[extropy-chat] Nanotech educations

[Adrian Tymes]

--- Chris Phoenix <cphoenix at CRNano.org> wrote:
> Adrian Tymes <wingcat at pacbell.net> wrote:
> > I'd have to know exactly what you know to pull the
> > best examples, but for a sampler, check out
> >
>
http://www.foresight.org/MolecularMachineSymposium/index.html#Topics
> > - and that's just the stuff organized enough that
> > there can be a conference about it.
> 
> Most of this is fundamental research and enabling
> technologies.  This 
> technical symposium is being pretty near-term:

True, and they have to be.  Supporting tech, by
definition, has to be developed before the final thing
can be developed.

> But there's nothing in the technical
> symposium about, for 
> example, CAD systems that can allow us to design
> with sub-molar numbers 
> of machines.  And I suspect at this point, it's CAD
> that'll turn out to 
> be the limiting factor in how quickly MNT can be
> developed.

...?  I'm finding AutoCAD works just fine for laying
out my designs, if I treat it like computer art (using
discrete pixel elements in 2D/3D arrangements; a
consequence of what they call "Manhattan geometry")
and I've only about 125 * 10^6 atoms - far less than a
mole - per voxel.  Granted, I need software to convert
to the file formats the machines use (like GDS), but
that's easy enough to pick up.  And I know AutoCAD and
similar packages have been used to draw theoretical
designs using individual atoms.

Or did you mean something else?

> What I'm complaining about is that general-purpose
> mechanosynthetic 
> exponential manufacturing is still considered as
> "where the tech might 
> lead" rather than "a major and powerful goal that
> will be achieved 
> surprisingly soon."

I'd suggest breaking that into two parts:

* a major and powerful goal: you've got that right,
and some long-term theorists do discuss it for that
reason.

* that will be achieved surprisingly soon: fully
fledged MNT probably won't be developed in the next
year or two.  Sooner than people would predict,
perhaps, but it doesn't look like we can leap to it
right now.  Unless you have a specific technical
architecture in mind which could be done with today's
technology...and I can think of at least a
couple...but those should be discussed as separate,
specific projects, succeeding or failing on their own
merits rather than just because or their similar end
result.

> I think this will delay MNT.  And make it happen
> more abruptly and 
> proliferate less controllably when it does arrive.

The former is bad; the latter, frankly, is arguably a
good thing.

> > Those saying nanobots are impossible have,
> themselves,
> > been largely silenced by disproof.  
> 
> Just a few days ago, Whitesides went on BBC and gave
> the Fingers Fallacy 
> to explain why Drexler possessed insufficient
> imagination.  Bleah, and 
> double bleah.  And I haven't heard Smalley retract
> any of his chemistry 
> mistakes yet, either.

Thus my disclaimer of "largely".  There will probably
be those claiming it's impossible even after someone's
done it.  The Flat Earth Society still exists, no?  A
closer analogy would be a certain infamous New York
Times column about Goddard's early rocketry work.

> > Which is why many
> > of them changed their tune (as I reported a few
> months
> > back in my NNI conference report): okay, they're
> > possible, but *here's a more feasable and more
> > immediate way to deliver the same benefits*...  
> 
> In these comparisons of feasibility and timeline,
> what cost and time are 
> they claiming for MNT?  Or are they simply engaging
> in rhetoric?

Mostly rhetoric, since they don't have timelines for
MNT - though they do point out that the fact that they
have timelines for their alternatives means that said
alternatives are probably closer to reality.

> > Cool thing, turning enemies into allies like that.
> ^_^
> 
> I'll believe they're allies when they stop telling
> people not to 
> research MNT.

Ah, but which "they"?  Like I said, there are a few
who will continue to disclaim; the key thing is that
there are more who have started to open up to it.
Relative numbers and sizes of research pools, not
absolute shutout - if you wait for everyone to agree
on a certain thing, you'll have a very long wait.

> >>And I do think it's important to plan ahead for
> such
> >>powerful technology.
> > 
> > I would tend to disagree, depending on the exact
> > meaning of "plan".
> 
> Identify ways we can hurt ourselves with it, and
> take whatever actions 
> are necessary to avoid the worst dangers.

I definitely disagree, then, based on the results to
date of attempts to do so.  But only directly with the
taking action part of it.

> For
> example, it looks like an 
> MNT-driven arms race is a very likely outcome, and
> would probably be 
> disastrous.  Planning would involve first, analyzing
> the danger in more 
> detail, and second (if it turns out that the danger
> is significant) 
> figuring out ways not to get into that situation.

At the NNI conference, an example of such an effort
was detailed.  Someone did a toxicology study on
buckyballs, reporting that injecting them into fish
throats until the throats were blocked off had harmful
effects on the fish.  (Well, *yeah*.  Most solids or
liquids will do that.)  The press reported it as, to
paraphrase, "NANOSTUFF IS TOXIC!!!"  The result was
seen as not harmful to overall nanotech research only
because no legislation had come of it.

