Re: George Kampis

[Tim Gwinn <***>]

> Judith
> PS: Tim; how would you say George's work/conclusions differ from my
> father's?

I hope I don't mangle this too badly. Roughly and very briefly, for Kampis
the argument is as follows. Systems can be thought of as being comprised of
"components". Components are not "things" per se - context (their role in a
system and how they arose) can play a role. Reductionism might be thought of
as the program of finding what are the basic fixed context-free components
which together are presumed to comprise some sort of "universal library". A
"component system's" behavior results from the particular components which
constitute that system. The dynamics of the system, then, can be mapped to
some algorithm. But this holds true only if the components in the component
system remain fixed and constant. If they change, then the dynamics change
and an entirely different algorithm will generally be required. Now that
system is modeled by two (or more) different algorithms, neither one of
which entails the other logically nor can the algorithms necessarily become
combined into one large algorithm; hence, that system is computable locally
and temporarily, but overall as a system is noncomputable..

The stipulation that "what is a system" is restricted to be those systems
which are computable mandates that the components of the system must remain
fixed and constant. This amounts to an arbitrary a priori restriction on
system identity and on physical reality itself. Self-modifying systems are
ones for which the system itself can alter its own components and thereby
alter the nature and behavior of the system. Self-modifying computer
programs are really only systems which are restricted to some recombination
of some universal set of components - a computer can only have a finite
instruction set and has limited inferential entailments. These restrictions
are by design and entirely artefactual: we have to specifically and
carefully design a computer so that it will in fact work this way and only
this way. By contrast, there are no such artefactual restrictions on the
physical world and the systems therein. Such systems can therefore
potentially self-modify in ways that outstrip our algorithmic notions of
entailment, and therefore such systems clearly can fall into the category of
being noncomputable. We can build computers and toasters and so on as
physical systems, by imposing appropriately designed constraints, but that
does not in any way imply that physical systems are therefore generically so
restricted in their flexibility.

This gives the main theme in very broad strokes and it may come across as
overly simplistic. Kampis addresses alot of subtle and interesting points as
he goes along. Clearly there's alot of kinship with the ideas in Rosennean
complexity, although it is framed and argued in very different terms. Kampis
goes into much more detail, and the discussion of "The Main Theorem" occurs
only after 200+ pages of preparatory discussion of the Newtonian paradigm,
the notion of state, dynamics, and so on. Alot of similarity to the
preparatory discussion in "Life Itself". Then follows another 200+ pages,
including discussion of Church-Turing, information, syntactics/semantics,
etc. I always found it interesting that both books were both published in
1991, although I consider this to be purely coincidence.

Regards,
Tim