Re: Howard's challenge

[John Kineman <***>]

Again, as pointed out in a previous post, but not recognized in this
discussion (so I'll repeat it):


> ... Von Neumann clearly did not equate construction (fabrication) with computation in any of his writing or his lectures. It is quite unreasonable to believe that he would fail to make this distinction since he virtually defined this distinction (program software vs. machine hardware) as it is now understood.

Software is fully specified and computable and exists in a 1:1
(non-complex) relationship with encoding structures, unless you consider
the whole software writting exercise as a social phenomenon, i.e.,
including the programmer in the system being modeled. Did von Neumann do
that?? It is the same argument as that for Goedellian incompleteness,
that completeness can only be achieved by including the mathematician.

> It was used only as the creative stimulus for the concept of a “universal” constructor” that von Neumann knew was not well defined and had many problems to be worked out.

Which is the issue I raised in the previous post -- what assurance do we
have that VN's approach - i.e., foundational assumptions - are the best
ones for a study of life?? Could they be more appropriate for physical
systems, and although extensible to life and society, not parsimonious
in these fields?? That's my suspicion, but again I would need to read
more VN to be sure, but will probably not do because of time
constraints. That exercise may be best left to someone interested in the
comparison - perhaps yourself.

> In several places von Neumann is clear about the inequivalence of construction and computation, specifically that formal and physical processes are different categories. He also says, “It is, of course, equally clear at which point the analogy ceases to be valid” (Von N, 1951. p. 318.) He discusses specifically the different effects of noise and how the choice of parts can alter the design of the constructor (the so-called “parts problem”), a problem that is meaningless in formal computation. That is why he said, “axiomatizing automata . . . throws half the problem out the window, and it may be the more important half.” Namely, why the parts of the construction are “the sorts of things they are.” (von N, 1966 (2) p. 77)

Well, this does sound to me very Rosennean. He is suggesting that the
"choice" needs to be part of the system. However, in what way does he
then recommend representing this phenomenon of "choice of the parts?" It
is one thing to say there is something missing (perhaps the most
important part) and another to come up with a way of dealing with it.
Did he do the latter?

> Judith: 2.) Both of those assertions are only possible from a mechanistic view point and will not change the foundational mistakes that my father discovered in his own scientific investigations. As such, the complexity that von Neumann is talking about bears no resemblance to the definitions that Rosennean Complexity deals with.
>
> HP: As I explain briefly above, nowhere can one find support for what you claim as the first assertion (equivalence of construction and computation). This was only Bob’s misinterpretation.
I don't understand. In the previoius quote, VN said that his idea of
"constructor" does not capture the "parts problem" i.e., the problem of
how the parts were selected and organized in the first place -- the
DESIGN of the constructor. That suggests very strongly to me that the
constructor is mechanical and that the non-mechanistic aspect of the
system is in what he is referring to as the design. If construction is
what the constructor does, then it is equivalent to computation, unless
one considers the design element - does he do that?

> On the other hand, von Neumann’s concept of “complexity” is certainly different from Bob’s. That does not invalidate it. It is simply that Von Neumann was asking a different question than Bob, and naturally he found a different answer. Von Neumann was essentially asking why in the presence of the inexorable noise inherent in the universe, life can grow in complication while non-living systems decay. Bob’s (M,R) systems did not address this question.
I believe it does, but one has to go farther with it than Rosen did.

> My interpretation is that Bob’s “inner workings” conceived of complexity synchronically or outside of evolutionary time, while von Neumann was thinking diachronically or of evolution in rate-dependent physical time.
Yes, that's my impression too. Within physical time, in my opinion, one
cannot escape computable mechanical structures but can at most make
simulations of reality. Rosen's meta-theory represented by the model
relation (interpreted ontologically) escapes physical time and thus can
get at complex aspects of design. The other approach, though laudable,
attempts to build into this kind of difference between various designs
by proposing there is a link between "complicated" mechanical structure
and true complexity of design as you defined it above. I understand RR
as arguing from first principles that there is no such link, except in
the limit. The gap may be closed arbitrarily small with enough clever
mechanical theory, but it will never bridge the gap fully. On the other
hand, RR's approach is wanting in another aspect, its ability to make
temporal predictions. So there seems to be a tradeoff between prediction
ability and completeness. In this sense, both views would be needed -
RR's on issues of design, and VN's in constructing predictive models. A
synthesis of the two would probably be very valuable, and someone should
attempt it (although not all of us, as there are other things to do also).

> HP: Yes, and I agree “correct” is a bad word. I should have said,
> “useful for answering certain specific questions about why the cell is
> organized the way it is.” (See my response to Tim about the relevance
> of von Neumann’s model.) It would more useful for this discussion to
> hear what other specific questions members of this list are trying to
> answer.


I'm trying to answer the question of how to design an information system
(for example as part of the Global Terrestrial Observing System) that
will propertly deal with natural ecosystem complexity. Since this is a
design issue, I have found RR's view extremely valuable. We are finding
that many mechanistic views of ecosystems have led us down a very
unproductive and expensive path in vein attempts to construct "good
enough" simulations of ecosystems. In fact, it appears that no matter
how close the simulation can get with further mechanical refinements,
the remaining gap, no matter how small, is catastrophic for management.
This is the non-linearity between Rosennean perspective and mechanical
perspectives and why I say they are not just "complements" nor are they
ever to be fully correllated. An entirely different approach to
management is needed. Building the theoretical basis for that is
extremely important, otherwise we have a lot of people telling Congress
and other funders that the whole problem is solvable from a mechanical
modeling perspective, and the money goes there because this is the
strongest hope of the ego, that we can do everything from a rigorous and
complete logical basis. The funding, of course, invokes the issue of
competition, and perhaps explains my earlier remark about that.

> Judith: Regardless, I would never, and will never, try to hide my father's conclusions from people in an effort to be politically correct.
>
> HP: I don’t know where that came from; but in any case it is my opinion that Bob’s fundamental conclusions will not be enhanced by continuing to claim that he invalidated von Neumann’s basic ideas, especially since they are not in fact inconsistent with his own with respect to the limits of physical description of life.
I would also suspect they do not invalidate any traditional ideas,
except the idea that traditional ideas apply to the whole system. That
is the part that is being invalidated - the conceit that the mechanical
program will eventually succeed or can be done well enough for practical
purposes. That conceit is true only in physical science, engineering,
product generation, etc.; admittedly most of human enterprise. But it is
not true regarding our management of whole systems, like ecosystems,
social systems, security systems, sustainability, etc. The problem with
discussing these theoretical differences in physics and theoretical
biology is that there the difference between approximation and reality
can be excused as part of the academic process. In the other fields I
mentioned, it cannot - we need to anticipate and manage correctly to
survive well and the two approaches correspond in the non-linear manner
I mentioned above. Furthermore, while I invoke these management fields
to demonstrate that there is a domain where the difference is important,
it is also important to investigate that difference in other domains
including physics and molecular biology. They may not produce as much
there in the way of man-made products or predictions, but the
completeness of our understanding of complexity and where it came from
is important to scientific acceptance in the other fields, which largely
defer to what is claimed to be our best scientific concept of reality to
base their approaches and designs.