Newsgroups: alt.fan.cecil-adams,alt.usage.english,sci.lang
Path: cantaloupe.srv.cs.cmu.edu!rochester!hood.cc.rochester.edu!news.acsu.buffalo.edu!dsinc!spool.mu.edu!uwm.edu!www.nntp.primenet.com!nntp.primenet.com!howland.erols.net!netcom.com!jqb
From: jqb@netcom.com (Jim Balter)
Subject: Re: WARNING  Popperesque Paradigm shift approaches
Message-ID: <jqbE2BD8G.7L9@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
References: <misraelE27EIG.G1v@netcom.com> <32ADF69E.25E1@sn.no> <58mvu7$77p@riscsm.scripps.edu> <58n43m$692@lastactionhero.rs.itd.umich.edu>
Date: Thu, 12 Dec 1996 18:41:04 GMT
Lines: 99
Sender: jqb@netcom.netcom.com

In article <58n43m$692@lastactionhero.rs.itd.umich.edu>,
John M. Lawler <jlawler@choplifter.rs.itd.umich.edu> wrote:
>Mark Israel <misrael@scripps.edu> wrote:
>>jlawler@umich.edu (John Lawler) writes:
>>   You're asking for rigour in the expression of a hunch; but for the
>>sake of argument, let's define "simpler" as "requiring fewer bytes to
>>store after naive data compression".

Over in comp.ai.philosophy, some of us have spent quite a bit of time
thrashing out a notion of rational behavior as being largely a matter of
parsimonious fit, which can be expressed formally in terms of Algorithmic
(Kolgomoroff) Complexity theory, which gives measures of compressibility.

>Fair enough.  That's all we can do for "random", after all.

I'm not quite sure what this means; formally, "random" implies "uncompressible".
Refer to Gregory Chaitin's work.

>>>> Linguists remind me of geocentric astronomers adding more and 
>>>> more epicycles to save the phenomena when it would be more 
>>>> profitable to question their basic assumptions.

No doubt this happens in every field, but unfortunately human behavior is much
messier and ad hoc than physics, which is why the "soft" sciences are so much
harder to make headway in than the "hard" sciences.  And which is why PET
scans that show activation patterns are so much more direct evidence of
hemispheric specialization than are psychological observations of
brain-damaged persons, to borrow a point from another thread.  Actual
observation of human neural organization via such methods has totally blown
away past wishful theories of the brain being some sort of amorphous malleable
interchangeable stuff; there is some fungibility, but there is also lots of
specialized structure and function.  So while some simple principles
(parsimony/compressibility of rules) may apply at some very gross level of
description, the details are bound to involve lots of epicycles.

>>   As for ways to question them, one of my chief problems with
>>linguistics is its lack of predictive power.  I think the solution
>>lies with artificial intelligence -- if we hypothesize that it is
>>possible to process language in a certain way, the fact that a 
>>computer program does do so is pretty good confirmation.

My experience programming language understanding systems led me to think that
this is pretty hopeless.  Just in order to disambiguate "a" from "the" in
continuous speech, we needed an entire semantic net, and that was with a
constrained 1000 word language.  The main thing to draw from AI is just how
damn complex "I" is.  There are many of us who think that the key to AI is
learning; the only way to develop language processing skills may be the same
environmental richness that humans are exposed to (despite my personal and
intellectual regard for Noam Chomsky, I think his "poverty of stimulus"
arguments are poppycock).

>Well, prediction is only possible in severely constrained contexts
>generally.  Electron behavior can be predicted statistically because there
>is in principle no difference between one electron and any other.  If we
>posited personality for subatomic particles as well as charm and
>strangeness, we might find the situation a little more complex. 

And if those electrons were made up of billions of electroneurons with
trillions of pathways among them ....


>>shughes@sn.no ("Simon R. Hughes") writes:
>
>>> Would an alternative interpretation of Mark Israel's prediction be this:
> 
>>> **************************************************
>>> * WARNING  Popperesque Paradigm shift approaches *
>>> **************************************************
>>> ?
>
>>   Something like that.  But I'm afraid I'm not up on my Popper; 
>>someone would have to enlighten me how one of Karl Popper's paradigm
>>shifts differs from one of Thomas Kuhn's.
>
>I don't find a great deal of difference, personally.

Neither did Popper, but Kuhn thought otherwise.

>But then I
>think Feyerabend's position is superior to either of those.

By what method would you support that claim?

I think the philosopher Joseph Agassi's treatment of Feyerabend, as in his
"Feyerabend's Defense of Voodoo: how to get away with murder" gives Feyerabend
the treatment due him.  A.F. Chalmers in his outstanding philosophy of science
overview, _What is This Thing Called Science?_, treats Feyerabend more politely
but sharply nonetheless.

>  "... a large subset of language can be handled with relatively simple
>   computational tools; a much smaller subset requires a radically more
>   expensive approach; and an even smaller subset something more expensive
>   still."

Isn't there a universal law of engineering that the first 90% of anything
takes only 10% of the work?
-- 
<J Q B>

