From: Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Date: Sun Aug 19 2001 - 06:14:06 MDT
On Sat, 18 Aug 2001 hal@finney.org wrote:
> Well, the specific point is that Whitesides said that the flagellum is
> powered by ATP, and Robert is pointing out that it is powered by a
> proton gradient instead (some cells apparently use sodium ion
> gradient).
Yes, but though a raven is perhaps not very much like a writing desk I
have a very nice collection of antique chess pieces.
I've read it, and it is an astonishingly vacuous piece. It reminded me
vividly as to why I stopped reading SciAm many years ago.
> Whitesides' larger point is quoted mostly above, that the similarity
> between the flagellar and electric motors is "largely illusory" because
> electric motors are magnetic. Everything about their design is oriented
Electric motors are either electric or electrostatic. Electrostatic motors
are described in experimental books for teenagers published decades ago.
Saying that electric motors are impossible at nanoscale is about the same
as saying that houses cannot be built because people cannot walk through
walls. That's true, as long as one keeps ignoring doors.
> towards organizing the magnetic fields to induce rotary motion. However
> magnetic fields don't work that well at the nanoscale and do not drive
> the flagellar motor. Hence his claim that they are not really all that
> similar in their operating principles.
Of course it would be stupid to use flagellar rotor to drive a nanosub,
unless one is limited in one's engineering to hybrid designs, or strives
for maximum biocompatibility (this would be a major point in using
biological components for coating hull and propulsion).
> Freitas points out that bacteria smaller than about 0.6 microns don't
> swim, below that size "locomotion has no apparent benefit" (9.4.2.4 of
> Nanomedicine). So Whitesides' error was in misunderstanding the
> intended size of nanobots; his reasoning would be correct for the 0.1
> micron size he was describing.
I guess if one exerts a lot of effort one can salvage parts of the
article. It would have been truly difficult to write a piece utterly
devoid of content, the author is good, but he's not that good, and SciAm's
editors still haven't completely flatlined on the EEG.
> I don't think that the rigid rods would have to be all that bulky.
Bulk becomes relative when you're looking at the reaction site. Try
manipulating a 1 mm objects with your human fingers meaningfully. Then try
manipulating two or three such objects meaningfully, using several people,
on the surface of a flat table, beyond where your fingers cannot go.
Moreover, we're not just talking structural components, you need
sensorics, you need cooling and fresh/spent moiety ducts (unless you
retract and cycle them offsite), you need power cables, while you have to
keep the mass down so you natural oscillation frequency doesn't become too
low, as you want to attain maximum processivity at given amount of juice.
I really recommend you download VMD, load some of the notoriousdiamondoid
parts (neon pump, bearing, manipulator), switch on space fill, and drag
them around, particulary loading several copies of the manipulator and
arranging them in front of a diamond wall (where the pump is embedded in)
will be illuminating.
Of of course the manipulator is just a toy, as it is lacking critical
functionality. The tool tip is empty, and it doesn't have sensors, and the
concentric coranulenes on the bucky rod don't have rods attached which
lead to bucky linear actuators.
> Merkle is just showing the end structures and they are quite thin.
Exactly, and this is why these structures are misleading.
> You don't have to get too fat to have strong rods, and in fact the tools
> used by Merkle don't seem to require huge forces anyway.
The point is not huge forces, the point is to maintain sufficient rigidigy
that thermal noise and rapid cyclic oscillations will not interfere, which
still not suffering from crowding.
> As for speed, that's not really an issue at this level of design.
> Merkle's main point was to show that you could come up with tools that
> would do the job.
The longer I think about it, the more I dislike the downsized robot arms
paradigm. The armies of tethered (juice, inks) robotic walker on a flat
surface nanolithoprinting small structures and manipulating (splicing,
welding, modifyng) random bucky parts seem to be pointing in the right
direction. The metaphor of picking up and putting down atoms is really
lousy, even if it's just formally so (transfer of a single group from an
activated moiety). You don't have to do that to self-rep, though it would
be of course nice if you could switch to slow-processivity but
enhanced-control mode for some critical components.
> The point of Merkle's analysis was to show how you could have a tool
> set that could build carbon structures, and, most importantly, also
> build more copies of the tools. This is the crucial element necessary
I don't see that anywhere in Merkles paper. He describes the reaction
site, he describes the structure around it, but always as isolated
occurences. He never tries to walk through a whole deposition circle in
machina, with all the strings attached. The paper is important, but he
most assuredly does not show how you could build more copies of the tools
with the tools he described.
So, it's too early to uncork the Dom Perignon magnums yet.
> for self replication and to my knowledge no other paper has showed a
> specific design that could accomplish this.
Correct: no paper has showed a specific design that could accomplish this.
I hope I'm going to see such a set of papers in the coming decades. The
next thing to do will be a doable bootstrap scenario (which will bee very
hirsute), and then we go into teotwawki mode.
> "Computronium" is not likely to be an early product of nanotech IMO.
Oh, computronium is basically round the corner, at least early 2d
computronium. It could be as little as 5-10 years away.
> It seems to be a very advanced material. To build mass quantities
Oh no, it's not a complex material at all, in fact the more advanced
versions of it will become simpler, as we learn to cut corners on
irrelevant aspects of the design.
It's molecular crystal, a hardware implementation of a specific cellular
automaton, with a probably very simple rule (few 10 gate complexity in
modern lingo). Simple orthogonal structure both at crystal and cell level,
no movable parts. It is most efficiently assembled by self assembly via
complementary patches, which suggests doing it in biology first. The
elementary cell of the crystal would be about virus sized, of course you
can shrink it dramatically if you use buckytronics, which can also stand
more heat, or work at cryogenic temperatures.
> of computronium you must have already solved all of the construction
> problems, so it will not work as a design strategy to rely on this
> kind of massive computation to solve the low level problems. (OTOH it
> may be possible to build sufficiently massive computers using bulk
> technology.)
As long as performance curve hasn't saturated too badly, you can stick to
Beowulfery, then you have to invest into a hardware force field engine
done in dedicated hardware. COTS has better payoff, that's why we don't do
this yet. But after a while we will.
-- Eugen* Leitl leitl
______________________________________________________________
ICBMTO : N48 10'07'' E011 33'53'' http://www.lrz.de/~ui22204
57F9CFD3: ED90 0433 EB74 E4A9 537F CFF5 86E7 629B 57F9 CFD3
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:09:55 MST