X-Message-Number: 8517
Date: Mon, 01 Sep 1997 12:08:22 -0400
From: "John P. Pietrzak" <>
Subject: Re: To John Pietrzak
References: <>

Hi Thomas

> I have several points to make about John's comments.

Go right ahead. :)

> First of all, there are problems with the original Turing Test.

Oh Ghod, how I hate the Turing Test.  Don't get me wrong, Turing was
a genius in computer theory, but I can't believe how he managed to
foist that piece of crap upon the world of AI.  You are indeed right,
his test manages to entirely miss the point of intelligence.

On the other hand, the "Turing Machine" is a very convenient
theoretical model for describing limits of computability.  That's what
I was referring to in my previous note.

> As for the ability to pass a Turing Test, or to do calculations, we
> CAN do that. But when we look at the actual circuitry of our brains,
> we do not find similar circuitry in computers. When I said that we
> weren't organized the same way as computers, I was referring to the
> organization of our brains.  At a sufficiently abstract level it
> remains fair to say that we are organized the same way --- but then
> at a sufficiently abstract level we and computers are organized the
> same way as a seashell or a bit of granite.

Well, this is the question, isn't it.  At what level does the structure
of the system make a difference.  Various computers over time have had
wildly differing structures; we've moved from a loom which shoved pegs
through a paper tape, through various iterations of structures which
manipulated mechanical switches, to vacuum tubes, transistors, and
microchips.  All these structures were *identical* in theoretical
computing ability.  However, around WWII and for a short time after,
there was a class of "analog" computers which used properties of
electrical current and resistance to perform calculations which are,
in fact, not equivalent to those performed by digital machines.
Digital computers can approximate the results given by an analog
computer, but it's not exactly the same.

So, how does the brain work?  If it's a state machine, then it doesn't
really matter how it's wired together, you can put together a digital
computer which is identical to it (eventually).  If it's an analog
(or other) form of device, then you may or may not be able to achieve
an equivalent mechanical structure.

(On the other hand, people in AI may not need to worry about this.
It may be that a state machine is sufficient to create "intelligence",
whether or not it is like human intelligence.)


> As for the organization of human brains compared to the organization
> of computers, my own reading --- which has gone on now for years ---
> in neuroscience suggests to me that our brains consist of a
> collection of neural nets (of a kind that so far does not exist
> except in living things --- I do NOT exclude the possibility of
> making such devices) connected by a very simple sequential system
> which gives us our awareness. Neural nets aren't programmed, they
> are taught; they can also carry out recognition tasks that normal
> computer organization of the same number of chips would find
> impossible. A corollary is that our memories are stored in the same
> "chips" (otherwise known as neurons) which carry out the processing.
> They are stored distributed among many such chips, not just one or a
> few.

My studies have been mainly in the realm of computer science.  The
modern computer exists as it does today because it's relatively
easy to define the abilities and limits of a state machine.  Pretty
much any system which manipulates switches can be thought of as a
state machine, and people have been cranking out smaller and faster
switches every year.  But that doesn't mean it _has_ to work that
way, there are alternative ways to construct computing devices.  If
a system of switches doesn't work, we can move to a different system
that does, if necessary.

At any rate, yes, your criticism of the Turing Test is right on.

John

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8517