Re: Roaming Assembler Size/Replication Rate?

From: hal@finney.org
Date: Mon Jan 15 2001 - 21:48:53 MST


John Marlow writes:
> Thanks. Basically I'm thinking of an all-terrain, omnivorous kinda
> guy--water, rock, soil; wouldn't really matter. Could be a single design, or
> something which produces environment-specific assemblers (water, rock, soil, etc.)
> if that would reduce size/complexity and, thus, increase speed of
> replication. Something stripped down for speed, but still adaptable. Larger
> size/number of arms will at some point prove counterproductive in terms of replication
> interval.

The design for such a beast is far beyond anything I've heard of people
attempting, even in the broadest terms. Omnivorous goo will be a
difficult problem. Consider the energy considerations: some things
cost a lot of energy to take apart, like rock. We may or may not be
able to reclaim the energy in putting the pieces back together. Plus,
some materials will be rich in what the goo likes to eat, while others
will be missing crucial elements except in trace amounts.

If you look at Freitas Ecophagy paper you see he did look at some
special cases that might be easy, like gray goo which only eats asphalt,
or only eats rubber tires. By having a fixed "food" the design becomes
much simpler, and if it is something that civilization depends on and
is standardized around the world, the goo could be maximally destructive.

> If I'm not mistaken, Drexler's logic (assembler arm can move back and
> forth 50 million times faster than a human arm (approx 150 million times a
> second))--times however many arms there are--has gotta yield an astonishingly
> fast replication rate--much faster than Freitas' stated rate. Particularly as
> Drexler envisioned a larger assembler (100-150 million atoms). But then
> Freitas' work is more recent than EoC, which is where I read this.

On basic scaling principles, a human arm is about a meter, while the
assembler arm is about 100 nm, a factor of 10^7 difference, which would
suggest that the assembler arm could be 10 million times faster than
the human arm. How fast it can construct things would then depend on
how efficient its movements can be, how long it takes to form bonds,
and so on.

> If (for simplicity) each back-and-forth-cycle is moving only one atom,
> aren't we talking replication of a 150 million-atom assembler in 1 second? And
> 9.3 replications of a 16 million-atom assembler?

I think Freitas was being conservative on the slow side, saying that even
if it "only" did 1 million atoms/sec you could still get macroscopic
quantities sufficient for medical use in an hour or two. You on the
other hand want to be conservative in the other direction and say that the
gray goo could potentially reach macroscopic size in seconds or minutes.

> I'm trying to get a handle on just how fast the things can reproduce
> initially, i.e.--before their descendants can't get out of the way fast enough to
> avoid slowing them down. (By which time the maximum replication rate would
> only be attainable by the "leading edge" assemblers.
>
> Any thoughts?

Your best bet is to pick a range of numbers and see how the scenarios
play out depending on whether we turn out to be on the high or low end
of the range. That is roughly how Freitas did it in his Ecophagy paper.

Hal



This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:04:55 MST