Re: Freitas on Gray Goo

From: Robert J. Bradbury (
Date: Sun May 14 2000 - 23:16:45 MDT

On Sun, 14 May 2000, Eugene Leitl wrote:

> Robert, this assumes that hotspots are in line of sight, that we have
> enough gear to watch the whole of the surface, and that behaviour of
> agents is unmoderated aggressive.

There are a whole host of problems that have to be solved for
bioecophagy. I suggested a number of clever work-arounds to
Robert, some of which were addressed, some of which may not
have been. The real *fine* point comes down to the discussion
of whether the machines evolve naturally or are consciously
designed to evade detection.

The natural biosphere is unmoderated aggressive, so I don't
view the "natural nanobots" scenario as much different from our
current reality except that it can disassemble the biosphere to
create a nanosphere but is constrained on a number of fronts.
How for example does a CHON (diamondoid) nanobot "grow" in the
desert or arctic?

If you are talking a multi-material (C -- Al2O3 -- SiO2) nanochemistry
then I think you are talking something created and not something
accidentally evolved.

> Given that behaviour will be at
> least initially set to stealth and dispersal, before the midgame
> behaviour will be expressed, these are not safe assumptions.

You are talking consciously designed evil-bots here, so you have
to assume (a) defense technology ~= offense technology; or (b)
a huge disparity (which requires conventional weapons responses).

> You have to be there in person, i.e. have the whole landscape covered
> with a distributed diagnostic and rapid response mesh, with sensors,
> intelligence, caches, rapid trasport and deployment as well as mop-up
> procedure. This is not trivial infrastructure.

No, as I pointed out to Eliezer, you can simply raise the environmental
temperature to a level sufficient to exterminate the vermin.
As a last resort, nuclear weapons do this quite effectively.
Robert, I believe, points out that the thermal conductivity
of air is pretty poor, so if you can identify infected regions
you simply incinerate them. What you need to balance this and
make it an effective strategy is the construction of current-tech
enclaves that are not easily breeched. Cheyenne Mountain
( comes to mind.

You propose an extreme situation -- the release of consciously
designed destructive nanobots -- you cannot expect the response
to be "polite". This is a "take no prisoners" war. The first
time you encounter one (like an nasty agricultural pest) the
response should be overwhelming and as effective as possible.

> Its presence will
> create artifacts, which is not desirable for a countermeasure. It will
> take it sweet time to be developed, deployed, and debugged.

No argument, a "surgical" defense, however desirable may not
be available. Best hope you are not living in locations where
the non-surgical defenses are required.

> Immune system is flesh policing flesh. Here we have dry solid-state
> devices policing themselves and vulnerable ecosystems. Imo, this is
> not comparable.

Perhaps. The badbots (in an engineered scenario) know that the
goodbots must eliminate them at all costs. They must either be
very quick or very stealthy. The badbots might have a "enter" and
change the genetic program of the host, and disintegrate protocol.
But that would have to survive organismal state monitoring and
correction by your self-bots. If the self-bots are programmed
to maintain homeostasis then their capabilities over and above
normal physiology should keep things from out of hand. The
only effective strategy I can see is to sabotage the self-bots
and that would require penetration of security protoocols.

If you have sloppy security protocols, then you probably
get what you deserve.


This archive was generated by hypermail 2b29 : Thu Jul 27 2000 - 14:11:14 MDT