From: Eugene.Leitl@lrz.uni-muenchen.de
Date: Tue May 01 2001 - 05:48:23 MDT
CurtAdams@aol.com wrote:
> Nanotech has done, to date, nothing. Nobody has even a simple nanomachine
I wouldn't put it in so absolute form, but guessing that you're not
using the lately inflationary diluted "nanotechnology", but meaning
molecular manufacturing, you're of course correct.
> going. Until somebody demonstrates machine-phase is even *possible*
Well, we do have parts of NEMS, especially those derived from
graphenes. Have you missed the jack-in-the-box electrostatic
linear actuators, which are made from concentric buckys and
use van der Waals forces as springs to pull back what you
protracted with a potential? That was flashy, as was the
selective fractionation of bucky batches through electric
pulse lysis of metallic fraction of the tubes, while transiently
biasing the semiconducting fraction into isolators.
While still expensive, production of buckys is being rapidly scaled
up due to their anticipated use in high-end structural materials.
Now buckys are relatively large molecules, but they're molecules.
It doesn't take a Merkle to realize that you can probably use a
MEMS microfab to fractionate these things (not necessarily
destructively) with sufficiently high processivity for practical
use (desktop device with a throughput of gram/hour, scalable to
larger sizes).
Maybe this isn't the catholic bootstrap route, but that was
exactly my point. The vein's so rich you get showered with emeralds
wherever you dig.
> it's way premature to expect it to even be practical, never mind superior
> to liquid-phase "nanotech" - i.e., the biotech being belittled.
Strictly speaking, we do know that machine phase exists. It's just
a fancy version of what biological enzymes do, after all. I can
cite you paper and verse, people have been doing chemistry with
the crude proximal probe tools we have. The question is, is the
repertoire of mechanosynthetic reactions sufficiently rich for a
structure to be able to deposit a copy of the structure itself?
Given how much we found poking around so little, the answer seems
to be a tentative yes.
But do we need machine-phase self-rep capable nanotechnology? Sure,
it would be nifty. But computronium built with self-assembling
molecular electronics is accessible with classical chemical and
biological means already.
> >Ditto for AI.
>
> AI can't even do grammar checks yet. Moore's law may pull it out yet. Or may
> not, depending on how efficiently Mother Nature used our neurons and how
> high a frequency of gamma rays we can focus.
There are two disparate issues at stake here: hardware and software.
Moore is about integration density, and will be going for about a decade,
and then after that, because people have finally started doing R&D
into molecular electronics in earnest. I don't know why it took two
decades to realize that's a worthwhile goal, but the train is picking
up steam, finally.
A less widely known law indicates that the ROI on transistors as
measured in CPU performance has started to run into diminishing
returns. The reasons are many, and most of them architectural. This is
a good thing, because it states that sooner or later people will
be forced to abandon the current axis of impetus, and explore a
hitherto negliged side alley. The results of this architecture change
will result into dramatic improvements of real-world performance
from the same amount of transistors.
Incidentally, this architectural shift should take at least a decade,
and coincide with the advent of molecular electronics. This is even
better, because the new architecture due to their orthogonality, small
grain size, scalability and intrinsic self-healing (through adaptive
reconfiguration) capabilities will allow even primitive molecular
electronics to be rapidly deployed.
The software issue. Well, the software issue. I do not expect
anything from the mainstream. Comp. sci. people, software engineers,
that's you. Sure, algorithms will be translated into circuits --
a desktop version of what silicon compilers are doing now. If you
undust your wizard book, you'll see how it's done -- it's not hard.
Swapping out pages will swap out circuits, not code. Crunch will
fall out of the bunghole, and be used for fancy dumb robots.
None of this will suffice to create naturally intelligent machines,
who're smart enough to be able to compete with us.
> I'm all for planning ahead; but none of the advanced stuff we talk about is
> imminent. AI is the only possibility for this generation. Nanotech is still
> a speculation. Biotech may produce wonderful stuff soon; but we're not
Strange, in my particular universe molecular circuitry is knocking on
the doors, and AI is soon to follow, unless we outlaw darwin in machina.
> about to improve brains yet because we don't know how, and we'll need
> to be *very* sure before we do it. Even an AI "singularity" won't be a
Well, we do have smarter mice, and smarter worms now. People are about
to sequence the chimps, and look at the delta in gene expression in
their brains, so I wouldn't be betting on a timeline. But given
how long it will take until the technique is validated, morally approved,
widely deployed and results into more IQ points for the average child...
The other fields should be noticeably quicker on the time scale.
> singularity. We'll shift from a 20-year doubling time to a 18-month
> doubling time, until Moore's law runs out. It'll be like the whole world
I don't expect the Moore's law to run out for a long time. If it's
timed right, molecular electronics will smoothly take over when photolitho
saturates.
> is a successful startup company for a few years to a few decades;
> then it will be over.
Ha! MorePolice, arrest this man. He's acting defaetist in public,
his extropian license needs to be revoked.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:07:24 MST