From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!garrot.DMI.USherb.CA!uxa.ecn.bgu.edu!mp.cs.niu.edu!rickert Tue Apr  7 23:22:17 EDT 1992
Article 4720 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!garrot.DMI.USherb.CA!uxa.ecn.bgu.edu!mp.cs.niu.edu!rickert
>From: rickert@mp.cs.niu.edu (Neil Rickert)
Newsgroups: comp.ai.philosophy
Subject: Re: Language as Technology: A Phenomenological Study
Message-ID: <1992Mar25.185007.21788@mp.cs.niu.edu>
Date: 25 Mar 92 18:50:07 GMT
References: <1992Mar25.080515.20086@a.cs.okstate.edu>
Organization: Northern Illinois University
Lines: 257


  WARNING: The following comments reflect my unconventional views.
	   I expect disagreement.  Please post your comments.

In article <1992Mar25.080515.20086@a.cs.okstate.edu> onstott@a.cs.okstate.edu (ONSTOTT CHARLES OR) writes:
>    When one reads Derek Bickerton's _Language and Species_, 
>one finds an unsurprising definition of intelligence: 
>  
>    1.) Language evolved.  
>...
>    5.) The book then tries to show that humans are more 
>    intelligent than other creatures by virtue of their 
>    language(in a top-down analysis). Thus, we have the first 
>...
>    It is noteworthy at this point to state that the AI 
>community generally accepts language as the determinant for 
>intelligence.  Further, understanding by a machine can be arrived
>at by means of giving a machine a language.(Even one capable of 
>dealing with something as esoteric and remedial as a blocks 
>world)  By 'language,' I mean not only the general languages of 
>English or French, but the logical language upon which the entire
>computer project relies. 

  Language is very important.  But it is far from the whole story.

  Chimpanzees are highly intelligent animals, although they do not have
anything we would consider language.  On the other hand, songbirds seem
substantially less intelligent, but they do possess some kind of language.
Machine languages have been around for a while, and if that were all that
were needed to make a machine intelligent we would not have these
interminable discussions on the CR and related topics.

  Here is what I believe is going on:

	The chimpanzee and the human possess a very powerful analog
	intelligence system.  It manifests itself in a remarkable ability
	to recognize, and for recognitions to generate various actions.
	This ability is often incorrectly dismissed as a mere unintelligent
	response to stimulus.

	The bird's analog intelligence is far less powerful, so in spite of
	its language abilities the bird does not match the chimp.

	The digital computer has language, but is completely lacking this
	underlying analog pattern recognition system, which is why computer
	programs sometimes seem to be so stupid.

  The most important aspect of human language is that it is a digital system.
In effect language has digitized humans.  Compare consumer electronics of the
'60s with comsumer electronics today.  The differences are enormous, and they
are almost entirely a consequence of the transformation from analog electronics
to digital electronics.  The difference between the chimpanzee and man is
even more enormous, and it too is the result of digital technology, in this
case the digital technology of human language.

  Roughly speaking, thought and consciousness consists of the ability to
simulate using our own intelligence system to perform that simulation.  Now
look at human technology.  Aircraft designers have long used wind tunnels
for simulation studies of new aircraft design.  These days they often use
computer modelling instead.  The effect of digital technology is that it
allows the easy creation of virtual realities, and thus simulation
no longer requires the presence of the physical hardware.  It is exactly the
same with human language.  As a digital technology it allows simulation
in the absence of physical apparatus.  Lion cubs can use their bodies as
a substitute physical apparatus to simulate predation, as they do in their
play, and this is not unlike children playing "cowboys and indians" games.
But, with the use of the digital technology of language we can simulate
things without the presence of physical apparatus, thus leading to
imagination, planning, etc, all of which are requirements of civilization.

 Songbird language appears to be an analog language, and therefore does not
offer the benefits of digital human language.

 Computers already are digital devices, and their machine code is digital
language enough.  Simulation, the equivalent of thought, is already available
to them.  What they lack is the underlying analog intelligence which would
allow them to recognize a wide variety of recurrent patterns, and to learn
to detect and recognize new patterns.

>    At this time, I would like to reiterate a study of the 
>phenomenology of understanding and intelligence which was given 
>at a lecture by Hubert Dreyfus at Oklahoma State University on 
>March 24th, 1992. The following comes from a paper he delivered 
>entitled, "From Socrates to Expert Computer Systems: The Limits 
>of Calculative Rationality." 
> 
>The outline goes like this: 
> 
>Utilizing a standard phenomenological study like that of 
>Heidegger we realize two things: 
> 
>1.) The notion of intelligence and understanding as 
>consisting of propositions and reductive methodologies 
>traces back to Socrates who consistently demanded the 
>"method" of how an expert comes to his conclusion.  The 

 This view of intelligence has predominated in most AI work.  The evident
failure thus far of AI to demonstrate human intelligence should be enough
to force us to reconsider this view.

>...
>2.) Expertness comes from the following method of learning: 
>...
>    A.) Novice - Needs precise rules and definitions to 
>    understand the new field.  Such as the fundamentals of 

  Not necessarily true.  It tends to be true in many of the fields we
consider most important, perhaps because our culture biases us toward this
view.  But precise rules and definitions are probably much less important to
an expert musician or an expert basketball player.

>    mathematics.  Or in the case of driving, at which point 
>    one uses the break peddle and at which point one uses 
                  ^^^^^^^^^^^^
	[Is this something you use to stop telemarketers? :-) ]

>    B-D.) Experiments with the rules to determine new 
>    outcomes.  Experiences an emotional component that 
>    helps curve learning.  Mainly, the learner experiences 
>    several cases.  As he advances from stage B to D he 
>    begins to loose track of the specific rules given in 1 
>    and begins to act more automatically.  Begins to 
>    develop intuitive skills. 

