Abstract
In our attempts to understand the evolution of biological, cognitive
and cultural systems, critical questions arise concerning the origin
of meaning. I argue that the key to success in attempts to create
computational systems that exhibit the same capacities as their
natural counterparts to evolve new and creative ways of interacting
with their environment, beyond that which is simply “programmed into”
the system from the start, lies in answering these questions.
The nature of the problem is laid bare when we consider the origin and
evolution of life. A fundamental question is this: how is it possible
for organisms, that follow their own goals and behave according to
their own rules, to emerge in a world governed by the laws of physics
and chemistry? More generally, how can agents and agency emerge in a
system governed by universal laws? And even once our agents have
emerged, how can the evolutionary process produce new agents that
interact with their environment through previously unexploited
modalities?
In this paper I describe work on a novel modelling approach which aims
to solve these problems and thereby allow us to produce artificial
evolutionary systems with greatly improved creative evolutionary
potential. This perspective sees organisms as entities whose
phenotypes are embedded within an environment viewed as a dynamical
system, and whose genotypes interact with the environment by
specifying constraints upon its dynamics, thereby generating the
phenotypes. That is, the abiotic environment has its own dynamics and
self-organisational properties; genotypes act to “sculpt” these
pre-existing dynamics by supplying constraints. From this point of
view, the most important distinction is not between organisms and
their abiotic environment, but rather between the environment as a
whole (including organism phenotypes) and organism genotypes.
Elsewhere I have presented initial results from a model based upon
this perspective, and demonstrated simple examples of the evolution of
new sensors and effectors, and of genome-regulated self-stablising
behaviour. Going further, we can generalise this perspective; in so
doing, we may find useful connections and analogies between
biological, cognitive and cultural systems, and thereby gain a better
understanding of how creativity may be instilled into artificial
systems.
The generalised picture describes a situation in which the constraints
of the system initiate dynamics, and the dynamics may feed back to
affect (select or modify) the constraints. In a situation such as
this, the system may exhibit behaviour which cannot be explained
purely by the laws of dynamics, but only with reference to the
particular history through which the system has evolved from its
initial to current state. This mutual interaction (or “creative
dance”) thereby brings forth novel forms of behaviour, the meaning of
which can only be understood by considering how the dance itself has
evolved over time.
This general description could be applied to a variety of other
systems, including the development of human cognitive processes, and
the development of human cultural traditions, institutions and
artefacts. Consideration of the extent to which such analogies hold
between these very different systems, and the commonalities and
differences between them, will surely lead to a much deeper
understanding of the generative causes of novelty and creativity, and
the origin of meaning, in natural systems. And such understanding
will suggest ways in which we may create artificial systems with a
much deeper capacity for creativity than exhibited by previous
attempts.