Re: Turing Machines and Physical Computation

From: David Longley (David_at_longley.demon.co.uk)
Date: 11/28/04 

Date: Sun, 28 Nov 2004 21:47:32 +0000

In article <U1qqd.577688$mD.266139@attbi_s02>, patty
<pattyNO@SPAMicyberspace.net> writes
>dan michaels wrote:
>
>> patty <pattyNO@SPAMicyberspace.net> wrote in message
>>news:<GtIpd.155286$R05.112941@attbi_s53>...
>>
>>>>ON THE COMPUTER "METAPHOR"
>>>>
>>>>'It has always bothered me that models of psychological
>>>>processing were thought to be inspired by our understanding of
>>>>the computer. The statement has always been false. Indeed, the
>>>>architecture of the modern digital computer - the so-called Von
>>>>Neumann architecture - was heavily influenced by people's (naive)
>>>>view of how the mind operated. Perhaps I had better document
>>>>this. Simply read the work on cybernetics and thought in the
>>>>1940's and 1950's prior to the development of the digital
>>>>computer. The group of workers included people from all
>>>>disciplines: See the Macy Conferences on Cybernetics, or "Her
>>>>Majesty's Conference on Thought processes". Read the preface to
>>>>Wiener's book on cybernetics. Everyone who was working together -
>>>>engineers, physicists, mathematicians, psychologists,
>>>>neuroscientists (not yet named) - consciously and deliberately
>>>>claimed to be modelling brain processes.'
>>>>
>>>>Reflections on Cognition and Parallel Distributed Processing
>>>>
>>>>D.A. Norman
>>>>(Ch 26, p534, Parallel Distributed Processing Volume 2)
>>>>McClelland J and Rumelhart D 1986
>>>
>>>I get that computer behavior and human behavior are essentially
>>>different. Why they are so different is certainly what we came here
>>>to discuss.
>>>
>> Hi patty, every couple of weeks I click on a post totally at
>>random to
>> find the same silliness repeated, with 0.999 confidence.
>> The problem with presenting a disembodied fragment is that it
>>possibly
>> implies what the fragmenter wants it to implie, but it hardly tells
>> what the original writer actually "meant". Follows are parts of the
>> **REST** of Norman's story .... [caps are mine] ...
>> .... here we are talking about a new form of COMPUTATION ... ...
>>these MODELS are highly parallel, with thousands of elements
>> INTERACTING primarily through actviation and inhibition ...
>> ... each ELEMENT is highly interconnected with perhaps tens of
>> thousands of connections ....
>> ...these NEUROLOGICALLY INSPIRED COMPUTATIONAL PROCESSES pose new
>> requirements on our understanding of computation, suggest novel
>> THEORETICAL EXPLANATIONS of PSYCHOLOGICAL PHENOMENA, and suggest
>> powerful new ARCHITECTURES for MACHINES OF THE FUTURE ....
>> .... carry on a tradition that has long existed .... a tradition of
>> BUILDING MODELS of NEUROLOGICAL PROCESSES ....
>> [and later on .....]
>> ... the whole point for the cognitive scientist is to UNDERSTAND
>> COGNITION ... to do so, we insist upon explanation of the INTERNAL
>> PROCESSING STRUCTURES THAT GIVE RISE TO COGNITIVE ACTIVITIES ....
>> .... this is why we have spoken of REPRESENTATION, of MECHANISMS of
>> memory and attention ...
>
>Point taken :) Making collages of fragments of other people's
>writings is art, not communication. What these collages refer to are
>mentalisms in the head of the author of the collage.
>
>patty

No, that is not the point you should take from this. That you do so just
makes you look as stupid as Michaels. What was it that Norman said about
the computer and the brain, and what has been shown about the
"properties" of so called ANNs apropos human and other animal behaviour?
What is it that they model do you think? (You haven't grasped how this
relates to natural stupidity)

The reason why you. Michaels etc don't pick up on any of this properly
is because you aren't trained in psychology. You don't have the context
or web if you like, to be able to make sense of it properly. The context
you have is that of an enthusiastic amateur with skills in other areas.
Those skills just don't equip you to read or talk in other fields and
that's the problem I have been illustrating. If you look to what many
people in "AI" or "Cognitive Science" get up to, you'll see this
repeated my many people. They commit the genetic fallacy. That is, they
may have skills in mathematics or programming, and they just assume they
have the ability or expertise in fields where in fact they have little
or none - in fact they have nothing more than pre-scientific "folk
psychology". Why would Michaels or you presume otherwise one might ask
(one can ask this same question of many of his luminaries too).

Read the section of "Fragments" on "transfer of training".

All you're actually doing here is reinforcing both your own and
Michaels' stupidity. The quote from Norman was simply to show what
people were doing back in the 40s and 50s. That they were doing that
doesn't mean that what they were doing was sound any more than Michaels'
or Norman's talk of ANNs being neurally inspired is sound! What these
models show is how we can model assumption violation, overfitting etc -
problems well known to those of us who use inferential statistics and
build models - hence construction, validation samples, reliability
measures, shrinkage etc. There is a subtle point being made about what
is being modelled which you are all missing.

See the two *other* sections of "Fragments" on a) ANNs *and* b)
hypothesis testing.

What is *your* behaviour illustrating? 

-- 
David Longley
http://www.longley.demon.co.uk/Frag.htm