Re: A question for all!

DON MIKULECKY (MIKULECKY%VCUVAX.BITNET@letterbox.rl.ac.uk)
Tue, 10 Oct 1995 10:27:14 -0400


Don Mikulecky, MCV/VCU, Mikulecky@ gems.vcu.edu

Reply to:

> Catharina Kennedy, ck@ics.inf.tu-dresden.de
>

Wow! After all this time....someone who seems to understand!

> I wish to reply to Don Mickulecky:
>> One big question still looms overhead: how much of this kind of "complex
>> ity"
>> can really be caught by computer simulation, if any?
>
> I cannot answer this question at present but I hope I can contribute some
> ideas for further discussion.
> A mechanistic system, as I understand it, operates under a sequential causalit
y.
> (i.e. A causes B which causes C etc). The notion of algorthim and Turing-
> computable is also inherently sequential (even with massively parallel models
> one needs a sequential algorithm to do something "useful"). This sequential/
> mechanistic paradigm is much easier for a human observer to visualise and it
> is therefore difficult to give up this approach.
>
> In contrast, an important idea being developed by the ICS group in Germany
> is that of *simultaneity*. Living or cognitive systems could be modelled
> as networks where one event can express itself *simultaneously* in different
> forms. The precise form is context-dependent (or standpoint-dependent). An
> example is a lecture being listened to by several people. Within the brain
> of each listener the utterances express themselves differently, since each
> listener has constructed his/her own individual context.

Here's exactly where I am confused. We can simulate an parallel system
(artificial neural network) on the serial processor. Does this mean
that a hard wired parallel processor has more to it than the simulation?
Is "simultaneity" more than parallel? This could be very important!

>
>> These ideas intertwine with an idea of Rosen's that we measure things
>> convenient to us inspite of the possibility that the natural system is not
>> using those observables in any meaningful way at all!
>
> I would understand this as follows: if we choose to make measurements
> according to a mechanistic conception, there will always be some non-
> deterministic effects that don't fit in. (It doesn't matter how "good"
> the conception is - in the sense that it "works" as an explanation a large
> part of the time - it will always be incomplete).

Yes, that's the idea exactly!

>
> The unexpected effects could be modelled as follows: they are the forms of
> expression (in our current context/measuring system) of sequences of events
> taking place simultaneously in contexts which are "inaccessible" to us (after
> all we can only pay attention to one context at a time and that is the current
> mechanistic model we are using). The other "contexts" are alternative position
s
> of description we could take but which could not be mechanistically "added"
> to the current model without contradiction. Effectively this is a form of
> dialectic.

Rosen calls it a "dialectic" in his book "Fundamentals of Measurement".
Also he discusses a notion of "error" which is introduced in this way
and relates it to our subjective notion of "complexity".

>
> I will try to give an example: somebody says something I don't understand,
> i.e. in my internal context, the expression has taken on an unintelligible for
m
> but in the other person's internal context it makes perfect sense. Instead of
> mixing these two viewpoints together into a single mechanistic system (which
> would lead to contradictions) they should be considered as seperate systems
> with a simultaneous non-mechanistic relationship between them.
>

That is exactly the subject matter of relational biology. Rosen calls
these things "linkage".

> In other words, a living or cognitive system could be approximated as a networ
k
> of event sequences where each event may or may not have *simultaneous
> translations* within the operation of neighbouring event sequences.
> It is precisely these "simultaneous translations" which cannot be reduced to a
> mechanistic system, and they are necessary if a system is to be "alive".
>

This is very suggestive....can you tell us more about this idea?

> As a possible step towards the future formalisation of such networks, an
> idea for the extension of classical logic has been proposed by German
> philosophers. I have written a summary of my understanding of their ideas
> in a seperate paper: "Logical Problems in Cognitive Modelling". If anybody
> is interested, Cliff has a copy on
>
> ftp://kong.gsfc.nasa.gov/pub/joslyn/Kennedy,_Catharina_paper

I will try to find it. I am clumsy on this net though.

>
> To return to the question above: I do not know how far we can approximate
> certain aspects of such simultaneous networks on a v.Neumann computer but
> it is just possible that a formalisation will be found and it will be a
> matter of developing a new parallel computer architecture which provides
> the operations to implement such simultaneous translations.

This is the question.

>
> Trying to define "simultaneous" will definitely require another posting.
> I hope I have made myself clear. My understanding of this issue is still
> developing.

May we join you?

>
> The ICS (Institute for Cybernetics and Systems Theory) in Bochum, Germany
> should have its home page early next year.
>
> I have only read a little of Rosen's work. Could anybody give a list of
> his publications?
>
I will send a bibliography I have put together to you e-mail address.

> regards,
> Catharina
>
> ------------------------------------------------------------------------------
> Catharina Kennedy, ck@ics.inf.tu-dresden.de,
> PhD student, Institute for Artificial Intelligence,
> Faculty of Computer Science,
> Technical University of Dresden,
> 01062 Dresden, Germany.
> Tel: +49 351 4575 490
> Fax: +49 351 4575 335
> ------------------------------------------------------------------------------
Thanks. You've been very helpful!
Best regards,
Don Mikulecky
P.S. Jeff Prideaux's tumor cell population model is highly parallel,
but simulated on the computer. A later version uses cellular
automata.