X-Message-Number: 8191
Date: Wed, 7 May 1997 09:12:03 -0700
From: "Joseph J. Strout" <>
Subject: Re: subsimulations

I haven't taken part in this discussion for quite a while; I try not to
jump in unless I have something to contribute (and time to contribute it!).
Hopefully this message meets that standard...

In Message #8185, Robert writes:

>c) If people in the real world are motivated to simulate a world full of
>people, the simulated people will also be so motivated, generating an
>exploding set of subsimulations, effectively causing the original computer,
>in the real world, to stop.

I don't see how this could be.  A simulation of a thing always runs slower
than the "real" thing, unless the simulation is significantly simpler than
the device running it.  So my 180 MHz Macintosh can emulate a Commodore 64
in real time; it can also emulate a Pentium running Windows -- but
substantially slower than a real Pentium.  Within SoftWindows, I may
perhaps be able to install a Mac emulator, and it will run slower than a
Mac emulator running on a real Pentium, and MUCH slower than the real Mac.

But suppose my Mac were fully multi-threaded (or instead of a real Mac,
make the outermost Mac itself a simulation running on a Unix workstation).
Another thread is simultaneously running which is, oh, finding prime
numbers; and this other process has at least as high a priority as the
simulation process.  As we run simulations within simulations within
simulations, does my prime-number search slow down?  Not a bit.  The
simulations just get slower and slower, *with respect to my time scale* as
an outside observer.  Or, equivalently, with respect to the time scale of
the outermost device.

But with respect to the time scale of the next-to-last device (in this
example, of SoftWindows), the simulation it's running hasn't slowed down a
bit either.  If it updated one simulation tick per 20 ticks of its own
clock, then it's still doing so; it can have no "awareness" that its own
clock is slowed down relative to the one above it.

If we leave the realm of practicality and talk about planet-sized computers
simulating billions of people, the same principles will hold.  Maybe we're
a simulation within a simulation within a simulation.  Now we build a
computer the size of Mars and start a simulation inside it.  It will run
along just swimmingly as far as we can tell.  If the beings inside do the
same thing, THEIR simulation will look slow to us, but just fine to them.
And never mind that one second of our own time takes a million years for
the ultimate creator, no matter what we're doing.

So no matter how far you nest, things won't grind to a halt.  As we peer
deeper into layer after layer of simultation, it looks slower to us, but
always fairly zippy to the simulators just above.  And no amount of this
will cause any slowing of the top-level simulation, which after all is just
simulating atom movements and basic forces, and couldn't care less whether
those atoms are arranged into clouds or computers.

Regards,
-- Joe

,------------------------------------------------------------------.
|    Joseph J. Strout           Department of Neuroscience, UCSD   |
|               http://www-acs.ucsd.edu/~jstrout/  |
`------------------------------------------------------------------'

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8191