  These intuitive skills are the underlying analog recognition system at work.

>    E.) Expert - Has an intuitive skills in his field.  
>    Comes to conclusions without necessarily relying on 
>    first principles.  Can not, without a painstaking  
>    process which is never complete, provide a detail 
>    account of "how" he came to the conclusion that he did. 

  This underlines the importance of the underlying analog recognition system.
Quite often when the expert does "provide a detailed account" his explanation
is only an ex post facto rationalization of his conclusion.

>    If we remain perfectly honest with ourselves, we see that 
>this phenomenological study is accurate.  We do, in fact, begin 
>to get an intuitive feel for what we are doing.  Further, that 
>feel is not necessarily one that is thought out in a language or 
>in terms of first principles.  However, the expression of the 
>conclusion is produced in a language. 

  Unfortunately most people are unable to be sufficiently honest with
themselves to accept this view.  Digital technology if great for creating
virtual realities.  And humans, with the digital technology of language
are forever creating virtual realities, and being "brain washed" into
believing they are real.  We even allow ourselves to be brain washed into
the belief that we are resistant to brain washing.

>    Further, Dreyfus claims that this does NOT DISPROVE that 
>there is some ultrafast first-principles calculation going on.  
>However, it DOES shift the burden of proof TO the argumentation 
>that there ARE first-principle calculations going on.  In other 

  It does no such thing.  All it does is suggest that people are looking
in the wrong place for the wrong kind of first-principle calculation.
The underlying analog recognition system still requires some form of
calculation, but it is not the type of calculation that would be used
in formal deductions.

>...
>    If we take Dreyfus's argumentation above to be acceptable 
>and if we think purely in terms of Heidegger, it appears that 
>language itself becomes a technology.  From the technological 

  Right.  But the crucial point is that it is not just a technology,
but that it is digital technology.

  It is not entirely certain whether the development of digital language
(the digits are the phonemes) is certain.  It could well be that some
cultures could have developed with purely analog languages.  But such
cultures would be at such a great disadvantage they would be expected
to die out once they come into competition with cultures with a digital
language.

>point of view, language serves the function of transmitting ideas
>and symbolizing those ideas for alternative studies.(Which, no 
>doubt, leads one to think of intelligence as a symbol system). 

  Ideas can be transmitted with body language, physical demonstration,
etc.  You don't need language for that.  Digital language allows
modelling/simulation, and allows the modelling to be separated from
physical artifacts.  In effect this makes abstract ideas possible.  Digital
language also greatly enhances memory quality and memory density [compare
the density and quality of a conventional audio tape and a DAT].

>    Accordingly, we realize that language is not the genesis of 
>intelligence(thus, we can dismiss Bickerton's teleology towards a
>language).  Further, we realize that language is itself a product
>of intelligence.  Thus, leaving teleology aside and maintaining a

  Sorry, but I disagree.  Language is a product of the ability of the
analog recognition system to make sufficiently fine distinctions that
it can recognize individual phonemes as digits.  The effective use of
language is partly a product of the underlying analog intelligence.  But,
being digital, language provides a technology which is essentially
arbritarily extensible.

>random-competitive model of evolution, language came about not 
>because of some sort of direct evolutionary process;  rather, it 
>came about as a product of an evolution that produced 
>intelligence which, in turn, produced a language. (I am meerly

  Disagree again.  Language evolved to provide communication.  Communication
is important for enhancing intelligence, and has re-evolved in many forms
in many creatures.  Language is only one form of communication.  I view
use of hormones as another example of communication, this time between
organs in a single individual.  The evolution of a digital language is
largely a cultural evolution, made possible by a sufficient degree of
discrimination in the underlying mechanisms.  And language is important to
our society primarily because it is digital and thus supports extensive
modelling.  With the increasing use of video for communication, the importance
of language for communication is steadily diminishing.  I do worry that as
as a consequence peoples ability to use language for digital purposes (such
as model building) will also diminish, resulting in a "dumbing down" of
society.

>...
>    At this point, I have to differ from Dreyfus's 
>argumentation by claiming that expertness can be arrived at 
>without the means of a language.  I think that the 

  I agree, and I believe I said as much earlier in these comments.

>...
>Whenever we think about the development of language itself, 
>whether in young children or in primordial man, we realize that 
>some experiential component is a necessary precondition for the 
>development of the language itself.  Thus, language comes about 
>through experience without needing explicit logical 
>explanations(which would, in fact, be impossible without 

  There is a paper in Science, at approx Jan 31st this year, which
demonstrates that a great deal of language (essentially the ability to
recognize the phonemes) is learnt by an infant well before the infant
utters its first word, and well before the infant could have any knowledge
of the importance or significance of language.

>language).  Further, we become expert in language and THEN begin 
>to utilize this technology to generate other forms of technology.
>...
>present-at-hand.  In this way, the language-oriented AI project 
>is doomed to fail because it attempts to reach intelligence by 
>one of intelligence's tools--the communicative and symbolic tool. 
>
>Further, it is impossible to think of an AI project that is not 
>language based with the possible exception of neural networking. 

  It is not impossible at all.  However it does seem almost impossible
to persuade anybody to take one seriously if following this approach.

-- 
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
  Neil W. Rickert, Computer Science               <rickert@cs.niu.edu>
  Northern Illinois Univ.
  DeKalb, IL 60115                                   +1-815-753-6940


