From: Stathis Papaioannou (stathisp@gmail.com)
Date: Mon Mar 03 2008 - 05:41:04 MST
On 03/03/2008, Krekoski Ross <rosskrekoski@gmail.com> wrote:
> yes, but thats not the point. You can have a random number generator
> (assuming such a thing exists) that outputs an arbitrarily long string of
> digits every n seconds. The number of molecular configurations in the human
> brain, or even the number of potential n second long simulations of the
> human brain is a vast but finite number, therefore eventually our random
> number generator will output an accurate n-second long simulation of our
> brain, but the number generator is not intelligent.
Certainly, the computation hidden in noise will not interact
intelligently (or even stupidly) with its environment. It lacks the
essential utility of a computation, which is to accept an input and
produce an output that can be then used for some other purpose. If the
thermal motion of atoms in a gas could be seen under the right
interpretation as an implementation of the calculation of pi, so what?
It doesn't do anyone any good.
But consider a special class of computations: inputless,
self-contained virtual environments with conscious inhabitants. If
these arise in noise, they're not going to pass any Turing test
because by definition they cannot interact meaningfully with the
environment at the level of their implementation. However, that we
can't talk to them should not make any difference to the inhabitants
in the computation themselves, who are intelligent with respect to
their own environment. Such forever hidden and inputless computations
must be occurring everywhere.
I see this as a consequence of functionalism, and unlike Lanier and
some others I don't see it as a problem for functionalism.
-- Stathis Papaioannou
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:02 MDT