From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Sat Dec 08 2001 - 20:58:12 MST
On Sat, 8 Dec 2001 Spudboy100@aol.com wrote:
> I wonder what kind of hardware would it take to generate an Earth
> simulation, with all its population, histories, biology and geology?
We went through this about 6 months ago, perhaps around the time that
Nick first proposed the argument. It got messy because people couldn't
agree on how much simplification could be invoked. I.e. The world
population, histories, biology, etc. are only simulated sufficiently
accurately enough that I'm "convinced" they are real. E.g. there is no
"real" (fully simulated) Eugene Leitl -- there is only a human fragment
AI that occasionially posts to several email lists I happen to subscribe to.
> What kind of hardware are we looking at? Similarly, what kind of software is being
> run, electrons, photons, photinos?
For a Matrioshka Brain, you can do at least 10^24 human mind equivalents
(a trillion trillion) based on rod-logic nanocomputers. You might be able
to push it to ~10^35 HMEs but beyond that will start to require subatomic
engineering or really robust quantum computers.
> Can a simulation that is run within a simulation effectively, be more
> complicated and capable, then the prior simulation?
I think it could be finer grained (i.e. we run molecular dynamics
simulations on non-molecular scale computers). So it might be more
complicated or capable for a very limited subset of what the external
simulation is capable of but it can't be more complicated or capable
of the entire external simulator.
Robert
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:12:24 MST