From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Tue Jul 06 1999 - 17:19:00 MDT
> "den Otter" <neosapient@geocities.com> wrote:
>
> people to visualize and thus better understand transhuman concepts,
> from nanotech to Jupiter Brains. I'll give it some more thought, though.
>
J-Brains are an obsolete concept at this point. To start with
they were a non-optimal form. Something the size of Jupiter
cannot use all of the power available from a star (without
melting) and if it were even relatively "solid" as Jupiter
is, you would waste most of your energy pumping around the
cooling fluid for your computers. [Friction losses mount
up when you circulate fluid through something the size of
Jupiter.]
Following in the tradition of "you need a cool name to
popularize an idea", e.g. "Black Holes", I have christened
an optimal computing architecture in the post-singularity
age - a "Matrioshka Brain" (M-brain). [Matrioshka's are
the carved wooden nesting Russian dolls.] This will be
a "legitimate" scientific concept (if there is such a thing)
as it will be discussed at the Bioastronomy '99 conference
in Hawaii occuring just before the Extro4 conference.
This is (so far as I can tell at this point) an optimal
computing (i.e. "intelligence") architecture. A M-brain
has several interesting properties
- different layers are constructed out of different "computronium"
materials that have an optimal computational throughput for
temperatures based on their distance from the power source.
This makes "optimal" use of the elemental abundances within
a solar system.
- each layer of a M-brain lives off of the waste heat of the next
innermost (warmer) layer. This allows a more complete utilization
of the power generated by a star. Instead of the ~30% utilization
you might get from a single-layer Dyson shell of "computronium",
you may be able to fully utilize 80-90% of the solar energy.
- An M-brain has between 1,000 and 10,000 *trillion* (> 10^3 * 10^12)
times the thought capacity of the *entire* human population!
M-brains have the property, that they do most of their communication
"inward" (across shorter delay paths) while radiating waste heat
"outward" (to supply power for slower layers). There may be
an alternate architecture involving nested layers of little J-brains
which are themselves structured like M-brains, but this architecture
has the problem that the waste heat of a single node heats up
all of the surrounding nodes (to both the interior and exterior).
There comes a point when the computational advantage you gain
(due to decreased communication path delays in J-brain type
nodes) is offset in the decline in computational capacity that
results from the increased use of materials to "shield" or reflect
incoming heat from nearby J-Brains and the power consumed by increased
cooling requirements.
Paul Vitanyi, I believe, has shown, that even if you have
"hyper-dimensional" (but not faster-than-light) communication
paths, you cannot escape the communication-delay problem
in computation. So ultimately the optimal architecture
will come down to the size of the problem you are trying
to solve. While a J-Brain architecture may for a limited
set of problems provide the quickest solution, and works
great if you only have a few of them in a solar system,
an M-Brain architecture (with the greatest computational and
communication capacity) seems to be the optimal solution for
the largest problems and/or supporting the largest thought
machines.
Robert Bradbury
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:04:24 MST