From: pcm@rahul.net (Peter C. McCluskey)
Sender: owner-polymath@its.caltech.edu
To: polymath@its.caltech.edu
Subject: Re: poly: New paper on Grey Goo problem
Date: Fri, 19 May 2000 11:42:50 -0700 (PDT)


 hal@finney.org (hal@finney.org) writes:
>http://www.foresight.org/NanoRev/Ecophagy.html is a paper by Robert
>Freitas on the so-called "gray goo" problem, the danger of uncontrolled
>nanotech-based replicators.  There has been some discussion of this paper
>on the Extropians list but polymath readers might find it interesting
>as well.

 Freitas assumes that if the goo consumes biomass in a way that produces a
certain amount of heat, that all that heat will be produced in a way that
constrains the spread of the goo by forcing it to operate at higher
temperatures. I see three strategies which the goo might use to spread
faster than Freitas' analysis indicates (I'm assuming that the goo is not
attempting to avoid detection at this point):

 1) The goo could perform some initial decomposition of the biomass using
moderate temperature solution based chemistry, and later switch to other
approaches as the local temperature rises as a result of the conversion.
This shouldn't cause large changes to Freitas' analysis, but it does indicate
that the probable need to do some operations at temperatures below 373K
doesn't mean the average temperature at which the goo operates must be that
low.

 2) Active transport of heat from the leading edge of the goo to regions where
the goo has already consumed the biomass. I haven't attempted to analyze how
this affects Freitas' conclusions.

 3) The goo has no apparent need to quickly convert any more biomass than
it needs for transport and replication. If the goo simply carves up much
of the biomass into small enough chunks to disable the biomass, and saves
more energetic decomposition for later, it can accomplish goals that humans
will treat as equivalent to total destruction before producing much of the
waste heat that Freitas shows total conversion would produce.
 I find it fairly easy to imagine that a replibot could convert 10% of the
local biomass into more replibots, and store the remaining 90% of the
biomass in a "stomach". This would appear to enable destruction to spread
an order of magnitude faster than Freitas' waste heat analysis suggests,
and I don't see any reason to expect that this is the most dangerous
strategy that the goo could adopt. I suspect that if waste heat is a
limiting factor for the spread of goo, that the limits need to be calculated
based on the minimum amount of energy needed to disable the biomass
(i.e. what is the minimum energy needed to replicate a replibot that can
klll X kg of biomass, and what is the minimum energy required to do the
actual killing?).

 Finally, I'm disturbed by the vague implications in the paper that detection
is the hardest part of dealing with the standard gray goo threats.
 There are plenty of examples of somewhat similar problems (AIDS, wildfires,
military invasion) where early detection is not enough to avoid major problems.
While it is easy to imagine nanotech improving our ability to contain such
threats, there is no obvious reason to be confident that the technology used
to contain threats will improve faster than the technology used by malicious
forces.
 I liked your analysis of the weaknesses in Freitas' assumptions about
containment at:
extropians: Freitas on Gray Goo.