From: Rafal Smigrodzki (rms2g@virginia.edu)
Date: Sat Dec 28 2002 - 17:28:03 MST
Ben wrote:
>
> My tentative contention is: There is no property that a "real" world
> would have, but a simulated world would not -- except the property of
> not being a simulated world.
>
### First off, if the simulation problem is still banned from this list, let
me say I am sorry, and any replies could be posted to WTA or ExI. But then,
on the other hand, I have a feeling that my thoughts on this subject might
have some bearing on AI, however remotely.
Some caveats: There are many implicit assumptions that go into even simple
deliberations, such as the feeling that the complex can be derived from
simpler objects, and the feeling, shared by many, that the simplest
mathematical objects are a form of ultimate, if limited, truth. It is very
difficult to think without such forms of reasoning, indeed, I would
immediately lapse into incoherence if I tried to shut them out of my mind.
Yet, on a very basic philosophical level, we cannot wholly reject the idea
that these forms of thought are not the only possible ones. We could live in
a very special simulation, designed to analyze the lives of constructs too
incoherent to exist in the "real world". In other words, even our concepts
of mathematics and logic could be in principle not generalizable beyond our
limited purview. This would of course invalidate any ideas we might have
about anything, including the simulation problem.
These caveats aside, one could try to derive some conclusions assuming that
our notions of mathematics are generalizable to whatever simulation level
(if any) we might be located in.
If indeed it is true that the complex is formed from the simple, then it
should be possible to delineate the minimum number of simple steps and
elements leading from a set of mathematical axioms and simple objects, to
the subjectively observed reality. There would be two classes of such
observable realities: the ones which could be mathematically explained
without recourse to any notions of "hidden variables" - complex phenomena
irreducible to mathematical transformations, and the ones where the
existence of such variables must be postulated.
To use a thought experiment: Let's suppose that a group of evil politicians
and their lackey scientists would take a number of newborns, and put them
into a VR environment consisting of a finite but very large Euclidean space,
populated by simple geometrical shapes. Certain interactions between to
shapes would result in food being delivered intravenously to the newborns,
and screens and electrodes implanted in the sensory and motor cortices would
allow the test subjects to interact with the environment. It is conceivable
that a society might develop among the minds who would perceive themselves
as simple geometric shapes (if allowed by the rules of the simulation).
Their scientists might develop a sophisticated understanding of mathematics
(the one shared with us) and local simulated physics (not shared with us).
However, any attempts at understanding the structure of their own minds and
bodies would be doomed without assuming the existence of highly complex
entities (brains, bodies, evil politicians) not directly derivable from the
simple mathematical objects and the observed version of reality accessible
to them. There would be a chasm between mathematics and physics. In effect,
they would have to postulate the existence of "supernatural" phenomena, an
irreducible cleft between body and mind, between structure and qualia.
If we, in our tinkering with the world, ever produce a simple set of
principles from which everything else, including the subjective experience,
naturally flows, then we would live in the first class of worlds - either
real, as real as natural numbers, or an ab initio simulation, which AISI are
one and the same. Should we ever stumble on a problem that cannot be reduced
to a concatenation of simple enough mathematical problems, this would be a
reason to suspect we live in a simulation. If you need to invent a god (a
conscious, goal-oriented entity) to explain the world, and even your best
efforts do not allow you get rid of this invention, there is a good chance
you are in a simulation.
Since the problem of qualia is to me the most vexing one in all of
philosophy, I am eagerly awaiting the ability to explore qualia in
Turing-capable machines. If there are such machines which would deny
experiencing qualia, while otherwise being capable of functioning and
self-monitoring at the same level and using similar information-processing
structures as humans do, I would seriously worry about the possibility of
the existence of souls, and other such rigmarole. Being the inveterate
skeptic I therefore expect that machines will have the same capability for
subjective experience as humans do, and it will be possible to tie this
experience to complex information-processing steps, built out of simpler
ones.
Rafal
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:41 MDT