Home > The Atkinson-Phoenix Nanotech Debate -- Page 2
The Atkinson-Phoenix Nanotech Debate - Page 2 - Emails
Commentary text by Chris Phoenix, except where otherwise noted. [Bracked Italics] indicates a correction. Italics used on text indicates a comment taken from an earlier email, written by the other person.
Last update: August 04, 2003.
Page 1: Overview
Page 2: Emails through June 27, 2003
Page 3: Emails through July 14
Page 4: Emails through August 4
Page 5: The wrap-up
June 17, from Bill Atkinson to Rocky Rawstern (editor of Nanotech-Now, which published my review)
Dear Mr Phoenix: A brief response to the on-line review of Nanocosm.
Most of the review's objections can be put to rest by my interviewees. Anatomical and functional distinctions between grey
and white brain tissue are those of Roger Fouts of The Chimpanzee and Human Communication Institute, Central
Washington University; laser spectroscopist Geraldine Kenney-Wallace has documented how chemical reactions commonly
occur in a few tens of femtoseconds; etc.
Of course conventional thermodynamic calculations forbid a Maxwell demon; my suggestion is tongue-in-cheek. Yet
consider that if CERN characterizes the Higgs boson by 2005, resultant innovations such as synthetic gravity might also
violate current theory. New data invariably encapsulate old theory as a special case.
I fully defend all observations and conclusions about Eric Drexler. Science demands skepticism about anything imagined yet
unproduced. One could argue details forever, but two facts seem incontrovertible. First: If all objections are answered,
where is the device? If A then B: Not-A: Therefore not-B. Second: Mr Drexler himself is on record that nanotechnology has
been empirically redefined, and must now be considered separate from the molecular assembler (New Scientist, 2003
From molecular switches to metallic ceramics and fullerene structures, from parallel failure-tolerant nanoarchitectronics to the
renascence of analogue, real nanoscience is rapidly producing tangible, bankable results. Mixing up such genuine work with
unproven speculation, no matter how fascinating, does no one any service.
William Illsey Atkinson, June 17th, 2003
June 17, from Bill Atkinson to Chris Phoenix
Subject: Review Response
Hi Chris: Don't know if you've seen this [response above]; I sent it this A.M. to Rocky
Rawstern. We probably have to agree to disagree, but hey - worse feuds
have ended in a surprising acreage of common ground.
June 17, from Chris Phoenix to Bill Atkinson
Subject: Re: Review Response
Rocky forwarded it to me. That's why I sent the email to Roger Fouts.
[I emailed Roger to find out if he really said what Bill reported about white matter.
He really did; he was using "white matter" to refer to myelinated neurons--the whole
neuron, including the gray cell body, not just the part with the myelin. I have
not heard this usage elsewhere, and my bio textbook contradicts it. So I think Bill
and I are both right on this one. --CJP]
I'll probably put a footnote in the review on the gray/white matter
thing. I'm still deciding what it'll say--probably something to the
effect that your statement was a plausible reading of what Fouts wrote,
but what Fouts meant was (etc).
However, this doesn't affect my other statements about the technical
inaccuracies in the book, and I stand by my statement that there are too
many of them to recommend the book. I didn't cover all of them in the
review. Your discussion of the buckytube golf club gives a very wrong
impression of the properties of buckytubes (and the physics of things
moving through air, and what it takes to move a golf ball, etc.) There
are many engineering details wrong in your description of the space
tower. Your design for the diamond-making cellular automaton--"Make all
carbon bonds covalent"--is unworkable and based on a false assumption
(carbon bonds are already covalent). I could go on for a long time.
Your "Maxwell Demon" suggestion might be excused as tongue in cheek on
page 79, but not on page 277. To defend that, you might have to say
something like, "resultant innovations such as synthetic gravity might
also violate current theory. New data invariably encapsulate old theory
as a special case." This may be true, but I might reply, "Mixing up ...
genuine work with unproven speculation, no matter how fascinating, does
no one any service."
I fully defend all observations and conclusions about Eric Drexler. Science demands skepticism about anything imagined yet unproduced.
Science demands skepticism, yes, but not ridicule. Science demands
testing of hypotheses, not insulting rhymes. And in fact, your
objections against nanobots are also unproven speculation. You don't
actually know whether assemblers can maintain their internal state with
sufficient precision; to state categorically that they can't is far less
justified than claims that some assembler designs probably can.
One could argue details forever, but two facts seem incontrovertible. First: If all objections are answered, where is the device?
This is a question, not a fact, and it has a simple answer. Answering
the objections doesn't build the device. No one pretends that it will
be simple, or could be done in a year by grad students. Since absence
of the device is easy to explain, it provides no evidence for its
To be scrupulously fair, the last objection won't be answered until the
device is actually built. But the question at this point is not whether
every single objection is answered to the satisfaction of the objector,
but whether there are any showstoppers in the laws of physics, the
design capability of computer-aided humans, or the realities of
economics or politics. There are no known showstoppers. You think
you've found a whole new crowd, like the claim that carbon is death to a
nanomachine, but this is just because your understanding is
superficial. You have also found some difficulties, such as the storage
required to specify a nanobot being larger than the bot itself. This is
not a showstopper, and perhaps one of your biggest mistakes is treating
every difficulty as a showstopper. With enough work, difficulties can
be overcome, and we already know how to solve every one you mention in
If A then B: Not-A: Therefore not-B.
The correct version is "If A then B: Not-B: Therefore not-A." The
version you gave is invalid. For example, if it's a horse, then it's a
quadruped. It's not a horse (it's a pig): therefore it's not a
quadruped. I harp on this because it demonstrates either carelessness
or lack of knowledge--which is also present throughout the book.
Someone who is expert in a field can afford to be self-assured. But you
cannot afford to believe things just because they seem sensible to you.
Many of your concerns about Drexler's work are quite easy to answer.
Many others are simply wrong, such as your claim that he thinks
nanoscale engineering is "same old stuff"--chapter 2 of Nanosystems
Second: Mr Drexler himself is on record that nanotechnology has been
empirically redefined, and must now be considered separate from the
molecular assembler (New Scientist, 2003 April).
I don't see what this has to do with the technical feasibility of
It does have significance for investors: since there are two fields
labeled "nanotechnology", they need to know the difference before they
can decide what they want to invest in. Drexler has done his best to
clarify the difference, by inventing a new term for his work: "molecular
nanotechnology." The confusion is certainly not his fault, and he is
not trying to exploit it--quite the opposite. A book that made clear
the difference between today's nanoscale technology and Drexler's
projections of molecular manufacturing would have been quite appropriate
and useful. But you chose to emphasize the difference through ridicule,
unfounded technical criticism, and personal attack. That was
unnecessary. And unfair to your readers. And, I think, led you into
intellectual dishonesty, such as complaining about Drexler's supposed
lack of provision for cooling immediately after complaining about his
analysis of cooling systems.
We could spend many hours discussing the specific areas where your
understanding is lacking and Drexler's work does not deserve the
criticisms you have leveled against it, but I suspect in the end, this
would do nothing to change your opinion of Drexler's results. It is
easy for me to see where you are wrong, but it will be a lot harder for
you to see where you're wrong. But if you're not willing to go there, I
don't know that we can find any common ground.
Some of your criticisms are worth answering. I have been working on a
paper that addresses issues of control and integration: Assuming we can
build a single assembler, how can it be integrated/bootstrapped into a
tabletop nanofactory capable of building products from a CAD file? The
paper is over 50 pages, and not done yet. It is not a simple problem to
solve. But I have not run across anything that stumped me, and I've
considered control, power, chemical distribution, heat dissipation,
component failure, assembly robotics, and several other issues. I
believe I've solved each of these in enough detail to show that a usable
design for the entire factory could be completed today.
There are two possible responses to this. One is to say, "Well, maybe I
was wrong about how impossible it will be to control trillions of
assemblers. Can I see the paper so I can check your assumptions?" The
other response is to say, "Yes, but assemblers are impossible, so what
you're doing is stupid." I could answer by defending the idea of
assemblers, but that's just playing whack-a-mole and will waste both our
time. So if that's the response I got, I'd answer that we can't find
common ground after all.
June 17, from Bill Atkinson to Chris Phoenix
Subject: Re: Review Response
Me culpa on the Modus tolens; that was a typo. So was p.277, in a list the
publisher prepared, and whose error I missed in proofing. I'll correct it in the
second edition. Thanks too for offering to post a clarification on the Fouts
thing: it speaks well for your intellectual honesty. Still, "what Fouts really
meant" seems a grudging allowance, and misleading at that. What Roger meant is
precisely what I say he meant - that white matter (rich in neurons with
myelinated axons) may well have different cerebral processing functions than
grey matter (poor in such neurons). You may catch me in other places, but this
isn't one of 'em.
The golf club is another jeu d'esprit, which I thought would give the
technically minded a smile. Joe III has considered the equation K = 1/2MV.V
(sorry, can't get a superscript with this E-mail program). Joe has seen,
correctly, that doubling mass doubles kinetic energy, while doubling velocity
quadruples it. But he's concluded, probably incorrectly, that an ultra-light
golf club moving at a higher velocity would impart greater specific impetus than
a heavier club moving more slowly. This approach has also been adopted by the
Army in its assault rifles: much lighter bullet, much higher velocity. As Joe
says, sniffing a rat, the new club must be tested. I wanted to convey some of
the uncertainty that accompanies all technical development, plus the chance once
always takes of being off base when attempting something radically new.
In the larger issue of whether a nanoassembler is possible, I have to say you're
making inroads on me. When I studied anthropology, one of the concepts that
interested me was acculturation - the fireworks and synthesis that happen when
two world-views collide. If I'm wrong I'll admit it; I'm willing to learn. About
covalency, for example: I worked from the datum that every other bond in the
benzene ring, not every single bond, is shown with the double line.
Beyond the details, however, there are issues here that you've shrugged off too
easily in your quest for a nanoassembler, and other issues that you haven't
addressed at all. Among the former: Rick Smalley. The guy's a beacon of science,
a Nobel laureate, an incredible experimentalist and theoretician: one who's
helped pioneer our knowledge of the nanocosm. I've read the IMM and open-letter
responses and conclude that Rick's arguments, based as they are on extensive
knowledge and genuine observation rather than what-we-think-should-be-in-there,
have in no way been dismissed. In fact I get a strong sense that the open letter
was really addressed to the faithful: Don't lose heart, we're going to do this,
Smalley doesn't know what he's talking about.
I also retain deep misgivings about Drexler's approach. When I read the New
Scientist interview, I thought: We agree! If he'd said that a year ago, and
showed he believed it, I wouldn't have been hard on him. In fact I never started
out to slag the guy. But the more I compared his legitimacy ratio [i.e. public
mindshare divided by scientific achievements] with those of real nanoscientists,
the more annoyed I got. Moreover, the websites with which Drexler is affiliated
continue to mix up new experimental data with what SciAm calls his "quaint
notions." Evidently Drexler continues his game plan of
legitimacy-by-association, appropriating the results of genuine R&D as if they
supported his ideas. It is exactly this obfuscation that I wanted to spotlight.
I still say, with Smalley, that the burden of proof is on you guys to show this
thing is possible. Not calculate it, not print equations on paper, but
demonstrate it. My God, Chris, think what you're trying to do! This makes the
Industrial and Cybernetic Revolutions look like walks in the park. A
self-directing, self-replicating assembler hasn't been achieved on a one-meter
scale. Trying to jump six or seven orders of magnitude down the pipe defies
understanding. I boggle, I truly do.
As for your answer - "Stay tuned, we're working on it" - I also address this
rhetoric in the book. Saying something will happen can never be disproven;
saying something won't can be disproven at any instant. In debating terms, I'm
the guy who's out on a limb. But I'm confident no one's going to saw it off. So
prove me wrong.
And, hell, you may! I grind my teeth to say it, but you have managed to do
something I'd thought was impossible: make me reconsider my position that the
nanoassembler cannot be achieved in principle. I do remind you, however, that it
may well be impossible in fact. In either case, i.e. in both scientific theory
and engineering practice, the burden of proof remains wholly on you.
Another point your review ignored: Even if the nanobot is possible, why bother?
Why break our brains developing something as incredibly fiddly, complex,
unproven, and improbable as a nanoassembler when self-assembly has proven itself
a highly flexible success for three billion years? I have to wonder if,
psychologically, the nanobot isn't simply the unnecessary, hubristic
self-projection of the engineer into the nanocosm - "I'm here, and here to
The day anyone demonstrates a working nanorobot in sequential AFM images, one
with onboard software storage, with interior or even exterior power source and
heat management, I will take out an ad in the New York Times publicly declaring
I was absolutely wrong, and offering a grovelling apology to Mr Drexler, the
Foresight Institute, the IMM, yourself, and your whole community. So do it:
prove I'm mistaken.
I don't think you will. Answered every argument, overcome every objection? You
haven't done enough genuine science to start listing your real problems -
neither experiments to find how the nanocosm really behaves, nor practical
engineering. Calculation might be necessary to make a nanoassembler, but it is
hardly sufficient. Every dollar spent on a technical will-o'-the-wisp like the
nanoassembler still seems to me to be a dollar taken from legitimate
June 18, from Chris Phoenix to Bill Atkinson
Subject: Intro and loose ends Re: Review Response
You have convinced me that you're willing to listen. In that case, I'm
very willing to talk--and to listen as well. Perhaps we can find common
ground after all.
This was getting really long, so after I wrote the answer, I broke it up
into threads of personalities and motivations, feasibility and
desirability, credibility of objections. Leftover stuff is below,
including some science facts.
What Roger meant is precisely what I say he meant - that white matter (rich in neurons with myelinated axons) may well have different cerebral processing functions than grey matter (poor in such neurons). You may catch me in other places, but this isn't one of 'em.
More precisely, white matter is rich in the myelinated axons of
neurons. The cell bodies are all in the gray matter. White matter
alone can't process anything, any more than power lines can generate
power or phone lines can generate signals. But I'm not trying to catch
you on this one anymore--Fouts' statement was pretty confusing.
The golf club is another jeu d'esprit, which I thought would give the technically minded a smile.
I think it'd give the technically minded a knowing smirk, and confuse
the others. It's not trivial to realize that people can't swing their
arms fast enough to make a super-light golf club work. But the idea
that a single buckytube could make a golf club shaft is worse: harder to
know it's wrong, and more directly nanotech-related.
In the larger issue of whether a nanoassembler is possible, I have to say you're making inroads on me.
Cool. And, though I know a lot about technology and science, I don't
know as much about economics and markets and user acceptance. You
haven't made inroads on me yet, but you could, especially in those
areas--though we might have to consider a variety of scenarios of MNT
difficulty without agreeing on which was most accurate.
When I studied anthropology, one of the concepts that interested me was acculturation - the fireworks and synthesis that happen when two world-views collide.
Yep. As long as the two sides are willing to interact meaningfully.
I'm not sure there's much synthesis coming out of the abortion issue.
But I think we can maybe synthesize something useful.
About covalency, for example: I worked from the datum that every other bond in the benzene ring, not every single bond, is shown with the double line.
The difference between single line and double line is not about
covalency. Both bonds are covalent. One is a single bond, and the
other is a double bond (which is stronger, and involves twice as many
electrons). Carbon wants to form four covalent bonds. In diamond and
in many organic chemicals, this is four single bonds. In carbon
dioxide, it's two double bonds. In graphite, it forms three covalent
bonds, and the extra bond kinda' floats. And in benzene, the double
bonds kinda' float around the ring; that's why it's sometimes drawn with
a ring of single bonds and a circle inside. But AFAIK, carbon doesn't
form any non-covalent bond.
Non-covalent bonds include ionic bonds (like in salt crystals) where
instead of two electrons being shared between the atoms, one electron
moves to the neighboring atom and the atoms are held together by
electrostatic forces. Of course this is only an approximate
description. And there are various other weaker kinds of interatomic
bonds/forces: hydrogen bonds, sulfur bonds...
Diamond contains *only* carbon. To convert a tree into diamond, you'd
have to break the covalent bonds between carbon and hydrogen, carbon and
oxygen, etc that occur in the tree's chemicals; and make the carbon bond
only to other carbons; and impose the regularity of the diamond
lattice. This would not be easy--in fact, it would be somewhat harder
than what nanobot assemblers are supposed to do.
Calculation might be necessary to make a nanoassembler, but it is hardly sufficient. Every dollar spent on a technical will-o'-the-wisp like the nanoassembler still seems to me to be a dollar taken from legitimate nanoscience.
Before we continue this particular thread, we should probably talk more
about the likelihood that a nanoassembler can be built, and the probable
utility if one is.
