[extropy-chat] Analyzing the simulation argument

[Hal Finney]

Dan Clemmensen writes:
> Assume a perfect simulation....
> 
> Here is the way I analyze such a proposition?
> 
> 1) logical and self-consistent? Yes.
> 2) consistent with observed phenomena? Yes.
> 3) useful explanatory power? Maybe.
> 4) falsifiable?  No.
>     Oops! our hypothesis is in trouble, but this is not absolutely 
>     fatal. We must still decide between
>     the assumption and the null hypothesis, so:
> 5) (Occam's razor) Is the system simpler with or without the
> assumption? 
> Without. That's it, ignore it henceforth unless new evidence arises.

The simulation argument is not an assumption.  It is an argument.
It is logic, not science.

The simulation argument can be expressed in the form, if A and B, then C.
The argument's validity depends only on whether it is true that A and
B together imply C.  An argument is valid if its logic holds.

The validity of this kind of argument does not depend on the truth of C.
The argument may be valid even if A or B were false.  Only if the argument
is valid, and A and B are both true, can we deduce that C is true.

In the case of the SA, A = "the human race is unlikely to go extinct
before becoming posthuman"; B = "any posthuman civilization is likely
to run a significant number of simulations of their evolutionary history
(or variations thereof)"; C = "we are likely to be living in a computer
simulation".

Occam's Razor is a tool for dealing with scientific hypotheses.  It has
nothing to do with logic.  A logical argument which uses many premises or
steps to reach its conclusion is just as valid as one which is trivial.
Fermat's Last Theorem is just as true as the Pythagorean theorem,
even though its proof is tremendously more complex.  Rejecting Wiles'
proof of FLT on the basis of Occam's Razor would be a mistake, applying
a tool to a domain where it doesn't fit.

Likewise, falsifiability is a technique appropriate for the domain
of science, not for logic.  We don't evaluate logical arguments or
mathematical proofs on the basis of falsifiability.  (Well, maybe some
of the debates about constructionism in mathematics might be thought of
as relating to falsifiability, but that's an esoteric detail not relevant
at this level of discussion.)

If we were dealing with a Simulation Hypothesis, which was simply
that we are living in a simulation, then you would be right to demand
falsifiability and use Occam's Razor.  But we are not.  We are dealing
with a Simulation Argument, and you should look at it as a piece of logic,
not a hypothesis about the nature of reality.

Now, I'm not claiming here that the SA is valid.  I'm just objecting
to this very common way of responding to it, as if it were a scientific
hypothesis.  If you want to reject the SA, reject its logic.  Or you can
accept the SA's logic but still not believe that we live in a simulation,
because you don't accept its premises.  This is the kind of analysis
appropriate to a logical argument.

Hal