[extropy-chat] Re: Social Implications of Nanotech

[Chris Phoenix]

On Sat, 15 Nov 2003 13:36:05 -0500, Robin Hanson wrote:
> We should consider both 
> conventional nanotech scenarios and radical ones, such as described in 
> Unbounding the Future and Diamond Age.

Would this audience be completely lost if you referenced Nanosystems
instead?  That's intricately technical, but does list quantifiable
technological capabilities such as power conversion densities (petawatts
per cubic meter) and computing power (petaflops per watt).  And makes a
reasonably strong case for the technological basis of dry
self-replication and flexible manufacturing.  

> Radical scenarios seem less likely 
> but have more severe social implications.  

If two things are equally allowed by the laws of physics, but the more
radical one is more profitable, then it becomes more likely.  I think
that's the case here.  Molecular manufacturing without self-rep cannot
be more than a niche technology, used for building sensors and other
components, and maybe a few small and incredibly expensive products. 
But MM with self-rep is what leads to general-purpose manufacturing.

> As a prelude to future modeling 
> attempts, I here try to identify five key assumptions to bridge the > chasm 
> between conventional and radical nanotech scenarios.

Seems like you're outlining scenarios, not assumptions.  Drexler's
projections are based on taking several assumptions together and seeing
what results.  If you take one assumption at a time (and assume that the
others do *not* hold), then you're building scenarios--which are far
less interesting than general diamondoid MNT, and I think also far less
plausible.

The assumptions underlying MNT:
1) Stiff mechanical devices (nanomachines), including actuators, power
transmission, bearings, and interlocks can work at the nanoscale.
2) Programmable nanomachines can do chemistry which is sufficient to
build programmable nanomachines.  (For diamondoid MNT, the chemistry is
diamondoid chemistry, and the products benefit from diamondoid
materials.)
3) Large systems can be integrated from nanomachines with fairly simple
engineering.
4) All this can be automated.

If all these assumptions are correct, MNT is extremely powerful.  Take
one of them away, and it simply doesn't work--you're not describing
MNT--you're not describing anything.  Other nanotechnologies can be
built from some of these assumptions (or variations on them) combined
with some other assumptions--usually with very different implications.  

So what I'm saying is, don't try to take parts of MNT and evaluate them
one at a time.  It's like trying to evaluate a car without a motor.  No
one would want it, so there's no reason to build it, and it wouldn't
have any effect.

I'll illustrate this by going over your scenarios and listing the
assumptions you appear to be making in each one.

... Having gone ahead and done that, I'll come back and note that some
of your five "assumptions" appear to be effects.  For example, "A local
manufacturing plant can create a copy of itself within a year."  There
are only a few technologies that can do this, so this is an effect of a
small subset of technological possibilities.  So then I get confused
when you talk about scenarios that don't fit any of these technologies,
but present them as being on the same continuum.

> 1.  Atomic Precision:  Atom-scale manufacturing is feasible; we put some 
> atoms where we want.

Assumptions: 
Mechanochemistry works to some extent.
Mechanochemistry can't close the autoproductive loop, but it can build
products.  (Questionble assumption IMO.)
The manufacturing systems are created by non-autoproductive systems.  (I
think this means they're likely to be *VERY* expensive.)

> Depending on how cheap this ability is, and which atoms, many new products 
> may be possible, including much cheaper computers, and perhaps medical 
> devices that float in our bloodstreams.

I recommend a different example.  Floating medical devices have been a
poster child for the anti-MNT crowd.  Perhaps sensors cheap and small
enough to integrate into all products would be more believable.

> 2.  General Plants:  General purpose manufacturing plants, using a limited 
> range of feedstocks, will displace most special purpose plants, like 
> general purpose computers have now displaced most special purpose signal 
> processors.  (This is mature "3D printing" or "direct manufacturing.")

Assumptions:
It appears you're not assuming mechanochemistry for this one.  
It's not clear whether these manufacturing plants are autoproductive.
It's not clear whether this technology transforms the feedstocks to
obtain new material properties.
>From what you say later, it sounds like these general purpose plants are
not assumed to be fully automated.  How can humans cope with machines
with such flexible behavior?  (Note that automation is different from
self-repair is different from high reliability, and I'm not sure which
you're talking about.)

