<nettime> Semiosis, Evolution, Energy - missing interviews.

[diana@mrf.hu (Diana McCarty)]

{Dear nettimers,
These two interviews appeared in part in the original posting
by Nell (Petronella) Tenhaaf, but were accidentally cut - due to
technical difficulties (diana at the helm), the introduction is the
same, but there is a full extra 17k of text that was missing.
The original post contained Nell's interviews with Stuart Kauffman,
Claus Emmeche, Arantza Etxeberria, and Roberta Kevelson.
Apologies to Nell, and confused readers. ~ diana}


These interviews took place during "Semiosis,
Evolution, Energy: Toward a Reconceptualization of the
Sign," a conference held at Victoria College, University of
Toronto, October 17 - 19, 1997.  This was a cross-
disciplinary meeting organized by Edwina Taborksy of
Bishop's University and Barry Rutland of Carleton
University, to investigate the idea that "all phenomena are
energy configurations belonging to one and usually more of
three distinct codal orders: physical, biological and
conceptual."  In particular, theoretical biology crossed
paths here with problems that have been formally studied in
the field of semiotics: interpretation, meaning and
subjectivity.  My own interests as an artist and writer have
been located in the territory of crossover between biology
and subjectivity for some time, although not from a base in
semiotics per se.  More recently my work has become focused
on issues of representation in the field of Artificial Life
and possibilities for engaging with these issues in my own
practice.

     Artificial Life or Alife is a set of computer-based
practices that took form in the early 1980s in the southwest
of the U.S., incorporating ideas from complexity theory,
chaos theory, Artificial Intelligence and theoretical
biology, especially evolution and genetics.  Alife is
concerned with synthesizing life-like phenomena in
artificial media such as computers or robots.  Currently, it
tries to bring understanding to issues of how the real world
works, but at its inception Alife programmers as well as
theoreticians were committed to the idea of making synthetic
life-forms that would literally be successors to biological
life-forms.  Evolutionary computation, or artificial
evolution which is discussed below, is one of the key
methods of Alife.

     One issue at the center of current theoretical biology,
that also affects Alife practices, is the tension between
the classical Darwinian evolutionary principle of natural
selection on the one hand, and the concept of self-
organization in nature on the other.  The latter is the idea
that implicit form emerges spontaneously at all levels in
the natural world, from the chemical to the organic.
Stuart Kauffman places self-organization at the centre of
his theory that life is not the result of randomness, but is
a probable emergent feature of the universe.  Further, he
has formulated the theory of the autonomous agent, a
construct that enables researchers to study and propose
answers to questions of self-organization and the origins of
life: what is a basic self-organizing unit, how does it self-
perpetuate, what are its sources of energy, and what forms
the constraints by which it is bounded?  This hypothesis
arose from theoretical biology and theories of dynamic
systems, but it has since been reverberating and resonating
throughout many other material and conceptual practices.


     Claus Emmeche is a Research Fellow and head of the
Center for the Philosophy of Nature and Science Studies in
the Niels Bohr Institute, University of Copenhagen.  A
theoretical biologist and philosopher of science, his area
of research is the semiotics of explaining emergent
phenomena.

NT:  There are actually two aspects of what you spoke about
that I'm particularly interested in.  One is the idea of
inner qualities of the organism that one could model, and
the other is the modeler's "frames of perception" that
evolve in relation to this.

CE:  Yes, how to conceive of the mental models that the
organism builds up in ongoing interaction with the
environment.  The first question is about qualitative
experience, or qualia as the philosophers call it, which
simply means that you feel something, you just experience it
directly.  You can feel pain or hunger or thirst, this is
just the technical term philosophers use for something we
all know because we all experience things: what it is like
to smell a good cake, what it's like to taste it on your
tongue, what it feels like to have a pain in your stomach.
These are immediate expriences, afterwards you can try to
conceptualize them or put words to them.  But the very deep
problem of having a biosemiotics, that is a biological study
of semiotic interactions, or sign interactions, is to
capture these experiential qualities of living animals.
Maybe single-celled organisms could also have these kinds of
qualitative experiences, although they may be very primitve.
I mean a single bacterium may simply experience how it feels
to be in a lower or higher concentration of glucose.

NT:  Speaking of that makes me think about a paper that I've
read by Lewis Wolpert [a developmental biologist in the UK],
concerning his conjecture about the first sense of top and
bottom, the first sense of axis in a single-celled organism;
and it extends into the drosophila fruitfly and beyond, to
mammals.  He speaks about the place where this organism
touched the floor of the primordial soup -- that point of
contact is a kind of inner representation.

