Re: Why AI does not require Reductionism

From: Christian Szegedy (szegedy@or.uni-bonn.de)
Date: Mon Jan 21 2002 - 05:03:29 MST


Paul Hughes wrote:

>Lets assume that he is right. Lets assume that
>the source of consciousness resides somewhere outside
>of all currently known physical laws. This still would
>not make a difference. In fact it would only
>strengthen the possibility of creating an AI. How?
>
Basically, you are right, but there is a more important point, you are
missing.
What Penrose is sceptic about is the Church thesis that states that the
every discrete aspect of any physical process can be modelled by a Turing
machine. This thesis is of course unprovable, but there is no counterexample
so far. Even the quantum computation models stay within the realm of
Turing-computability, "only" the efficiency of some algorithm could be
increased theoretically. (This is also experimentally unproved, and some
scientists doubt the possibility of practical implementations of this.)
Only if we knew all physical laws, this would be a well posed
mathematical question, but probably a hard one. I don't know whether the
Church thesis is proved even for the Newtonian mechanics.
Of course, if there existed a grand unified theory, there would be still
a chance that
the model is not complete, so some uncertainity will always remain.

The main question is therefore: is the human consciousness modellable by
a Turing machine (a digital computer). This is one of the most interesting
question posed in science so far, even if it is far from well-defined
or decidable.

Best regards, Christian



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