June 18, from Chris Phoenix to Bill Atkinson
Subject: Dealing with objections Re: Review Response
Beyond the details, however, there are issues here that you've shrugged off too easily in your quest for a nanoassembler, and other issues that you haven't addressed at all. Among the former: Rick Smalley. The guy's a beacon of science, a Nobel laureate, an incredible experimentalist and theoretician: one who's helped pioneer our knowledge of the nanocosm.
The nanocosm is huge, and Smalley is very good at exploring a tiny
corner of it. If he expressed an opinion about biocompatibility of
dendrimer drugs, no one would listen very hard--it's nanotech, but it's
not his field. And although bucky structures and diamondoid are both
made of carbon, nanobots aren't his field either. Much of his knowledge
and observation doesn't apply. That's why I'm comfortable evaluating
his arguments for myself on their own merits, without paying much
attention to his reputation. He hasn't given nearly as much scientific
attention to diamondoid mechanosynthesis and nanoscale engineering as
Drexler has. But he's very smart and an excellent researcher; if he
claims to have found problems with Drexler's theories, I'll listen.
I'll even listen to his intuitions and projections. But since he's
applying them in areas he hasn't directly studied, they're at least as
questionable as Drexler's choices of equations and constants.
But some of Smalley's arguments are flat wrong. He talks about fingers
gripping atoms. This has never been proposed, and he should know that.
This particular argument is no more than a strawman. I wouldn't expect
most people to realize this, but there's no question about it. To
dismiss this argument does not require dismissing Smalley's technical
knowledge--only disavowing the use of fingers to grip atoms, which
doesn't hurt Drexler's case since he never suggested it.
For a while, Smalley was saying that chemistry requires dozens of atoms
moving in intricate trajectories in a small space. That was wrong, and
in his field too: burning hydrogen involves chemistry, but only a few
atoms at a time. That makes it easy to dismiss him--perhaps too easy.
I have not seen more technical arguments from him. I don't know what
work he has done to back up his popular pronouncements. But the
pronouncements themselves do not convince me that he's even thought
about it very much. And if he hasn't given it at least a little deep
thought, then Nobelist or not, I'm not very worried about his
pronouncement that it's impossible.
I've read the IMM and open-letter responses and conclude that Rick's arguments, based as they are on extensive knowledge and genuine observation rather than what-we-think-should-be-in-there, have in no way been dismissed.
The arguments about the number of degrees of freedom required are worth
listening to. But here, Drexler has a strong case, and Smalley has
damaged his credibility a little by his earlier "dozens of atoms"
argument. Drexler described several structures in Nanosystems for
holding reactive atoms stiffly a small distance away from a surface. If
Smalley did some work to demonstrate that such structures wouldn't work
as claimed, I'd pay very close attention, and I'd worry more than a
little. But so far, Smalley hasn't seen fit to spend nearly enough time
to criticize Drexler's work in detail.
And meanwhile, we've had at least two very different examples of
covalent chemistry done by scanning probe microscope. In one, carbon
monoxide was attached to iron. In the other, a single atom of silicon
was plucked from a covalent silicon surface and then replaced in the
same spot. Sure, it was a surface that's especially easy to work with.
But if I understand Smalley's arguments, they would've said such a feat
was flat-out impossible. Now that mechanochemistry on tetrahedral
covalent lattices has been demonstrated, I think Smalley needs to decide
how far to narrow his arguments.
I still say, with Smalley, that the burden of proof is on you guys to show this thing is possible. Not calculate it, not print equations on paper, but demonstrate it.
We're getting there. Pieces of it have already been demonstrated. I
agree we won't know for absolute certain sure until it's demonstrated.
But we can start talking now about possibilities, and to some extent
even probabilities and capabilities.
Some of the big problems can legitimately be solved on paper. For
example, assembler control: don't mount the computer, communicate from
outside--and the fact that there are several ways to communicate under
consideration makes it more likely to work, not less. Pressure pulses
are so simple it's hard to imagine them not working.
I do think it's significant that Nanosystems is chock-full of testable
hypotheses, and not one of them has been disproved yet. It's not just a
big book full of equations; it's a big reliable book full of equations
that, as far as we can tell, have been applied correctly.
In either case, i.e. in both scientific theory and engineering practice, the burden of proof remains wholly on you.
I disagree that the burden of proof is wholly on us in the area of
scientific theory. When Engines of Creation came out, the stuff was
largely unproven, and much of it was fantastic. But Nanosystems appears
to establish quite a lot of scientific theory pointing directly at
nanobots. It doesn't seem fair to expect us to do even more work before
being willing to scientifically criticize any of it.
The day anyone demonstrates a working nanorobot in sequential AFM images, one with onboard software storage, with interior or even exterior power source and heat management, I will take out an ad in the New York Times publicly declaring I was absolutely wrong, and offering a grovelling apology to Mr Drexler, the Foresight Institute, the IMM, yourself, and your whole community. So do it: prove I'm mistaken.
What do you think of the Feynman Grand Prize? It's a somewhat simpler
set of criteria, but it's well on the way to what you describe.
Keep in mind that the need for heat management depends on device size.
A single nanobot can be cooled by conduction and requires no explicit
heat management. A solid cubic meter of Drexler's nanocomputers
couldn't be cooled by any means.
I don't think you will. Answered every argument, overcome every objection?
Answered every known showstopper, and a fair number of the practical
You haven't done enough genuine science to start listing your real problems - neither experiments to find how the nanocosm really behaves, nor practical engineering.
We certainly have started listing the real problems. Heat, power,
stiffness, control, design, and some chemistry. We also know quite a
lot about how the nanocosm behaves. We certainly don't know enough yet
that we could train a mechanical engineer to design nanobots. But I
feel pretty comfortable in saying that nanobot design will turn out to
be accessible to engineers. Remember, all we need is one assembler
design plus a few modular functional designs to turn out a whole flood
of products--and the product designer won't even have to know much
(Discussion continues in "Feasibility.")
June 18, from Chris Phoenix to Bill Atkinson
Subject: Feasibility and desirability Re: Review Response
My God, Chris, think what you're trying to do! This makes the Industrial and Cybernetic Revolutions look like walks in the park. A self-directing, self-replicating assembler hasn't been achieved on a one-meter scale. Trying to jump six or seven orders of magnitude down the pipe defies understanding. I boggle, I truly do.
And the Industrial and Cybernetic revolutions aren't even over yet.
Their logical end point--which we're steadily moving toward--is a
convergence of rapid prototyping and automated assembly, with extreme
levels of miniaturization. Hm... sounds familiar. I think that even if
we never worked directly on MNT, we'd get there around 2030 or 2050
The reason for trying to do manufacturing at a smaller scale is that
some things get a lot easier at that scale. At the human scale, it
takes millions of different kinds of operations to make a
general-purpose factory. But if you're working directly with chemistry,
it may take only a few hundred operations, repeated millions of times.
Computers take advantage of this; a single logical operation, NOR (or
NAND), is enough to build an entire computer capable of editing your
Yes, some things are harder at the nano scale. We're not used to
working there. And it has to be kept very clean and orderly. And
bootstrapping will be a pain and a half. But the practical advantages
make it look worthwhile. Bearings with no static friction--this may
sound incredible, but it's not: in fact, mainstream buckytube
researchers have reported it already (though I suspect they were
overeager to see it). Motors with incredible power densities.
(Electrostatics are inherently efficient, and work better at small
scales--MEMS already use them.) Super-strong materials.
But the biggest draw of molecular manufacturing is the ability to build
a wide variety of parts, with a wide variety of properties, and
incredible precision, using only a few operations in programmable
sequence. This is why we think we can get self-rep at the nanoscale
before the human scale.
As for your answer - "Stay tuned, we're working on it" - I also address this rhetoric in the book. Saying something will happen can never be disproven; saying something won't can be disproven at any instant. In debating terms, I'm the guy who's out on a limb. But I'm confident no one's going to saw it off. So prove me wrong.
Give us a decade. :-) This one, we'll have to wait and see on.
But on issues of whether it's worth doing (assuming it can be done), I
think I can give you a strong argument today.
And, hell, you may! I grind my teeth to say it, but you have managed to do something I'd thought was impossible: make me reconsider my position that the nanoassembler cannot be achieved in principle. I do remind you, however, that it may well be impossible in fact.
We know that protein nanoassemblers are possible. But they're hard to
work with, and produce only so-so materials. If I knew all we'd ever
have was protein, I'd probably fold CRN and work on dyslexia or AI.
Are diamondoid nanoassemblers impossible? We can't know for sure.
There are no known showstoppers. There's suspicion from Smalley, an
excellent chemist, that mechanochemistry may have too many practical
limitations. There's partial evidence from Drexler that we can probably
do at least some diamondoid. There are lots of unsolved questions.
Many of them will take hard work, but it's pretty certain that we can
find solutions. Some of them may turn into showstoppers. My own
experience, both as an embedded software engineer and as a nanotech
theoretician, tells me that there probably won't be showstoppers. There
may be crippling costs: at energy costs of $1000/kg (calculated by
Merkle for a very primitive assembler design), not many products would
be worth building. But even at that point, it'd be worth building
weapons, so we'd have to start considering policy. And Merkle's design
can probably be made more efficient by two orders of magnitude once it's
integrated into a nanofactory. Just replace the ratchets with a
reversible logic drive.
Another point your review ignored: Even if the nanobot is possible, why bother? Why break our brains developing something as incredibly fiddly, complex, unproven, and improbable as a nanoassembler when self-assembly has proven itself a highly flexible success for three billion years?
The exact same things are true of digital computers and their software.
Fiddly, complex, and improbable. A bit (logical 1 or 0) is a good
candidate for the most unnatural thing in the universe. A computer does
trillions of bit operations every second, with negligible error. And
why? So that I can move my mouse around the screen and fix typos
without having to retype the whole page. Or look up a phone number
without having to leaf through a Rolodex. Or just play Minesweeper.
But computers are still useful. Improbable: handling bits. But that
lets them deal with any kind of information: movies, weather
simulations, databases. Likewise, a programmable assembler should let
us build a wide variety of products without retooling and retraining.
Complex: A computer has about twelve levels of abstraction from the
silicon to the GUI. We throw away most of its potential power, just so
we can think about fields instead of variables; variables instead of
registers; registers instead of transistors. We throw away more power
by ignoring analog logic values. Likewise, usable MNT will require
intense simplification at every step, and most of its potential power
will be inaccessible. But what's left over will, I think, still be
Fiddly: Like I said, computers do quintillions of operations per
hardware error (software errors are usually more frequent). This is
possible because bits have built-in error correction--the flip side of
throwing away the analog. Likewise, diamondoid mechanochemistry is
hard-edged on purpose: by using only stiff machines to build only stiff
things, we expect to get ridiculously small error rates--enough to
remove a lot of the fiddle. Biochemistry includes about one error in
every ribosome. Test tube chemistry copes with fractional yields at
each step. These processes produce lots of error, and must deal with
it. But mechanochemistry offers the chance to overdesign right at the
beginning, so that we don't have to worry about errors until we get up
to the micron scale. If the theory holds, this will open a new and
potentially very useful design paradigm.
You haven't done enough genuine science to start listing your real problems - neither experiments to find how the nanocosm really behaves, nor practical engineering.
