From: Carl Feynman (carlf@atg.com)
Date: Sun Aug 31 1997 - 11:35:49 MDT
> = Nicholas Bostrom
>> = CarlF
>> > = Nicholas Bostrom
>Carl Feynman wrote:
>
>> >Isn't the
>> >design work fairly tractable (Drexler has already produced some nice
>> >designs) and it is mainly the lack of molecular tools that prevent us
>> >from starting building things?
>
>> (1) The minimum self-reproducing device (mycoplasma genitalium) seems to
>> require about a million bits of information. ... a car or ... OS/360
>
>There is a difference between the three systems you mention, on the
>one hand, and nano self-replicator on the other. Mycoplasma
>genitalium, automotive vehicles and comersial software are all
>required to be fairly optimized. To build an optimised nano
>self-reproducing device would be much harder than simply to make
>something useful that can replicate.
We're talking about gray goo here, right? That has to live in the wild,
without getting all its little cogs and conveyor belts clogged up with
natural molecules that just happen to fit into its various nooks and
crannies. I don't see any way to solve that kind of problem without
actually putiing it in the world and seeing what fails. If you're talking
about something that can only replicate in a sterile tank of ultrapure
feedstock, I'd be willing to believe that it could be less 'optimised', but
it is also far less potent economically and militarily.
Further, a certain minimum level of 'optimality' is needed for a
self-reproducing device to function at all: if the mean time until failure
is less than the mean time until reproduction, growth is impossible. Thus,
a self-replicator is analogous to an orbital rocket: there's no use for a
launcher that only gets up to 90% of orbital velocity.
>For example, a universal Turing
>machine has been constructed in Conway's Life world. The entity is
>very big and it was hard, but nothing near a thousands of genius-year
>task, to do it. The feasibility stems from the fact that you have
>identical components that you can put together into bigger identical
>components, and so on, and at each step you need only consider the
>apparatus at a certain level of abstraction. If this is the right
>analogy for nanotech, then the design work would seem tractable, once
>the right tools are there. But I will take your opinion on this
>issue into account in my future thinking. And debugging is also a
>complication.
I agree that Conway's Life self-replicator design could be carried out to
completion in a short time. However, it is not a good analogy. The Life
world is perfect, so engineering reduces to mathematics. Moreover, Conway
designed his machine to function in an infinite vacuum, so the complexity
produced by the impingement of the world is not present. I doubt that a
Life self-replicator could exist in a sea of random pixels, inasmuch as any
design for Life machinery I have seen will be completely destroyed by a
single erroneus pixel.
>
>
>> (3) Drexler and Merkle are two very (very!) smart guys. They have labored
>> for years, and designed what? Some bearings, a transmission, a PLA, and a
>> Stewart platform? And you claim this shows how easy it is? I'd hate to see
>> an engineering task you considered hard!
>
>Those are again optimized designs.
No they aren't. Thay are pretty much the first designs they came up with
that didn't explode in simulation. And they totally ignore issues of
manufacturability, optimization of peformance, radiation tolerance, failure
rates (short of instantaneous explosion), and some other things I haven't
thought of that will turn out to be totally critical once we try to build
these machines. All these things can be ignored for a proof-of-concept
simulated prototype, which is what they needed to prove their point.
>What about the rod logic
>computer?
The only part of it designed to atomic precision was the PLA (programmable
logic array) I mentioned above. Even there, only the rods were designed to
atomic precision, not the frame that holds them, so maybe I shouldn't count it.
>And I didn't Drexler give it as an exercise to his students
>to design an autoreplicator when he taught a course at some
>university? I am not sure about how detailed these designs are,
>though. (Could anybody tell us this, please?)
I wasn't in that course, but I've read Nanosystems closely, and the
replicator design in there is at a fairly high level of abstraction. It's
detailed enough to convince me that nanotechnology is a feasible technology,
which was its purpose. It is by no means designed to atomic detail.
More later. Must sleep.
--CarlF
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:44:47 MST