X-Message-Number: 8550
Date: Sat, 6 Sep 1997 20:57:12 -0700 (PDT)
From: John K Clark <>
Subject: Digital Shakespeare

-----BEGIN PGP SIGNED MESSAGE-----

On Fri, 05 Sep 1997 "John P. Pietrzak" <> Wrote:

        >>Me:
        >> Pi = 4 - 4/3 +4/5 -4/7 +4/9 - 4/11 + ...        
        

        >Calculating Pi (using the above formula) is, in fact, quite simple.
        
        >The following algorithm (in pseudocode) is sufficient to do it.
        
        >[...] Unfortunately, because it contains an infinite loop, this
                >algorithm will not terminate. 
             

The task I gave you will certainly terminate and it doesn't have an infinite 
loop, I didn't ask for the last digit of PI, I asked for the trillionth and 
that's very finite. You've repeated that it's all quite simple so I will 
repeat, what is it?
             


        >Still, the real problem here is that you and I have a different
                >definition of "simple".  
             

I have no definition of "simplicity" or "complexity" just as I have none for 
"intelligence" and for the same reason, these things are best explained by 
example. That's not to say definitions haven't been attempted, you're pushing 
the AIC measure of complexity, 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. The AIC is largest for random strings because the smallest 
program that can produce a truly random string is the string itself, it's
completely incompressible. Accordingly, the gibberish a monkey produces by 
banging on a typewriter contains far more information than a Shakespeare play 
of the same length. This is almost the exact opposite of what we usually mean 
when we say something is complex. Another rather serious problem  with AIC is 
that we can never find it, we can only find an upper bound. Greg Chaitin 
proved a few years ago that in general you can't be certain that there is not 
a program shorter than any you know about that will generate the string. 
This also means that although most strings are random and thus incompressible, 
you can't prove that any particular string is random. 
             
There have been other attempts, like a measure of how long a program must run 
to produce its result, not how big the program is. Still others think the 
amount of memory a machine would need to produce the output from a program is 
the best definition. Unfortunately the monkey still beats Shakespeare.
          
         >>Me:

         >>In the end all Shakespeare did was put ASCII characters in a
                  >>sequence.


        >Absolutely not!  There is a great deal of difference between the
        
        >sequences of characters originated from  Shakespeare and a random
                >sequence of characters.
         

Interesting, it would have been even more interesting if you've given me some 
hint as to what those differences are.
           

        >You're confusing instances of behaviour with ability to behave.
        
Huh?


       >Certainly, a computer may be able to present to you something that,
              >with respect to your previous knowledge of what it has done, is novel.
        
Absolutely.
        

        >But what you can't do, is create a new (digital) computer which can
                >do something that a previous computer couldn't possibly do.


You may not be able to build such a machine, but you might be able to build a 
machine that could build a machine that could.



        >The concept of the Turing machine is powerful because it does place
                >absolute limits on computability.


Yes, and I see no reason to think humans don't exist under the exact same  
limitations. 
          


        >In my opinion, the way that you act is not a good indicator of your
                >intelligence.  


Then who in hell needs intelligence, it's a useless concept! I don't care if  
I'm dumb as dirt as long as I act in a smart way.

I've asked this question before but received no answer, since behavior can  
tell you nothing about intelligence why do you think Einstein was intelligent?
                        


       >Your statement here is more of the form "X acts_like A implies X = A".
       

OK, so tell me, if something is moving swiftly is it swift?.
                        


        >If you can define "talks intelligently" in an objective form, you
        
        >would have a valid (and valuable!) implication to use in determining
                >intelligence. Unfortunately, there really isn't any objective        
        >definition for it...


And yet you use the word "intelligently" every day, I don't think it's just a  
sound you make with your mouth like a burp, I have a hunch you mean something  
by it, even if you can't define it. 
                          

        >Eliza has passed the Turing Test a few times.          


Yes, Eliza can pass the Turing Test if the tester is an utter moron, he would 
conclude that the program is as smart as he is, and in this case he's 
probably correct. Eliza wouldn't fool me, I guarantee it.
                          