[I was wrong: the books aren't Zyvex projects; they're private projects
of Freitas and Merkle. And the schedule on Diamond Surfaces had slipped
since the last time I looked. See http://rfreitas.com/#Replicators --CJP]
Zyvex is coming out with a book, hopefully later this year, on
diamondoid surface chemistry. There's another one on assembler
requirements. Freitas, one of the authors on both of them, is an
extremely careful and thorough worker. (The other author is Merkle, and
we'll probably disagree on his reliability. But Freitas wouldn't let
anything be published with his name on it unless it was absolutely
solid.) I don't know what's in the books, but I'm eagerly looking
forward to them.
There are sub-problems we haven't looked at much yet. Aside from one
section in Nanosystems and one paper by Merkle, I haven't seen much
discussion of how to make "binding pockets" for sorting or organizing
molecules from solution. (Freitas discusses this in Nanomedicine, but
he's assuming a more advanced nanotechnology.) In theory, there might
be a showstopper here. And although there's been some work on
mechanochemical tools and reactions, we need a lot more.
But in many areas, we've made great progress at breaking down problems
into subproblems that we should be able to handle. I'm now confident
that given an assembler we can make a working and useful
nanofactory--there are no showstoppers left. And my impression, based
on non-detailed conversations with them, is that Freitas and Merkle have
reached a similar stage on the assembler design. [When I double-checked, Freitas expressed
discomfort with a blanket "no showstoppers" claim. --CJP] And if we have a robot
capable of doing general mechanochemistry, programming that robot will
not be too difficult. To me, the biggest question is about funding, and
the second biggest is about sensing--the job will be much easier if we
can actually look at what we're doing as we do it. There's a company
called AngstroVision that claims to have developed a sub-wavelength
non-proximal 3D optical sensing technology that'll go down to voxels of
a few nanometers. They're still in stealth mode, but if they're right,
we'll be able to debug our designs a lot more quickly.
June 18, from Chris Phoenix to Bill Atkinson
Subject: Personalities and Motivations Re: Review Response:
In fact I get a strong sense that the open letter was really addressed to the faithful: Don't lose heart, we're going to do this, Smalley doesn't know what he's talking about.
I've known Drexler for years, and I've talked with him about the
letter. He meant it to be a serious challenge to Smalley: quit
criticizing strawmen and address the actual proposals. I do not believe
he was grandstanding at all. He copied the letter to some futurists and
journalists of course, but most of the list was scientists, government
people (at least one senator), academics, science magazines...
I also retain deep misgivings about Drexler's approach. When I read the New Scientist interview, I thought: We agree! If he'd said that a year ago, and showed he believed it, I wouldn't have been hard on him.
Hm. What he said in the interview is cautious, but there's nothing in
there that backs off from Nanosystems. It looks to me like both he and
the interviewer were focusing mainly on the near-term stuff, and not
talking about diamondoid at all. But he says: "Doing studies of
molecular manufacturing at this point is like studying space flight
before Sputnik or studying fission chain reactions before the Manhattan
Project. But in all these cases you use standard applied physics methods
to study the consequences of known phenomena." He clearly still
believes that molecular manufacturing will be possible and significant,
and that we can already study some aspects of it.
In fact I never started out to slag the guy. But the more I compared his legitimacy ratio [i.e. public mindshare divided by scientific achievements] with those of real nanoscientists, the more annoyed I got.
Most nanoscientists aren't working to gain public mindshare. Drexler
has for decades been concerned about the policy implications of MNT, and
in the mid-eighties decided to go public with his concerns. This was
not a scientific thing to do, and if judged in a scientific context, it
is certainly annoying. But it doesn't mean the science he's done is
invalid. It also doesn't mean his science is any more valid, just
because more people pay attention to it. In fact, it has no bearing on
the validity of the science.
Drexler's going public is a lot less self-serving than some of the other
scientists who have done so. For example, there's a group of dyslexia
researchers who have developed and commercialized an auditory-based
method of treating dyslexia. They have gone public with the claim that
dyslexia has nothing to do with vision--which is demonstrably
incorrect. But they get lots of news stories and lots of publicity for
Drexler didn't do it to get funding, or money, or even fame. He did it
to get people talking about the policy implications of MNT. If I were a
nanotube or dendrimer or quantum dot researcher, I might be annoyed and
envious at his publicity. But as a member of the public, I'm glad he
did it. It might have been better if he'd found someone else to do the
publicity side. But I'm not sure anyone but a researcher would have had
the credibility at that early stage.
Moreover, the websites with which Drexler is affiliated continue to mix up new experimental data with what SciAm calls his "quaint notions." Evidently Drexler continues his game plan of legitimacy-by-association, appropriating the results of genuine R&D as if they supported his ideas. It is exactly this obfuscation that I wanted to spotlight.
The legitimate research is supporting more and more of his technical
claims. Since many of his claims have been attacked in the past, I
don't blame him for pointing out when the research contradicts some old
(or not-so-old) criticism.
There's a subtle point here. The external research supports the
technical claims. He reasons from the technical claims, to product
claims, to policy claims. The external research does not (yet) directly
support product or policy claims, and he doesn't say that it does. But
the policy issues are important, and he feels he has to talk about
them. And to defend the technical claims that the policy issues are
based on, he has to talk about external research. Maybe he could make
it clearer--but he'd be misread no matter what he did. I don't think
he's deliberately trying to obfuscate.
I have to wonder if, psychologically, the nanobot isn't simply the unnecessary, hubristic self-projection of the engineer into the nanocosm - "I'm here, and here to dominate."