CE:  And the idea of having an inner representation is of
course very important to all of this study.  We have the
idea that you make somehow an internal scene.  You would
like to model the world or make a kind of replica of the
world by building up for yourself, or for the animal itself,
a scene or a mental map or whatever you'd like to call it.
And the problem there is that you can do models, for
instance, by using robots or using simulated animals on a
computer where you have these small creatures running around
on your computer screen.  And you also have for each
creature a model of how this creature is representing its
world, that is, its immediate environment and its neighbors
and its predators, enemies and so on.  But this is the
functional aspect of the representation, this is a question
of what the algorithm is or what the program is, what kind
of physical tokens this organism is using in order to
represent its world, in order to somehow gain knowledge
about the world.  But this is just the outer side of the
coin, because when we are talking about signs in the
universe or signs that we recognize as signifying something
for us, we are talking about a coin with two sides to it,
the external side and the internal side.  And the internal
side is really what it feels like for the organism to have
these experiences.

That is what I see as the really hard problem, I'm not the
only one who points to it as a really deep and hard problem
to solve.  Part of the problem is that you're also
interested in somehow bridging the gap between human minds
and animal minds.  But animal minds can come in various
degrees of complexity.  So you can easily imagine
chimpanzees having a mind which in many respects is very
similar to the human mind; then we can go down in complexity
and say what about dogs, what about rats, what about mice,
what about insects, bugs, do they have minds in some sense
and how should we conceive of that.  And this is the problem
of the scale of continuity between what we as humans
experience, because for our experience we have both the
external and internal point of view.  When we as biologists
do modelling we always capture these things from the
external point of view.

NT:  Now is that because of the rules of science, or is that
just because it's the only thing that's possible?

CE:  I think it's primarily because it's the only thing
possible.  I mean, we have to start somewhere.  As
scientists we try to figure things out in a precise way that
we can describe to other persons, because if we couldn't do
that, it wouldn't be science.  We have to be able to
describe precisely what we are doing, and when we want to
describe precisely what a little animal in an environment is
doing, I mean if we make this as a model, we have to
describe precisely what the mental representations are in
that animal.  And we can do that for some simple models
where we can, for instance, use neural networks in a
physically embodied system like a little robot, a little box
that you can follow.  You can open it up and analyze its
network, what it has learned.  So this is the external side.
But you are not sure, I mean you cannot know if this little
artificial box is really experiencing anything in a
qualitative way.

NT:  Now just a comment about that.  Because one of the
notions I came across at ECAL this summer was that if you
are modeling in an internal and an emergent way, even if
it's initially algorithmic but then it develops properties
that you haven't pre-determined, isn't there an additional
problem of reading back what has emerged?

CE:  That's a big problem, because all this research is
really trying to get after emergent properties.  For
instance, you can have the little bug become better and
better at acquiring knowledge so as to avoid obstacles in
its environment and so on, finding food, remembering where
the good food was and where the bad was.  So you can
conceive this network as evolving emergent representations.
But, it's very hard to avoid putting our own conceptions
into that network.  So that when we look upon the network --
and this is really just a simple network of nodes or you
could call it neural cells and the connections between them
which have a certain strength -- but when we look upon it we
would like to make sense of this little network, of this
little bug.  When we do that, we come very close to making
the fault of anthropomorphizing the little bug.  This is
really a hard problem, because of course we would like to
figure out how this insect or little animal can go about
doing his things, in a way that doesn't involve any
mysterious principles.  But we cannot know beforehand how
this creature somehow chooses to configure its own world.

I can give you an example.  The things which are important
for us when we walk around in the sunshine are not the same
things which are important for the bat when it flies from
tree to tree in the dark, depending on its specific sonar
systems.  The bat's world is radically different from our
world.  Of course it's the same physical world, but we do
not share the same Umwelt as the bat, its internal
perceptions, the bat has a bat-specific Umwelt and we have a
human-specific Umwelt [this is a term proposed by Jakob Von
Uexk=FCll in the early 1980s in the semiotics of biology, to
describe how an organism lives in a subjective universe
which is a niche within the environment].  So every time you
have a new species you have not only its morphology and its
anatomy, but you have also a species-specific Umwelt.  That
is, how is the world experienced by this animal?  We have
certain clues to figuring out what the bat's Umwelt is
because we can do experiments, we can measure how good it is
at using its sonar system in order to detect differences in
its environment.  But this is only an indirect way, again we
can by this indirect way figure out its Umwelt.  But we can
never know how it feels for this bat to be such a creature.
So this is the hard problem.

NT:  I guess the other question that I framed is a bit of a
gross question because it takes these ideas about
subjectivity and kind of butts them up against the
historical objectivity of science.  Do you see any problem
in that respect?

