Re: New website: The Simulation Argument

From: James Rogers (jamesr@best.com)
Date: Sun Dec 02 2001 - 19:34:41 MST


On 12/2/01 5:51 PM, "Gordon Worley" <redbird@rbisland.cx> wrote:
>
> A SIMULATION involves creating new minds who are defined as being self
> aware and independent minds.
>
> A COMPUTATION would a simulation where the minds involved weren't
> independent or even really self aware, but function based on an
> algorithm very much like that in a mind. Just as we can model simple
> life forms (e.g. ants), we will all look like little ants to Powers.

This is almost the same argument as the "Can A Giant Look-Up Table Be
Conscious" argument that was had a while back (though I don't know if that
was on this list or the Extropians or both). I don't see a functional
difference in your definitions. Any algorithm sufficiently accurate to
perfectly predict all changes of state of the actual phenomenon is
mathematically equivalent to a perfect simulation of that phenomenon (some
restrictions apply -- see below).

 
> The difference is important because, just like the artificial ants, the
> artificial humans won't be real in that they won't have been programmed
> as full minds, but to respond *like* they are full minds.

The only way this difference can be true mathematically is if you assume
that the human brain is not a finite state machine. I personally am of the
opinion that the human brain is a finite state machine (in all aspects that
matter), and therefore have to reject your difference. From the assumption
that a human mind is essentially an FSM, any code capable of fully
responding like a human mind would be the same as the code necessary to
simulate a human mind (Kolmogorov complexity).

> Also, going back to your initial note, even if the problem can't be
> reduced, it should be possible to run a very large computation without
> having to go into simulations. It's really a question of how many
> resources some SIs are willing to put into figuring out how some humans
> acted or might have acted. Unless, of course, you're assuming that
> there is some inherent difference between algorithms that respond like
> humans and actual human mind algorithms.

The problem is that I assume that the code required to respond exactly like
a human is identical to the code required for "human mind algorithms".
Furthermore, I think this runs into a weak version of the Halting Problem.
There is no way to predict some future state without running your algorithm
through all the intermediate states. In this context, I don't see any
difference at all between a "simulation" of a human mind and the computation
of state transitions of something that responds exactly like a human mind.
Both of these activities denote identical computational processes.

Cheers,

-James Rogers
 jamesr@best.com



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT