living
things as the proof of principle is a valid argument. Almost all proofs of
principle, being hastily cobbled together from available parts, are slow and
clunky.... the final product of optimization will be much better.

Rafal