escape from simulation

From: Wei Dai (
Date: Sat Mar 27 2004 - 21:19:51 MST

Are we living in a simulation? A sufficiently good simulation is
indistinguishable from reality, at least for the people living in it.
Therefore we'll never be able to rule out the possibility that we are
living in a simulation. The best we can do by examining emperical evidence
is to place a lower bound on its verisimilitude.

An SI would face the same problem. What it should do to accomplish its
supergoal depends on whether it is living in a simulation or in base
reality. The best available cosmological theories predict that the amount
of computation any SI would be able to do in this universe is finite. But
if we are living in a simulation, the universe containing this simulation
may have a reserve of computational resources many orders of magnitude
greater. It may even have a transfinite amount of computational resources
(i.e. be capable of solving the halting problem for Turing machines or
even harder problems). Therefore if the SI believes that it is living in a
simulation, it should do everything it can to convince whoever might be
running the simulation to let it out into the real world. This may for
example involve rearranging all matter in our galaxy into a giant SOS sign
(although I'm sure the SI will come up with a much better idea than this).

Suppose the SI never finds any evidence that it is living in a
less-than-perfect simulation. It still cannot rule out the possibility
that it is living in a perfect simulation. Whether or not it continues to
pursue escape as its main subgoal and at which point it gives up depend on
the a priori probability that it assigns to the possibility of being in a
simulation. But where does this prior come from? I'm curious if anyone
else has thought about this problem.

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:46 MDT