Close. The nanobot is certainly a projection of engineering into the
nanocosm. As such, it's unnatural; so it must be invented by humans; so
in a sense it's a self-projection. But I don't think it's hubristic.
Now, if you want to talk hubristic, look at Three Gorges Dam. But
that's a projection of the politician, not the engineer.
Hubris depends on cultural standards. Back in the early 1800's, the
idea of making a railroad engine go at 40 MPH was the height of
hubris--except to the engineers. Today, hubris isn't even an issue in
trains, cars, airplanes, even rockets. No one says, "Isn't NASA
hubristic to try to make rockets go faster than 10,000 MPH?"
But engineering, and to some extent engineers, are divorced from
cultural standards. An engineer doesn't ask, "Is it hubristic?" but,
"Is it practical?" From the outside, this can look like either hubris
or humility--depending on the observer, not the engineer.
As for unnecessary, we'll have to wait till we know more about the
practical value. But I hope I can convince you that it's reasonable to
think that the practical value might be high. (See the Feasibility
June 19, from Bill Atkinson to Chris Phoenix
Subject: Re: Intro and loose ends Re: Review Response
You can imagined me blackened and smoking, having just picked myself up off the floor;
you keep blowing me out of my chair. I'm so glad we're taking the time to E-mail each
other: I can feel my head stretch hour by hour. At this point I'm going to go for a
run and think about some of your points. More later.
PS. You heard about the agnostic-dyslexic-insomniac who stayed awake all night
wondering if there was a Dog? (Dyslexia all through our family)
June 19, from Bill Atkinson to Chris Phoenix
Chris: here's my most recent posting on The Well -
I've had the good fortune over the last week to develop a parallel
dialogue on some of these points with Mr Chris Phoenix, who by day is a
mild-mannered imbedded-software engineer and by night puts on his
superhero costume and flies off to make detailed calculations about
molecular assemblers. We started off by beating each other up on
various points, moved to a grudging mutual respect, and are now (I
think) on the brink of a very exciting synthesis of ideas.
Anthropologists call the process acculturation, though it often feels
like a very noisy and painful collision.
My position in the book Nanocosm is that most of the nanobot crew,
along with their fellow travellers the corpsicle people, are off in
la-la land, doing endless calcs and getting no nearer to a real device.
I'm not convinced I was wrong yet: there are many points I made in the
book that have not been fully answered to my mind. But I'm starting to
suspect that with certain modifications, the nanoboosters could make
contributions to legitimate nanotechnology - giving vision to the
traditionalists' spadework, while informing their own work with more
discipline and rigor.
- I'll send you some of the comments on this - I can also post your own
response(s) if you like. That'll at least save you a tenner.
June 21, from Chris Phoenix to Bill Atkinson
Subject: Re: FYI
This is a very encouraging start. I look forward to addressing the rest
of the points. Either I can read the book again and address them
chronologically, or (which is probably faster) you can ask them in the
order you're most interested.
Yes, I'd like to see the Well discussions. Thanks!
More discipline and rigor... the rigor of MNT work varies widely, even
within the same author. Sure, a lot of the stuff is blue-sky
daydreaming. But a lot of it isn't. I predict that you'll be surprised
to find how much of it is already backed up by rigor. Of course, some
of it is backed up by rigorous *theory* without experiment. But some of
the theory is quite reasonably extrapolated from experiment. I think
it'll have to be assessed case by case--a big job. But even if it's
unsupported theory, it's still good for giving vision; there are a lot
of areas where the theory is not brittle and will work even if it has to
I used to be an embedded software engineer; I quit a few years ago.
Sorry if I said something misleading. I worked at Electronics for
Imaging, doing high-end embedded software, for six years. Then I quit
to do dyslexia correction and research. Now I'm spending all my time on
CRN. And yes, I consider this dialog as CRN business. It's important
to establish whether nanobots can really work, and if so, what they
might be capable of. (Would you consider CRN a neutral site for posting
I got another email from Fouts; it seems that you are exactly right in
your understanding of what he meant to say. He meant to include the
neuron cell bodies. So either Fouts doesn't know what "white matter"
means, or I don't, or there are two usages. I'm still researching to
learn which, but I've found at least one reference to the meaning I
thought it had. At this point, I'm thinking of deleting that complaint
from the body of the review entirely, with an asterisk'd footnote saying
that I removed an unwarranted criticism of you, since you interpreted
Fouts correctly, and a clarification of what I originally wrote.
Unfortunately I may not be able to do this till Monday.
[A few emails are omitted, where Bill and Chris discussed the logistics of posting this dialog, and Chris's correction of the "white matter" question in his review. This message, from Bill, sums it up pretty well.]
June 27, from Bill Atkinson to Chris Phoenix
Subject: Re: Intro and loose ends Re: Review Response
Hi Chris! Herewith at least the start of my responses to your last notes.
I like your idea to flag items commented on later in the dialogue with a mark of some
kind - maybe a line drawing of a Tennessee wrassle between an engineer and a science
writer? Or, here's an alternative possibility: run all the original text as is, with
editorial comments where necessary in square brackets. For example, you could insert a
note saying "comment since verified/amended in personal communication with _____" This
would keep the dialogue fresh and avoid the impression that we've all been running
around ex post facto, shoring up our positions! Above all I want to retain what we've
created here: the feel of a work in process - the acculturation I refer to at one
point, where extreme opposites start to grope toward common ground. Like a live debate
before an audience: a work in progress, rather than an exchange of doctrines in
The Fouts material seems increasingly irrelevant to our main points. As he says, he
himself isn't a neurologist; nor are we. So let's go with what we've got, and move on.
Interesting stuff, but tangential.