In other words, identifying possible dangers as we
come to know about them is a good thing.  Acting on
the possible threats before we have enough data is
not.  If planning causes action before we can be
reasonably certain of the exact nature of the threat,
and our only control over the action is whether to do
the planning that causes it not, then don't plan.

(And yes, I realize this is kind of the Precautionary
Principle, subverted to extropian ends.  The irony is
not lost on me. ^_^; )

> We can know with pretty good confidence that certain
> things will become 
> techincally possible.  Like, the
> avionics/electronics for a jetliner 
> will weigh less than 10 kg and cost less than $200.
> 
>  From this, we can predict that certain things which
> are stable today 
> will become unstable.  And if one of those things is
> geopolitics, then 
> we might want to make a plan for a new way of doing
> geopolitics.

But by the time you get this far in practice, you've
made so many assumptions that - if you look at the
volume of assumptions - you can't have pretty good
confidence in this result.

> It could be argued that it's better to figure it out
> as we go along--to 
> wait until things start to fall apart, and then
> figure out where to go 
> from there.  I'm not being snide--this may actually
> make some sense if 
> we have no clue how the present systems work.  But I
> think we have 
> enough of a clue to do at least some of our homework
> in advance.

A little, perhaps.  Study and observe.  Just don't let
the studies slide into action before they're ready.

> > The
> > knowledge will be created and spread.  We can plan
> for
> > some of the aftereffects, trying to make sure most
> > people have access to this new power (but perhaps
> not
> > those who would abuse it the worst), but that
> > infrastructure is already in place.  
> 
> I'm not at all sure you're right.  Especially since
> it may not be abuse 
> that kills us, but simply old reactions in new
> situations.  Like arms race.

Ah, but solutions for the old reactions have already
been developed, no?  Like the arms race: develop
defenses ASAP once the nature of the threat is known
(and thus defenses can be developed); until then,
study the potential threat to determine it's nature.

> You don't think it's possible to identify likely
> instabilities and 
> vicious cycles, and work to steer clear of them?

Possible, yes.  Advisable *at this time*, no, based on
an analysis of previous instabilities/vicious cycles
caused by premature efforts to do so.

> I'm proposing
> _planning_ in the 
> sense of gathering information, projecting possible
> outcomes, and 
> sometimes choosing between them.

The first part of that is good.  The last part of that
has practical problems, if we try to act right now.

> >>MNT is not technology-limited. 
> > 
> > I strongly disagree.  
> 
> Sorry, I misspoke.  MNT _research_ is not
> technology-limited.  If we 
> wanted to start a crash nanofactory program today,
> we could write useful 
> software without waiting for any lab results.  And
> by the time the 
> software was done, the lab results could be
> achieved too.

That is what I understood (the hardware part, anyway),
and disagree with.  The state of the art in hardware
is, itself, limiting the research, and thus the lab
results.  Which is not to say there aren't problems
to solve on the software side too, just that they
aren't that dominant.

> > The state of the art available
> > to most researchers in precise placement of
> varying
> > types of substances is limited to the 10s of
> > nanometers range*, 
> 
> Huh--I thought DPN could get down to 5.  And surely
> ebeam of masks can 
> get below that.

This is ebeam of masks, actually.  I've heard DPN can
get down to about 15-30 for complex shapes; 5 is for
straight lines, which by themselves aren't that
useful.  (That may also be what you've heard about
ebeam.  I've seen individual dots made with ebeam that
have better resolution, but the processes involved
were not useful for anything but making small dots,
and specifically not for constructing useful systems.)

> > and that's a slow and complex
> > process for any significant volume of end result. 

> 
> Does "significant volume" mean ten nm^3, or 1
> micron^3?

mm^3 and above, where one can start applying nanoscale
properties to macroscale objects (like space
elevators, or enough nanobots to hunt down all the
cancer cells in an average adult human body).

> MNT is not just eutactic manufacturing.  It's
> exponential 
> manufacturing. 

Only if the assemblers make the assemblers.  Like I
said, once you have a device that can assemble things
at the atomic level, you have a device that can
assemble things at the atomic level.  Building more
assemblers may take a (lot of) resources and time,
which is why some people are proposing ways to make
assemblers other than just relying on the assemblers
to self-replicate (which works until the volume of
assemblers available can exceed the manufacturing
throughput of these other methods).  This is likely to
shrink the time to rapid deployment to less than the
time to develop it (measured from, say, publication of
Nanosystems).

> most researchers are quite uninformed about the
> goals or the 
> implications.

This I'll agree with, if you add "long term" in front
of "goals and implications".  But it's the same way
that, say, those researching better ways of achieving
LEO are uninformed as to the details of lunar
colonization: the data's there, and has possibly come
in front of their eyes more than once, but it makes
little immediate difference to them, so they ignore it
for now.  When it's closer to reality, expect more to
start paying attention.