X-Message-Number: 8600
Date: Tue, 16 Sep 1997 10:16:33 -0400
From: "John P. Pietrzak" <>
Subject: Re: Digital Shakespeare
References: <>

John K Clark wrote:
[ On the AIC as a measure of _algorithm_ complexity ]
> If you were attempting to show how your ideas of complexity differs
> from AIC I fear you have failed, at least I'll be damned if I can see
> any difference, just like you AIC is not concerned with the length of
> the results or how long it takes to calculate them, it's only
> interested in the length of the algorithm.

If I may, allow me to quote from an earlier message of yours (Cryonet
#8550): "[...] the Algorithmic Information Content (AIC) of a message
string is the length of the smallest computer program that can produce
the string, and that's just another way of asking how much the string
can be compressed."  Now, may I ask, what does string compressability
have to do with algorithm complexity?  Suppose I take two strings of
the same length, one of which is very uncompressible, the other being
very compressible, if the algorithm to print out the uncompressible
string consists of a single (gigantic) printf() statement, does that
mean that _algorithm_ is more complex than the compression algorithm
for the second string, which is smaller but contains more statements?

> I would also say that if the "complexity" of a problem has absolutely
> nothing to do with actually obtaining an answer to that problem then
> the word "complexity" should be retired because it has no use.

You are equating the cost of obtaining an answer to a problem with the
complexity of an algorithm solving that problem.  I am saying they are
different; related, perhaps, but not equal.  E.G., quicksort is
certainly a much more complex algorithm than bubble sort, but for
practically all cases works in much less time in solving the same
problem.

[ On the origins of Digital Shakespeare ]
> I think it's important to remember how this argument started, you
> said the basic unit of a computer, a switch, is very simple and I
> agreed, you then state that because of this everything about a
> computer and it's programs must be simple. I said and continue to
> say that is a completely nutty idea.

In a sense, that is what I believe.  Within the realm of algorithms,
some are certainly more complex than others; but speaking of computers
as a whole, the lowly Turing Machine is able to perform any algorithm
that any past, present, or future digital processor will ever be able
to do.  There is just so little to the TM, I often have trouble
understanding why people believe computers are so complex: it takes
very little effort to understand the workings of a TM, and after that,
the rest is just the costs of time and space, nothing else. 

[ On John K Clark's *Examples of Intelligence* ]
>          >I don't know why you chose them.
> 
> You don't understand why I group those thing together and give them
> the label "Intelligence"?

No, I think I understand why, but I have not been told by you, so it
is behooven on me not to assume.  (Besides, I'm pretty sure that my
reason for why you chose those examples is not the same as your reason
for why you chose them.  But, more on this below.)

> I think you do, I see no other reason you would say in the same post
> " I look at the world, just as John K Clark does, and find that there
> are certain things I associate with intelligence; novels, poems,
> theories, symphonies, rockets".

I belive you are taking that quote out of context: I go on from that to
state that after selecting the "what", I need then to find the "why"
that they were chosen.  In other words, I selected those examples
without consciously understanding why I did so.

>       >In fact, _you_ don't know why you chose them; if you did,
>       >you'd have some sort of definition of intelligence.
> 
> Utter and complete nonsense.

Nonsense in that you _do_ know why you chose them?  In which case, tell
me why (i.e., give me your definition of intelligence).  Nonsense in
that you _don't_ have, somewhere inside of you, a definition of
intelligence?  If so, tell me how it is that you didn't just choose
randomly.

> Most people on this planet have never even seen a dictionary and
> can't give a good definition of anything, yet their behavior is far
> from random [...]

Yes, that's _exactly_ what I'm trying to get at here.  There's more to
humans than just pure categorization by training of neural nets.  These
people come into the world with some structures *already in existence*.
They have a built-in bias towards the construction of their categories.

Simply stated, this means that there are rules, axioms, DEFINITIONS
of things like intelligence stuffed somewhere down inside the brain.
Just because people like you and me cannot determine exactly what they
are, the simple fact that we DON'T choose randomly means that there
must be some reason for _why_ we choose as we do!

> To say that all knowledge must be in the form of definitions is crazy.

This I don't understand.  Give me knowledge without definitions.

> You admit you have no definition of intelligence [...]

I admit that I have no conscious definition of intelligence (or, to
put it more accurately, I have too many inadequate definitions for
intelligence).  However, something somewhere inside me guides my
choices for what is and what is not intelligent; I do not choose
randomly.

[ On my "test for meaning" ]
> I've tried very hard but I can find nothing to disagree with in the
> above and in fact I think it's a very wise procedure, there is a
> technical term for it, it's called "The Turing Test".

It is certainly a similar test.  And, if you would note a bit further
down in my diatribe, you would find that I find MUCH to disagree with
in it, in fact it is the achilles' heel of my definition.  It is a
purely subjective and poorly understood process.  In so far as I am
able to replace it with an objective test, I will finally have my
true definition of intelligence.  Until then, I'm still just nibbling
at the edges of the problem.

(BTW, I'm off to Boston tomorrow on a sales trip for my company, so I'm
probably not going to be able to get to my e-mail for a while.  I'll
try to pick this up again later next week when I get back...)


John

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8600