From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Sun May 12 2002 - 15:31:22 MDT
On Sun, 12 May 2002, Lee Corbin wrote:
> First, would someone define computronium? From context, I thought that
> I knew last year what was being discussed---now I'm not so sure.
>From Anders "C" page: http://www.aleph.se/Trans/Words/c.html
> COMPUTRONIUM: Matter supporting computation, especially artificial
> substances suitable for high efficiency computation. [Eugene Leitl].
>From Eliezer's CFAI glossary: http://www.singinst.org/CFAI/info/glossary.html
> Matter that has been transformed from its natural state into an optimized,
> maximally efficient computer. (A true Extropian would argue that this is
> matter's "natural state".)
> What constitutes "computronium" varies with the level of postulated technology.
> A rod logic nanocomputer is probably too primitive to qualify as computronium,
> since large molecular aggregates (hundreds or thousands of atoms) are used as
> computing elements. A more archetypal computronium would be a three-dimensional
> cellular automaton which attached computational significance to each individual
> atom, perhaps with quantum-computing elements included. More exotic forms of
> computronium include neutronium, Higgsium, monopolium, or - my personal
> invention - an interlaced structure of positive-matter and negative-matter
> monopolium wrapped up in a fractal Van Den Broeck warp. (The total mass is
> zero, so the whole doesn't undergo gravitational collapse. If paired negative
> and positive matter can be manufactured in unlimited quantities, the fractal
> Van Den Broeck warp can continue extending indefinitely and exponentially.
> Threading the system with wormholes keeps latency down. And the whole thing
> fits in your pocket.)
[As an aside -- for some of the more exotic materials such as neutronium,
Higgsium, monopolium, to qualify as *real* computronium there are some
not-so-insignificant issues of lifetime and I/O to be resolved. Somebody
really needs to keep an eye on Eliezer -- not only is he going to be
downright dangerous when he gets older, but he will try to slip things
(such as the above) past us when we aren't looking...]
Looking back in my notes, I seem to find that Robert Freitas "invented"
the term on May 18th 1999. So its almost 3 years old now -- up to about
200 Googles. I suspect this was first use since Robert has the sometimes
annoying habit of inventing words.
(If anyone knows of an 'older' reference than that I would like to
know about it.)
So you understand the history of where computronium came from you
have to go read the Matrioshka Brain paper(s) and think about them
in more detail. It became apparent to me when developing the concept
that you could have computers operating at *vastly* different
temperatures ranging from perhaps 10K to 3000K made out of very
different materials with very different architectures (no magic
physics here (such as Eliezer is leaning towards), simply good
nanoengineering.
Furthermore, when thinking about *what* those computers would
think about, it became clear that computer architectures can
be optimized for the computation they are doing. The Japanese
GRAPE computers and the Transmeta being cases where that ranges
from the extreme hardware level to the firmware level. The
architecture can vary at an even higher level such as computers
optimized to run neural networks vs. computers optimized as
cellular automata.
During the development of the MBrain ideas, it became clear that
one would have a tradeoff -- do you take a longer amount of time
computing a desired result using the default general purpose
hardware that happens to be at hand -- or do you take some
amount of time computing the "optimal" hardware architecture,
build that, do the computation using it, and get your results
sooner than you would using the general purpose hardware.
So "computronium" in my mind is always semi-optimized
special purpose hardware. It may always be in a state of flux --
being rearchitected and reconstructed for new problems one
wants to think about. It is in a sense "living computational
matter".
Computronium is always up against a couple of limits --
element and energy abundances. If the type of technology
you want to build requires specific elements -- and they
happen to be relatively rare -- you may not be able to
build as much of the optimal computronium as you desire
without breeding said specific elements (that pushes
you into time and energy constraints). Furthermore as
you produce increasingly exotic computronium for specific
purposes there may be a negative consequence of lengthening
scene change times -- it takes *time* to convert all the
gandolinium you needed for your spin state based quantum
computer back into sodium for liquid cooling in your rod
logic computer. If the energy requirements/losses during
scene (architecture) changes are high enough you will
fall back on a more general purpose architecture. In
reality I suspect we get different architectures optimized
for thinking about different things (presumably like the
human brain is today).
I don't see these problems being solved until someone
comes up with an optimal theory of what advanced intelligences
want to think about and the best architectures for accomplishing
that. Thats a pretty post-singularity event in my estimation.
Robert
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:14:01 MST