Thanks for the chalk-talk on covalent bonds. Perhaps because I interviewed so many
nanotechnologists who came to their research via chemistry, I tend to think of
nanotech, at least today's nanotech, as chemically based. Interestingly enough, I'm
getting feedback from several of my chemist interviewees that there's a push on, at
least within the US NNI, to fund nanotech R&D that's done mostly (or even only!) by
In the book I wrote glowingly about nanotech's ability to synthesize a variety of
disciplines into something science hasn't had for several hundred years: a unified
view of nature - a single discipline, looking at a single thing. This might still
happen; but I may well have underestimated the foot-dragging by existing disciplines.
There's an enormous amount of vested interest within those disciplines in keeping
themselves walled off, and regarding other disciplines as unworthy competitors for
My next response will, at your request, focus on some of my own areas of expertise,
including marketing communications. I have a hunch this will surprise you.
June 27, from Bill Atkinson to Chris Phoenix
Subject: Re: Feasibility and desirability Re: Review Response
One thing I didn't boggle at was the nanoscale frictionless bearing. It's already
demonstrated in chemicals called rotaxanes, to which I refer in the book. So yes,
there's a precedent for the actual improbable (as opposed to the plausible
impossible). On the high nanoscale / low microscale border, the bacterial motors that
power whiplike or rotational flagella would be unlikely if friction loss were any
higher than negligible, or (another way of stating the same thing, I suppose)
oscillatory or rotational efficiency were anything less than extreme - 100% less
I will grant you that sheer fiddliness, by itself, is not necessarily a barrier to new
technology. Even steam, when you examine it historically, was incredibly complex.
Engineers in our grandfathers' or great-grandfathers' day weren't less intelligent, or
even less educated, than we: they just worked on different data sets. Try finding a
crew today who could break down, diagnose, repair, and reassemble a Pacific 351
locomotive in six hours, a turnaround that was routinely done in a good shop c.1938.
Or locate a good flint-knapper who could take a lump of chert and give you a
laurel-leaf point that would kill a mammoth.
My problem with the MA technology you describe is that it's still completely
imaginary: all calc and no grease. Workable technology, indeed any technology worth
the name, has a fairly short cycle between drawing board and lathe: between idea and
trial. In fact the better the technology, the shorter the oscillatory period. It's a
tennis game: imagine/try/iterate. MA work seems to have stalled in the Imagine
half-cycle. Thus denying itself any feedback, it's evolved wilder and wilder designs -
absolutely none of which have been put to the test. It was my growing annoyance in
realizing this that really led to my complete loss of patience with Eric Drexler. The
guy really has dispensed, it seems to me, with the necessary reality checks.
As have you. You tell me your recent work has "solved problems." No: it's only done
sums and solved equations. A true solution requires demonstration: and that in turn
demands that fateful encounter with reality, in the person of the nanocosm. Elsewhere
both you and Drexler like to cite the Manhattan Project, with its endless calcs and
forecasts. But the Project met the description I set out above, for an actual
technology: a high-frequency iteration between theory and practice, paper and nature,
screen and reality. Kept isolated, untested and therefore unproven, the greatest calcs
in the world are infertile. I say, with Gibbon's 1781 critic: "Another damned, thick,
square calculation! Always scribble, scribble, scribble! Eh! Mr Phoenix?"
No doubt you'll say Yes, but it's a preliminary: we're coming to the shop work. Good,
then: I'm glad to hear it. Do it. Until you do, you won't know which of your
calculations are workable, and which just seem workable - and for all their internal
elegance, are either unworkable or simply untrue. I keep coming back to this point,
only to be told: Give us time - five years, ten years, fifty years. But something
admitted to be vital, and not yet done, in a way suspends dialogue. Come back to me
when you start your shop work. Till then I'll stay a skeptic.
That being said, I have to admit - Gad, this is dragged out of me! - that some of your
ideas are conceptually fascinating. I like the way the extreme complexity and
intricacy of your world view's design and construction may be - I say, May Be! -
offset by the extreme simplicity of its operating methodology. There seems a parallel
between your view here and the main point I was trying to make in my CA discussion:
ie. that at the nanoscale, manufacturing success absolutely demands a stripped-down,
cut-rate, compressed, streamlined minimality in directive algorithms. Otherwise the
whole concept of the MA collapses under the weight of its own complexity.
Another reality-based question. Can you really create a nanoscale clean room, without
a single bit of unwanted crud in it, ever? Because a single atom in the wrong place at
the wrong time would gum up the works. This may be your tallest technological order.
Again, I'll believe when I see. But I grant (aghh! the tooth is pulled) it may be
conceptually possible. Nice calc.
But again: Why bother? The super-strong materials to which you refer are already being
achieved by self-assembly, using sputtering (AIST Tsukuba) or other techniques for
self-deposition and self-organization (Integran Technologies, Toronto). It's as if
you've designed a nutcracker that works by firing a hammer into orbit, then falls to
Earth to smash the walnut. Even if it's possible, why bother? It's unnecessarily
complicated. Worse, even from the eng-calc viewpoint, it's inelegant - it doesn't tap
into that organizing information that seems evenly dissolved throughout nature. Why
compel when you can persuade? The molecular assembler seems nothing more than a
steam-propelled gondola, or that orbital nutcracker.
That's assuming, which I do not believe, that it can ever be shown to work. For
example, I don't see how you can claim the nanoassembler will have universal
application when it will work only on, and with, stiff materials. "Wide variety of
parts and properties"? Only if they're stiff! (You can paint my Ford any color, as
long as it's black.)
We could throw rocks at each other forever on this one. That's why I think my later
response is going to surprise you. I think this whole issue could give you folks your
only workable chance to get the respect that the mainstream scientific communities,
and I too, have so far denied you.
I know, I know - What a tease!