[extropy-chat] CRN's position on gray goo

[Chris Phoenix]

Robert Bradbury wrote (rearranged):

 > Focusing on "gray goo" involves a very specific assumption that
 > it can be developed or evolve quickly

CRN does not focus on gray goo.  See http://www.crnano.org/BD-Goo.htm

I mentioned gray goo in the context of a particular scenario, in the 
context of which gray goo was impossible (see below).  My point in that 
instance was that it would be good news for the nano industry if gray 
goo were actually impossible.

As far as I can tell, machine-phase gray goo is theoretically possible, 
but not our biggest nano worry (see below, and the URL above).

 > Trust me -- gray goo is possible because green goo is possible.

I use gray goo to refer to non-biological machine-phase replicators. 
Certainly, engineered green goo is possible.  Even fully constructed 
nucleic/protein/fatty acid self-replicating machines appear possible. 
That doesn't prove that goo using machine-phase chemistry is possible.

 > If it cannot adapt very quickly then it isn't a problem.
 > If gray goo is dependent on human design then it will probably
 > be centuries before it even might become a problem.

Gray goo requires a fabricator, a chemical plant (metabolism), a shell, 
an onboard computer with blueprints, mobility (either robotic, floating, 
or piggyback), and energy (either very efficient chemistry or solar 
collection), integrated.

None of these is conceptually difficult, especially if the chemistry can 
be inefficient.  Integrating them into a small (10 micron?) package 
could be very difficult, but I wouldn't want to bet my life on it.  And 
a mm-scale insectoid could be pretty destructive and require special 
effort to clean up.

 > AIs are not gray goo.  AIs are the problem that...

We need a Center for Responsible AI.  If someone wants to start one, 
I'll be happy to give him/her advice.  But CRN can't focus on AI as well.

 > I personally do not believe one can get effective gray goo unless
 > one invests millions of person-years of human engineering (not
 > likely to happen anytime soon) or one allows for self-evolving
 > AI design engineers.

I don't think any of the components/requirements listed above would take 
more than 100 person-years.  If designs are readily available for 
non-goo products (which will probably be the case), then adapting them 
might take less than one year.

 > Note the
 > assertion here -- biotech is a much bigger problem than nanotech.

In the short term, you are right.

 > I would assert that CRN (crnano.org) is focused on the wrong
 > problems.  Much more important are the risks of biotech and AIs
 > over the next 10-15 years.

CRN is CRN.  Along with CRAI, please start a Center for Responsible 
Biotech.  I mean that seriously.  But I know much more about 
machine-phase nanotech than about biotech, so I can't do CRB.  There are 
several big problems; there's nothing wrong about picking one to focus on.

If I thought that we definitely had 10-15 years before tabletop 
molecular manufacturing factories, I would not be making so much noise 
about the risks thereof.  But I don't think that.  I think it's much 
simpler than most people think it is.  And we're closer than most people 
think we are.  I wouldn't be at all surprised if we saw basic 
nanofactories by 2010.

And some of the problems will be harder to solve (take more time to 
prepare for) than bioweapons.  And some of the benefits will greatly 
mitigate current problems, including biotech problems.  And knowing in 
advance when computers are going to get a millionfold cheaper is 
important in planning for AI.  So I think what I'm doing is worthwhile.

Chris

-- 
Chris Phoenix                                  cphoenix at CRNano.org
Director of Research
Center for Responsible Nanotechnology          http://CRNano.org