[extropy-chat] Re: Social Implications of Nanotech

[Chris Phoenix]

On Sat, 15 Nov 2003 12:27:02 -0500, Robin Hanson wrote:
> I said from the start that the reason I started this conversation is that
> in a few weeks I will participate in an NNI conference on the social
> implications of nanotech.  So I definitely need to take their scenarios
> seriously.  I'd also like to discuss scenarios like you prefer (though
> perhaps because of that they won't actually let me talk and won't
> invite me again).  And I'd like to discuss these in terms of their
> economic implications.

I knew you were talking to the NNI.  What I didn't know is that the NNI
had preexisting scenarios; sorry if I missed that.  I thought the
scenarios you were presenting were ones you had designed to try to
explore the problem space.  So I've been suggesting that some of these
were not doing the job, because they were not technically plausible.  

> So what I really want is to break the gulf between these very different
> views into component assumptions, expressed in economic terms.  This
> allows compromise and debate in the form of "I buy these assumptions but
> not those", and positions held can be more easily translated into
> scenarios with intermediate social implications.

So you're not just trying to put together a range of scenarios between
"no MNT" and "MNT", but between a preexisting/artificial position and
"MNT".  That is a rather different, and I think harder, problem.  Now I
can see why we've been talking past each other so much.

Suppose you were talking about agricultural productivity.  El Nino is
important.  You might make one scenario in which we had an El Nino in
2005, and one in which we didn't.  But can you express the proposition
"2005 will be an El Nino year" in economic terms?  And if not, what do
you call things like that, and at what point do they enter your scenario
process?

The reason I ask is because "Mechanochemistry can build diamondoid" is
comparable in some ways to "2005 will be an El Nino year."  It's a
statement of physical fact, and has to be decided explicitly in order to
know which scenario you're in.  You can discover relations between
weather and productivity without knowing what the weather will be.  But
you can't predict the productivity without knowing, or making
assumptions about, the weather.  And I don't think you can make useful
manufacturing scenarios without talking about the underlying technology.

This is a separate question from computing the effects of various
outcomes.  You can predict that if productivity is lower than X, there
will be famine.  Whether X is a plausible number at any time in the next
Y years is not addressed.  

> >Above, in the discussion leading to the two dimensions of variability
> >for scenarios, was my best understanding of the differing *technical*
> >assumptions that people are making--and how to unify them, by
> >recognizing that something similar to MNT will happen eventually > whether
> >or not dry nanomachines work, but if it's far enough out it will blend in.

> I'm having trouble parsing your text into an itemized list of assumptions.

I guess I'm still not understanding what you mean by "assumptions".  If
I don't get it right in this email, we probably need to talk on the
phone.

> And some of the assumptions I can identify have to do with how the
> self-replication scenario plays out, rather than how it shades into others.

And I'm not at all sure that the self-replication scenario can "shade
into" others.  I guess one type of shading would be: what if a single
factory could duplicate itself, but required much labor to set up the
result in a new location?  But with MNT, a self-replicating factory
could easily be as small as a cubic centimeter, and would have to be
completely automated.  So in practice, I'm not sure that there is any
shading to be found.  This is the basis of much of my criticism of your
scenarios.

Are these assumptions of the kind you're looking for?

Total automation is: Easy; Very difficult; Impossible.

Nanoscale engineering: Is Easy in general; Each instance requires
special design; Each instance requires special techniques (research).

Nanoscale (dry, stiff) machinery: Is easy and efficient; Is not
generally efficient; Is impossible.

Self-replication of manufacturing systems: Is easy and compact; Can be
done with difficulty; Is impossible for practical purposes.

General (programmable) manufacturing: Is easy; Is doable in limited
domains; Is not efficient enough to be worth developing.

Mechanical chemistry: Is a useful manufacturing process; Is too limited
to be useful.

The interesting thing is that by themselves, the "easy/useful" choice of
each of these assumptions is questionable.  Want to build a car with
100% robots?  What's the point?  Build a self-replicating system? 
Why--what's it good for?

But if you choose the "easy" choice in all of these simultaneously, you
find an MNT system.  It really does make all of these easy.

