Artificial Intelligence
----------------------- Considerations: a brief starter by James Coupe (1/12/03)------------
Discourse centred upon representation and notions of selfhood must inevitably
take account of perhaps the most romantic attempt to construct a reality in
the last half century: artificial intelligence (AI). AI began as the science
of making machines that can do the kinds of things that humans can do. Representation
is a central concept within this model of AI, with computer programmes being
designed to construct, adapt, and link representations in the production of
intelligent responses. This linking of representation with intentionality has
also proved to be one of AI's greatest problems. If a system is programmed to
recognise particular
symbols in the world -- such as a chair, a bed, a word, etc. -- then how can
we get it to recognise it on its own terms rather than in terms pre-defined
by the system's designer? Within such a system, the system becomes a representation
of the designer's intelligence rather than something with its own intelligence.
For instance, if the system can't lie down, then why should it interpret the
bed as something to lie on? What is occurring in such instances is an anthropomorphising
of artificial intelligence -- building machines that try to simulate human behaviour,
and override their natural impulses as a result.Figure 1: Classic AI Representational
Model. There are direct mappings between objects in the world and the designer's
own internal concepts, and between the designer's concepts and their counterparts
in the AI system's representational domain. There is, however, no direct, designer-independent,
connection between the AI system and the world it is supposed to represent,
i.e. the AI system lacks ‘first hand semantics’ or ‘contents
for the machine’.
The approach that we are pursuing within I, Project is that if a machine is
to be intelligent then its logic should be based upon itself rather than a human
being. Any machine that is able to display human intelligence would always be
a simulation, whereas a machine that had its own logic would not be attempting
to imitate or represent something else, and therefore could arguably be considered
as "real". This approach can be aligned with developments that have
been made in AI over the last
twenty years to deal with the fact that classical AI systems did not have the
capacity to relate their internal processes and representations to the external
world. Searle, for example, criticized purely computational AI systems as "lacking
intentionality" in his famous "Chinese Room" argument, but believed
that physical systems causally connected to their environments would have the
potential for intelligence. That means, instead of accusing AI of being materialistic
(for its belief that(man-made) machines, could be intelligent), Searle actually
accused AI
of dualism, for its belief that disembodied, i.e. body-less and body-independent,
computer programs could be intelligent.
Such an approach, understood as the pursuit of situated and embodied intelligence,
agree that AI should be approached from the bottom up; first and foremost through
the study of the interaction between autonomous agents and their environments
by means of perception and action. The idea is that these autonomous agents
can actually ‘experience’ and interact with their environment rather
than simply manipulate its symbols. In other words, artifacts are no longer
expected to learn human language, but their own language, i.e. a language that
is about ‘the world as it appears to them’ and that helps them to
communicate with other agents (no longer
humans) in order to better cope with that world. Recent research into such
a model of AI, as carried out by the likes of Noel Sharkey and Rodney Brooks,
has focused around the building of robots. The robots are programmed to construct
their own world-view of their environments, and respond to them so as to avoid
obstacles, follow objects, etc. Nevertheless, such an approach is limited by
the fact that the robots' behaviour is understood and interpreted as a success
or a failure according to human criteria. The robots can be seen to imitate
real-world organisms such as insects, and so consequently do not have their
own intrinsic "life tasks".