[FRIAM] Morphogenisis

[jon zingale]

"""
Is carrying out an algorithm more like “computation” or is “building a
limb”? Is a salamander’s limb “computed”?  If so, who computes it, or is
that a violation of the language of computation.
"""

>From the summary of Chemero's "Radical Embodied Cognitive Science":

"""
Radical embodied cognitive science is a direct descendant of the American
naturalist psychology of William James and John Dewey, and follows them in
viewing perception and cognition to be understandable only in terms of
action in the environment. Chemero argues that cognition should be described
in terms of agent-environment dynamics rather than in terms of computation
and representation. After outlining this orientation to cognition, Chemero
proposes a methodology: dynamical systems theory, which would explain things
dynamically and without reference to representation. He also advances a
background theory: Gibsonian ecological psychology, “shored up” and
clarified.
"""

Which I find helpful to juxtapose against Valiant's "Probably Approximately
Correct":

"""
The assertion that the Halting Problem was not computable by any Turing
machine was identified with the claim that it was not computable by any
conceivable mechanical procedure...Extensive efforts at finding models that
have greater power than Turing machines, but still correspond to what one
would instinctively regard as mechanical processes, have all failed.
Therefore there is now overwhelming historical evidence that Turing's notion
of computability is highly robust to variation in definition. This has
placed Turing computability among the most securely established theories
known to science.
"""

Two steps in Turing's process I find worth highlighting are:

1. Abstraction of features of particular machines to the general.

2. Discovery of a limiting set of robust properties of generalized machines
such that these properties could be identified universally in any
sufficiently capable mechanical process.

In part, I mention the question of universality because (here in the still
hours of a sleepless night) I cannot help but feel that metaphors often
attempt to point to universals[1].

While I am never really sure that I get non-computation in the sense of
Chemero, chatting with EricC about Gibson makes me feel like I can almost
see it. Where Valiant emphasizes in computation that which is universal to
anything we can sensibly call mechanics, Chemero and others place the
universality squarely on the side of representation, or in some extreme
cases of nominalism, rejecting the universality altogether.

>From another perspective, a difference between these two points of view
relates to the question of agency. In Valiant's description, one doesn't
seem to care *what is doing the computation* as much as that a mechanical
procedure is executed at all. Alternatively, Chemero is concerned with
subjectivity. For him, there are well-defined agents and they have
environments. As far as I can read, each disagrees terminologically on
whether agent-based models compute. While Valiant is comfortable calling
what these agents do in their environment a computation, Chemero is not.

[1] Nick, regarding universals, while I have heard you denounce the
stability of a universal, like beauty, what is your bet on computation?



--
Sent from: http://friam.471366.n2.nabble.com/