If you try to "sneak up" on MNT by making a scenario with the "easy"
choice in all but one of these assumptions you run into two problems. 
First, the results of your scenario change drastically, so you can't
"sneak up" on MNT results this way.  And second, it's not technically
plausible due to interactions between the assumptions.  For example:
Mechanical chemistry plus general manufacturing implies easy
self-replication.  Mechanical chemistry plus self-replication implies
general manufacturing.  And easy self-replication plus general
manufacturing probably requires programmable chemistry.  

If you want to talk about relations between capabilities and results,
some useful questions/choices about capabilities may be:

Are factories cheap enough to build for convenience?  Do they sit idle
much of the time?

Does general-purpose manufacturing work?  Can a single person design
and/or make a complex product without knowing anything about
manufacturing?  (In the way that a person can make a brochure without
knowing anything about typesetting other than how to specify fonts.)

Is material strength likely to get significantly better than today's
metallurgy and polymer chemistry?  

Does product complexity affect manufacturing cost?  (If I design in a
computer per cubic millimeter, does it cost any extra to build it?)

Are manufacturing systems suitable for home use?  Service bureau?  Or
just specialized/centralized production?

Can nanoscale components (sensors, computers, eventually actuators) be
produced and/or handled cheaply enough to integrate them with general
products in combination and in large number, or are they difficult to
integrate, leading to products being built around one or two nanoscale
components for specialized purposes?

Can nanoscale components be rationally designed and straightforwardly
built, or does it take lots of idiosyncratic research to design and/or
build each new nanoscale component?

If this is useful, I could come up with several more.  But again,
consider the limitations of going straight to scenarios instead of (as I
said above) working out relations between capability and effect, then
building scenarios based on specific technologies and their effects on
the above capabilities.  I think the latter will produce scenarios that
are both more defensible and, in the case of MNT, far more powerful as
the relations synergize and produce extreme results.  

> I teach law and economics and a standard issue there is whether to enforce
> contracts between someone lost in the desert and someone else who agreed
> to give him water for some enormous sum.  Such contracts are not actually
> enforced, on the basis that monopoly pricing would lead to low quantity,
> i.e., some people would die from inability to pay, but the counter-argument
> is that enforcing them encourages more people to look for people to save.
> I'm not sure if this addresses your "market" question though.

I think it does.  If you don't have enforceable contracts, you don't
have a very reliable or useful market--there's no way to defer payment
or to know the value of a promise.  So it sounds like, in at least some
sufficiently extreme circumstances, markets can't work well.

Now the question is, if the cost of providing something is much lower
than the value, and the value varies widely between users and is
unknowable to the provider, is this a situation where no useful market
can exist?

> >...  The criticism of MNT theory has been of such low
> >quality, and people are so willing to believe blanket assertions without
> >testing them for plausibility or relevance, that by now I don't worry
> >too much about panels of experts.  ...

> If you care about what other people think who might listen to these
> experts, then you might want to try to convince those experts.  In which
> case you need to state to them as clearly as possible, in terms they are
> familiar with, and in a style they can respect, how you come to disagree.

What I meant by "I don't worry about" is that I don't usually worry that
they are right that MNT can't actually work.  I'll cheerfully shrug off
the proclamations of a Nobel prize winner--if he's talking outside his
field, and saying things about chemistry that aren't even true outside
of solution chemistry.

I am usually quite good at communicating with experts (in science--I
don't know as much economics :-) in their own context and explaining why
I disagree with a particular key part of their thinking.  I've done
quite a bit of this.  See for example my debate with Bill Atkinson (a
generalist, not an expert, but the principle applies):
http://www.nanotech-now.com/Atkinson-Phoenix-Nanotech-Debate.htm
And my current extensive discussion with Richard Jones:
http://www.quicktopic.com/24/H/UMabNQK2xXW

Some experts seem to be both hasty and entrenched.  Even there, I try;
I've made some attempts to communicate with Smalley, got no reply, and
may not try again.  But if any expert is willing to take the time to
examine assumptions, I can and will take the time to give them a very
useful discussion.

Chris

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