> As with computers, this requires that the efficiencies of special purpose 
> devices be overcome by the scale economies and lower design costs of 
> general purpose devices.  When transportation costs matter, products would 
> likely be made at the general plants nearest to each customer.

This does not go very far toward analyzing the effects of MNT.  We
almost have this technology today.  

This scenario requires an extremely flexible fabrication technology,
which it does not in any way specify.  So there is no way to evaluate
the effectiveness of this technology or the scope of its implications. 
You can make projections anywhere from "no significant effect" to
"massive disruption of everything."

> 3.  Local Production:  Small general plants, located in or near homes, 
> dominate manufacturing.

Assumptions: like 2, only more so.  Now the functionality has to be
packed into a much smaller space.

> This requires that production processes be almost fully automated, with 
> human intervention rare. Such high automation seems harder to design.

I still don't understand how a manufacturing system building nanoscale
products or nanoscale components can hope to work without complete
automation.  This seems to be a counterfactual scenario.  

Automation need not be hard to design if the function being automated is
sufficiently simple.  For this, it helps to have "plenty of room at the
bottom" so that you can design your product components to be
snap-together.

>   Here 
> costs of transportation and labor for manufacturing are mostly eliminated; 
> what remain are costs of design, marketing, regulation, feedstocks, and 
> rental of general plants.  As with PCs today, open source product design 
> and file sharing of stolen product designs could become issues.

> 4.  Over-Capacity:  Local general plants are so fast/cheap that they are 
> usually off, like PCs now.

This sounds like not even a scenario in its own right, but an extension
of previous scenarios.  If the cost of building a general plant is lower
than the cost of coping with fluctuating demand (e.g. delayed
availability, warehousing, shipping), then the plant will be built. 
Tweaking the parameters of scenario 3 will lead to scenario 4.

"idle" is a more accurate description than "off".

> For most products, the main marginal costs would be feedstocks and 
> marketing.  Fixed costs of design, regulation, and marketing would dominate 
> total costs, as with software and music today.  

Um, what about profit-taking?  Remember I'm not an economist, so this
may not be the right term.  What I mean is that if someone has a
near-monopoly (e.g. the music industry) then they can simply charge
higher prices and pocket the difference.  With the amount of regulation
MNT may inspire, this seems like a very plausible source of cost.

> Like software and cable TV 
> companies that now offer a small menu of product packages to 
> price-discriminate via anti-correlations in item values, future consumers 
> might be offered a few lifestyle packages that cost most of their salary 
> and entitle them to designs for clothes, furniture, food, etc.  This would 
> require high concentration of or coordination by sellers of consumer good 
> designs.

Life for rent... brrr!  

This would have other implications in other areas than lifestyle product
packages.  Black market in designs for illegal products?  Force
multiplier and deployment expediter for militaries?  These may be more
important than the commercial effects.

I think the question of whether manufacturing plants are usually
idle--whether they're cheap enough to exist for convenience in some or
even many contexts--is a very important question to ask.  Am I right
that if they exist for convenience, this completely removes many
frictions that currently exist in our economy?

> 5.  Self-Reproduction:  A local manufacturing plant can create a copy of 
> itself within a year.

I don't at all understand why this isn't implied by 4), assuming that 4)
makes use of mechanochemistry or otherwise produces programmable
nanoscale features.  (And if 4 doesn't include mechanochemistry, in what
sense is it a nanotech scenario?)

> This is one possible route to achieving over-capacity of local general 
> plants. This route, however, has the potential to give a large and sudden 
> cost advantage to the commercial or military power that first develops 
> achieves it.  How large an advantage depends on just-prior costs, and how 
> sudden depends on self-reproduction time.

Yes.

>  Self-reproducing military or 
> terrorist weapons become a concern.

In general, I don't think so.  Most products will not drag production
systems around with them--that would be very inefficient.  

Chris

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