Re: Life without evolution/evolution without life

[Howard Pattee <***>]

Tim,

You have to begin by asking: What is the question that von Neumann is
trying to answer? And also: What questions is he NOT trying to answer? Also
you must remember that in Theory of Self-Reproducing Automata (von N) are
informal transcribed lectures with the editor, Burks, filling in large
sections. Von Neumann?s discussions of self-reproduction extended over 4
years, so there are some differences of emphasis. One needs to read the
whole story. An additional source is his Hixon Symposium (von N, HS) talk
(Vol 5, No. 9, Jeffress, ed., Wiley, 1951).

Von Neumann?s fundamental question is: What is the threshold of complexity
of a system that can evolve increasing complexity rather than degrade?
Increasing complexity can be interpreted broadly as open-ended emergence,
endless variety, or evolutionary novelty.

He concludes that this will require a type of self-reproduction, and this
requires more than a symbolic Turing machine. He says, ?For the question
which concerns me here, that of self-reproducing automata, Turing?s
procedure is too narrow in one respect only. His automata are purely
computing machines. Their output is a piece of tape with zeros and ones on
it. What is needed for the construction to which I referred is an automaton
whose output is another automaton.? (von N, HS p. 315)

Consequently, he warns that with respect to his choice of parts for the
automaton: ? . . . this choice is neither rigorously justifiable nor
humanly unambiguously justifiable. All one can do is to try to submit a
system that will stand up under common sense criteria.? (von N, p. 77) That
is, it is not a formal logical argument as Rosen incorrectly assumes in his
critique.

What von Neumann did NOT ask was about physical implementations. As von
Neumann warns, ?One has resigned oneself not to explain how these parts are
made up of real things, specifically, how these parts are made up of actual
elementary particles, or even higher chemical molecules.? However, unlike
Rosen, he doubts that a relational model alone is adequate. In constructing
a relational (?axiomatic?) model, he says, ?One does not ask the most
intriguing, exciting, and important question of why the molecules or
aggregates which in nature really occur are the sorts of things they
are . . .? (von N p. 77)

I?ll skip von Neumann?s ?common sense? derivation here (I?ll do it later if
you wish). His essential conclusion is that even entirely abstract
relational models of replication require two disjoint categories of
components: 1) components that can be described temporally by a dynamics
(?constructions?), and 2) components that are non-dynamic structures
(?quiescent descriptions?) that can be described as memory-stored symbolic
constraints on the dynamic constructions.

Von Neumann did not discuss the physical conditions necessary for symbolic
descriptions. He simply assumed that his abstract concept could be
materially implemented. I took up the problem of the physical requirements
for implementing symbolic descriptions, and this is what led to Rosen and
my discussions of Hertz and his modeling relation. (My intellectual
dependence on von Neumann is described at
http://wwwc3.lanl.gov/~rocha/pattee/pattee.html) The best single Rosen
paper for illustrating how we agree and disagree is Causal Structures in
Brains and Machines, Int. J. General Systems 12, 107-126, 1986. Also this
paper most clearly shows how he misinterprets von Neumann (p. 114).

Von Neumann?s admittedly incomplete discussions of replication are often
credited as the beginnings of artificial life studies as well as a stimulus
for the discipline of biosemiotics.
I would emphasize that Von Neumann?s conclusion is entirely consistent with
Rosen?s conclusion that dynamics alone (the generalized Newtonian paradigm)
is too restricted to describe life, and that complex systems require more
than one model. However their logics and levels of abstraction are quite
different. Von Neumann is actually making what I think Rosen recognized as
a relational model that competed with his own (M,R) ? systems. They are
both abstract models that do not employ any particular material
representation, but that would hold for any material embodiment. This might
explain Rosen?s motivation for continually objecting to it on the same
grounds for over 30 years.

Howard