CE:  Yes, I see that many of the attempts in the new fields
of complexity studies, artificial life, cognitive science,
are really trying to explain subjectivity.  But the question
is, whether we possess methods of explanation which are
really good enough at dealing with this specific subject
matter, that is, subjectivity, experience, qualitative
feeling, what it feels like to be something.  Because
traditionally in science we want to be very precise, we want
to explain things in objective ways, we want to have a sort
of behavioristic or external perspective.  At this meeting
here, there is a discussion going on about whether we can
configure some internalist way of having emergent
explanations.  That is, when we understand a system, this
understanding depends on the creation of new ideas or new
concepts within our own minds.  And I think this is a
possibility we should be serious about, that we could
enlarge the notion of natural science not only depending on
objective modes of thought but also somehow involving
methods from the humanistic sciences, like the idea of
empathy or the idea of trying to interpret something which
makes sense if you can really figure out what this small
world of another organism looks like.  But of course it's
very controversial as to whether we are still doing science
when we do these things.

NT:  It's a mix of the so-called hard sciences and soft
sciences, isn't it?

CE:  It is.  And most of these scientists still want to keep
it within the basic explanatory framework of hard science.
This is also what you see in the Santa Fe complexity
studies.  They are making attempts to do mathematics of
complex systems, physics and biology of complex systems,
that is, to use the methods of science in order to
understand these very strange phenomena.  Because it is very
strange that such a thing as a human brain, which is simply
just biological cells and their communication, can create
mental phenomena.  I mean, this is still a mystery.

NT:  As Stuart Kauffman said yesterday, there is so much at
stake, because you do need to install a sense of synthetic
science instead of a reductive and atomistic science to even
start this work.  So I guess that's why the complexity
theorists are so insistent on staying within the framework
of science.

CE:  That's right.  In order to be synthetic, you have to
have a notion of what this kind of whole you're trying to
explain really is.  And once you begin to describe that in
detail, you somehow get involved in the reductionist method
or the decomposition method.  So I agree with many of the
people from the complex systems sciences, that we should try
to combine the reductionist method of decomposition with a
new way, by simulations or new constructions or maybe new
ways of experiencing things, of synthesizing or developing
the very objects that we want to explain.  And this is what
you see in Artificial Life, we create creatures.  These are
not creatures we can see in nature, these are totally
artificial creatures.  Basically they are built upon
abstract ideas which are put into computer programs.  But if
we are lucky enough, you sometimes see within these models
that new emergent and very surprising phenomena appear, for
instance, the ability to self-reproduce, the ability to
evolve and so on.

NT:  Well I think that the concept of the autonomous agent
as Stuart Kauffman proposes it and then, as Jesper Hoffmeyer
says more metaphorically, the concept that ideas are
autonomous agents, I think this is really rich in itself
because it offers a way to start reintegrating things at
different levels. [Jesper Hoffmeyer is a researcher and
professor in the Institute of Molecular Biology, Unversity
of Copenhagen].

CE:  Yes, I think so.  And there it's important to walk on
two legs.  That is, you have to have both the traditional
scientific way of explaining it, and then on the other hand,
you have to be very aware of the difference between your
model and the real world, which is always much more complex
than your model.  Here I think of Jesper Hoffmeyer's ideas
about agency as something which is very much dependent upon
complex surfaces of semic interactions, that is, surfaces at
which you have signification processes in the semiotic sense
going on, and where the various surfaces really define the
agents, I mean the inner side of the agents.  This is one of
the promising notions, I think, in the field.

                         ---------------------

     Arantza Etxeberria is in the Department of Logic and
Philosophy of Science at the University of the Basque
Country, in Donostia, Spain.  She spoke about problems in
Artificial Life practices, in particular how to build
"embodied" agents in artificial worlds through the
integration of physical properties of the model's
materiality, and how to overcome adaptationism in the design
of evolutionary models.

NT:  You were saying in your talk that you think there's a
place for art in the practices of artificial evolution.

AE:  Well I was talking about sighted evolution.  I was
making a distinction between the idea that evolution is
blind, and sighted evolution.  The thing is. if you take
natural selection in any context, it is very difficult to
make it really blind.   We assume that it's blind, but even
biologists doing models have a rough idea of what the
fitness is.  That's assumed to be exerted by the
environment, but it's thought out in advance and then seen
as just from the outside.  When we want to do an artificial
evolution, our biggest problem is how not to intervene, how
to get the system to use its own proper fitness function.  I
think that the most interesting solution for this at the
moment is coevolution, because it's so difficult to get the
agents interacting with an environment so that they all have
a common history, so that the organisms or the simulated
agents and the environment will inform each other.  Lots of
people have tried to do this by inducing coevolution between
two kinds of agents.  Sometimes it will be predator and
prey, or mates for sexual recombination.  And it's been very
interesting to see the use of games to evolve agents that
are playing games with some competitor, so that all the
conditions of the game change when the competitors change
their strategies, so that the strategies are not fixed
because you always have to account for what the other is
going to do.  Both are evolving, both are going to try to
make it more difficult for the other.  So that's sort of a
common history.

