Re: EngYog-Sothoth for Fun nanotech

From: Robert J. Bradbury (
Date: Thu Jul 26 2001 - 03:22:45 MDT

Anders wrote:

> I would rather say we are striving for more complexity than
> more evolution.

Here, here (I've been watching Masterpiece theatre tonight...)

> add extra DNA repair

Yes of course -- take out the SOS response that actually
*increases* the mutation rate, then paste on some of the
Deinococcus radiodurans repair mechanisms, then do some
molecular modeling to improve the proof-reading. You
can drive the error rate very low.

> possibly make some of the potentially dangerous genes
> brittle in the sense that any mutation there will disable
> the cell (how to to achieve that?).

I love it on those rare instances when I'm actually ahead
of Anders on the curve... :-)

First you make the genetic code brittle in terms of no
"redundancy". To do this optimally, you have to study
the mechanisms driving transition & transversion frequencies
and "rig" the code such that the most likely mutations
generate either STOP codons or "radical" amino-acid changes
(i.e. + charge to - charge, hydrophobic to hydrophilic, etc.).
I haven't tried to do this yet -- I suspect to really
get it right requires a fair amount of simulation based
on existing genomes.

It looks like one of the things you really want to do is
disable insertion sequences/retrotransposons (the July
20 Science paper on the virulent strain of S. pneumoniae
says its got a genome with 5% insertion sequences --
much higher than most bacteria). So probably you want
to add in transcription sequences for anti-sense RNA
to any known sequences that can hop around.

You can drive the mutation rate to very low levels if
you work on it enough. Nature doesn't generally seek
really low mutation rates, so this is an unexplored
part of the phase space.

Entropyfoe wrote:
> These new organisms will be dependent totally on the supply of the
> synthetic amino acid to produce its proteins.

It doesn't have to be a "synthetic" amino acid -- you can simply
delete the synthesis pathways for a couple of the more essential
amino acids. If you have changed the "code", then horizontal
gene transfers from existing organisms coding for those genes
isn't going to do you any good.

Back to Anders:
> I wouldn't want to bet that for a given simple organic molecule
> there is *no* metabolic pathway in the biosphere that produces it.

True, but if that pathway only exists in organisms found in deep
sea vents, its going to be hard to transfer it to a field in Iowa.
Plus(!), if the genetic code is different, the transfer of the
pathway is worthless. The risk is the probability that the transfer
of "random" DNA is likely to produce a functional pathway in the
organism with an alternate genetic code. If I transfer random bits
off of my Cisco router (with a 68000 processor) what is the probability
that I will get executable code on my PC?

> Linda Nagata did a nice trick in her _Limit of Vision_ to show
> how you could get around such a metabolic constraint on a GMO
> both short term and long term

I'll have to read that -- I'm suspicious that she may have leaped
the fence of statistical improbability. If there your error rate is
10^-50, *it* can happen, but I'm not going to lose any sleep over it.

> I think one has to base the metabolic key on something more tricky to
> manufacture in cells.

One or more patents are forthcoming... :-)

> I envision mass production of billions of nano components for
> nano-electronics, nano-computing or optics.

Yep, one micro-organism per part. Its more expensive than you
would think however to get there. Its going to require a
non-trivial amount of software engineering to leap the gap.

> It might not have the wide potential of hard drextech, but it
> is a likely candidate for both early investment and bootstrapping
> more advanced forms of nanotech.

Drexler, 1981; Drexler, 1994; Merkle 1999
Bradbury, 2001 (forthcoming -- the devil is in the *details*)

Eliezer wrote:
> My own reaction was "If you built a self-reproducing car, *then*
> I would start worrying about its descendants learning to live off
> tree sap."

You have to realize Eliezer, that the 40 trillion bacteria living
on or in your body *would* eat you, *if* they could. Their "creativity"
potential hasn't reached the point where they pose a serious threat
to your continued existence. The S. pneumoniae example I cite
above is a good example of evolution mastering "creativity^2".
It isn't enough to simply be able to mutate random letters in
the genetic code -- several billion years of evolution may
have done that and not produced very much because most mutations
are dead-ends in the phase space. When you learn how to take
"working" sequences, replicate them, and dedicate them to other
purposes, now you have speeded up evolution significantly.

I'd predict that the Cambrian Explosion
probably was associated with the development of retrotransposons.
(How else do you get to copy the homeobox genes resulting in
body patterning?)

Now, "Intelligent Design" (ID) is about to produce "creativity^3".
So I'd predict that since ID will be able to minimize creativity^1
and creativity^2 it is going to triumph big-time. That isn't to
say that ID might not be able to produce creativity^3.5 --
but it isn't clear to me how one would do that without AI.

> But I'd still be worried about the new organisms learning how to
> produce the synthetic amino acid on their own using chemical processes.
> Even if the original amino acid included a rare atomic type as an
> integral part of the chemistry, I'd still be worried about the
> organism learning to build a substitute.

Enzymes have little ability to distinguish isotopes. The
incorporation of isotopes into biomaterial is based on simple
statistical mechanics regarding the abundance of the isotopes and
their mobility (based on mass) of being more bio-available.

Regarding the organisms "learning" how to produce a natural
(or synthetic) amino acid -- creativity^1 is going to take
a *long* time (with an alternate genetic code); creativity^2
has a better chance but it depends on the similarity of
the path to pre-existing paths and the rate of creativity^1.
Suppression of both levels of creativity is going to make
it difficult for the organism to "learn". (I'd say "learn"
is a poor description of the process -- "discover" might be
more appropriate.)

Furthermore, there are ways you can engineer into organisms
"fail-safes" that would prevent them from deviating much from
their "pre-programmed" path without self-destructing. Whether
these would be necessary is unclear to me at this time.


This archive was generated by hypermail 2b30 : Fri Oct 12 2001 - 14:39:57 MDT