        >What you yourself have is a way of distinguising Einstein or Hawking
                >from a rock or a tree or a bug or a politician.  

Yes.


        >What you don't have is any reason for giving this category the title
                >"intelligence".


I do have a reason, the most powerful reason a neural net can have, I put 
things in a certain category if they're like Einstein and unlike a bug.  
If you dislike my label "intelligent" for this group of objects then pick  
another word.


        >Quite a variety of what I term "analog" computers have been 
        >constructed already; machines which perform mathematics by adding or 
        >subtracting actual amounts of electrical current rather than by 
        manipulating switches.


On a good day for brief periods you might get close to 10-bit performance on 
one of those junk heaps. Infinite number of internal states my eye.
               


        >no matter what type of new, fancy processor you come up with, if it
        
        >is Turing Machine equivalent, then it can do anything a Turing
                >Machine can do, and it can't do anything a Turing Machine can't do.  


All true, but show me something you can do that a Turing Machine can't.
                


        >This is the most damning aspect of the test: that the results are
                >purely subjective in nature.          


Just like the way we judge the mental abilities of other people.
                
In #8545  Andre Robatino <> On Fri, 5 Sep 97 Wrote:
               

         >it's not just a question of whether the architecture is truly
         
         >optimal, but whether evolution is capable of figuring it out. A lot
         
         >of biology seems to be set early on in evolution and then conserved
                  >in higher organisms.  


Very true. The trouble with evolution is that it has no foresight, every 
change must be an immediate improvement. If people acted like that they 
couldn't even change a tire when it got worn because when you took the old 
tire off you've temporally made things worse, now you have no tire at all.
               


        >Seth Lloyd's quantum simulators do indeed have an infinite number of
        
        >possible internal states.  Don't let the fact that the input and
                >output are discrete fool you. 
               

If "different" internal states  produces the same output for the same input 
then exactly what's different about those states? 
               


        >Thus even though the input and output are discrete (when measured),
        
        >the output has a nonzero probability of being wrong if one attempts
        
        >to use it for any form of digital computation.  So instead of the
        
        >uncertainty being manifested by a small error in the output, it shows
                >up as a nonzero probability of getting the wrong discrete answer. 
               

The output is either correct or it's not, there are no other possibilities. 
Are you saying the output is correct if the machine was in one internal state 
but the exact same answer is wrong if the machine is in another internal 
state???       



        >Though an electron's measured spin has just two possible values,
        
        >+1/2 and -1/2, its spin state is described by a superposition of
        
        >pure spin states of the form a|+1/2> + b|-1/2> where a and b are
                >complex numbers. Thus it lies in a continuum. 
               

That's true, by measurement I can tell if an electron is spin up or spin down, 
but that still doesn't completely define the axis, it can be anywhere on the 
surface of a cone. I will go further, the quantum wave function itself is 
continuos, but that does us no good because experimentally we can't measure 
the quantum wave function F(x) of a particle, we can only measure the 
intensity of the wave function [F(x)]^2 because that's a probability and  
probability we can measure. 

You complain that I'm only talking about observable qualities, but we're   
talking about a machine here, a computer, and that's the only thing it can  
deal with. It would be as if I said I had a new type of airplane that flew by  
pulling itself along lines of longitude and latitude.
                                                        
                                              John K Clark    

-----BEGIN PGP SIGNATURE-----
Version: 2.6.i

iQCzAgUBNBIiXX03wfSpid95AQF7YgTvc/SOrza09YzG7tkS/omUhZM0GFQ+ewjr
A9hHecBlj+ARYadvIRuMsqDOY+EuCf0D/hgXdOOPcOpi4CDh5kYwtf4eBnp/QgrU
wXwqE31TxQdlgbFhptPJlL2H14c0tbQLeaNkzHmeEPKxhPft+TM55CuNkyF6qIyE
BsftIMMmLzVK1x400g7MqQ5k5/2HpHVjtiHdCenInbMu8VqDc9E=YIUj
-----END PGP SIGNATURE-----

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8550