NT:  But that just makes me think, before we cycle back to
where the art comes in, that makes me think that there seem
to always be incredible presuppositions built into the
models, for example those kinds of models usually assume
competition.  At ECAL, there was a really interesting paper
about the emergence of cooperation from a competition model,
without game theory, without preordained rules.  One is
always backtracking to what the basic assumptions are.  I
know and you would probably agree, that in the evolutionary
computation world, which you call artificial evolution but
we're talking about the same thing, in that world there's a
generally accepted metaphorical use of Darwinian theory.

AE:  I think that natural selection has a big component of
competition.  And for example, other ideas of biology like
self-organization are much more into cooperation.  But
that's very difficult to achieve if you start out thinking
about Darwinism, I think.

NT:  You're coming from two different poles there, from the
forces of evolution on the one hand and internal self-
organization on the other.  So you're saying that within
Alife, coevolution is an interesting modeling strategy.
Then you've proposed that there's  a potential artistic
strategy using intervention in the evolution proces.

AE:  Well, no, I was saying two different things.  I was
saying that artificial evolution has been used with several
purposes in mind.  I think that's very important, in fact
it's a very pluralistic field.  So with respect to sighted
evolution, I was saying that some artists have used models
of evolution in which the selection is actually done by the
modeler, very directly.  And those are interesting models.
The final product, in what I was referring to as art, is
different.  But I think that it's very important to take
into account that Artificial Life models can have very
different purposes, maybe not only being a contribution to
theoretical biology but also for example to get art into the
picture.  And not only art, but also models for
understanding education, some people have worked on that.  I
think that very broadly we could say Artificial Life
productions are either models that try to grasp the nature
of certain phenomena in the world, or something that I would
call instantiations.  That has to do with art, that is what
I would call poetic science in a very appreciative way,
because a lot of people say poetic science in a very
disparaging way.  And I think that it also has to do with
the first goal of Artificial Life, of exploring life-as-it-
could-be.  Because in fact we are not trying to model
anything, but we are trying to understand how certain
phenomena happen, through artificial models, and that kind
of undertanding is either scientific by creating new theory
or new models, or even artistic.

NT:  But do you think it's interesting artistically because
art can always be interested in new ways of creating models
of life?  Or do you feel it's because these biological
issues, or scientific issues in a larger sense, are the
current issues of our time, in the way they shape the
material world through biotechnologies, or reproductive
technologies.  Can we dig a little bit further at why you
feel it's so intereting for art?

AE:  Well, it's maybe neither of the things you said.  From
the very beginning, there has been a very big discussion in
Artificial Life as to whether the models that people were
doing were actually life or were not life.  That's the big
discussion between what they call "strong artificial life"
and "weak artificial life."  Well I think that discussion is
sort of stupid, or nonsense.  I think that all of them are
productions.  But, it's very interesting to analyze the
purposes we have when we are building the models.  And those
purposes are of course, understanding phenomena which are
complex and for which we don't have good analytical
scientific models.  My thing is that, in my opinion, it's
very difficult to get good reproductions of life in
simulations.  I don't believe that computational models can
reproduce life, that you can produce something yet it's
living.  But you can get some fantasia, as Marcel Danesi
[Professor of Italian and Semiotics at U. of T.] was saying
yesterday.  You can have an understanding through them, and
I think that that pluralistic way of understanding models,
according to the purposes of the modeler and the kinds of
things they want to achieve through the models, can produce
an increasing ontology in the world.  Actually artificial
systems have given us more ontology, more things that we
have to analyze so as to understand what they are.  So now
there are certain artifacts we don't know.  Art can enter
into that picture because these are things we interpret once
and need to interpret again, which is a source of
creativity.  Maybe that creativity is also linked with
understanding.  I really think that there are very different
ways to access these new phenomena, this new ontology that
is being created.  And it's important to be very pluralistic
and leave aside the discussions about whether what we're
doing is really life or not.  Because that's not going to
take us anywhere.

NT:  That's a really interesting way of putting it.  Within
the art practice that I'm familiar with, a recent phenomenon
was incredible fascination with media production.
Deconstruction is so tied up with that idea, with seeing how
all of the real is already mediated.  Artists take that up,
consider it and communicate it.  If we're now involved, as
you say, in a growth of the ontological or a growth of
artificats within the technoscientifically-mediated real, of
course artists would also be engaged with this next level of
mediation of the natural.



---
#  distributed via nettime-l : no commercial use without permission
#  <nettime> is a closed moderated mailinglist for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: majordomo@icf.de and "info nettime" in the msg body
#  URL: http://www.desk.nl/~nettime/  contact: nettime-owner@icf.de