Ruth Nuttall
Title: Artificial Minds Author: Stan Franklin Publisher/Date: The MIT Press (Cambridge, Massachusetts), 1997 Price: GBP14.95 Reviewed For The 'Artificial Intelligence' Journal.
Stan Franklin, Professor of Mathematical Sciences and co-director of the
Institute for Intelligent Systems at the University of Memphis has written
this book in the hope of providing an overview to the relatively new field of
artificial systems. This area of research has no standard textbooks --
`Artificial Minds' is an attempt to fill this void.
The author was motivated to write this book by the desire to bring
together the various fields which comprise the wider field of cognitive
science. Cognitive science has important implications on how one views the
mind, and the newest research has doubly important philosophical implications.
The author also has his own agenda about the nature of the mind. He argues
that rigid Boolean distinctions between mind and non-mind should be rejected
in favour of a continuum from less to more mind. However, this agenda does
not bias his overview of cognitive science, as the author is careful to
highlight what is his opinion and what is not. The book takes the form of tour
through Artificial Intelligence (AI) and Cognitive Science, beginning with the
history and tradition of AI and then moving on to cover each major development
through to the present day. It ends with a discussion of the future of the
field, and whether the author has proved any of his own views on the nature of
the mind.
Chapters One, Two and Three discuss concepts of mind, in order to
clarify what it is that Artificial Intelligence is trying to recreate.
Chapter One also outlines the seven points about the nature of mind that the
author is simultaneously trying to justify, so that the reader is aware of
this and can judge for themselves whether these points are merited.
Chapters Four and Five discuss one of the traditional areas within
Artificial Intelligence -- symbolic AI -- and the first major philosophical
debate between AI researchers. This debate considers whether or not the
mind can be recreated using only symbolic AI. The conclusion the author comes
to is the standard view that explicit rule based systems have perhaps too many
problems and inadequacies to be the ideal candidate to create an artificial
mind with.
Chapter Six introduces a major development in AI which this first debate
engendered -- Connectionism, which is a completely different approach to
creating artificial systems. This chapter provides a solid introduction to
the field, explaining the basic idea -- to use the brain as a model of mind
-- and the mathematics underlying this in an admirably clear fashion. This
exploration of the Connectionist view leads on to the second major debate in
AI -- whether Connectionism can model mechanisms of mind that symbolic AI
cannot, or whether it in fact adds nothing new. Chapter Seven explores
this issue, forming an analysis of the critics of Connectionism, including
Fodor's well known attack and Chalmer's equally infamous defense of
Connectionism. This chapter concludes that both symbolic AI and Connectionism
have valuable contributions to make in AI.
After Connectionism, the next major development in AI was evolutionary and
genetic systems, and these are dealt with in Chapters Eight and Nine.
Together these areas have the generic name of Artificial Life (Alife). As a
very recent aspect of AI there are few standard AI textbooks which adequately
cover this subject -- for example `Artificial Intelligence' by Rich & Knight
has very little about Alife issues. In these two chapters the author covers
the main issues and work in this area, and provides many references to the
main texts in Alife.
The next four chapters deal with more exotic and disparate approaches to AI.
Chapters Ten and Eleven discuss approaches from Minsky's multiplicity of
mind model, through Jackson's Pandemonium model of mind to Brook's subsymbolic
models of mind, and his subsumption architecture. Chapter Twelve deals
with the important area of dynamical systems, and also explores some
implementations of models of mind, such as Edelman's `Darwin III'.
Chapter Thirteen deals with creating artificial memory and creativity. This
is an important inclusion, as it is clear that an artificial mind would be
incomplete without memory, and also somehow less than human without any notion
of creativity. Most AI textbooks ignore or gloss over these issues, so it is
pleasing to find a chapter devoted to them here. At this point, the author
argues that all the various mechanisms of mind explored so far tend to back up
the notion that the mind is comprised of a multitude of differing mechanisms.
This is a valid point to make, and is well argued by the previous thirteen
chapters.
And so, one comes to the last major debate in AI history -- that of the issue
of representation. The issue is whether representation is needed to model
the mind, or as Brooks would argue representations `get in the way'. In
Chapter Fourteen, by presenting many examples of representational and
non-representational systems, the author argues that both are needed for a
whole and accurate artificial mind.
The penultimate Chapter Fifteen, is a slightly spurious and out of
place discussion of future developments in AI which are highly fanciful and
generally unjustified. As quite a short chapter, it would not have affected
the book if it had been left out. However it does serve to give a flavour of
some of the more exciting aspects of the future research in AI. The final
Chapter Sixteen reviews how the author's initial claims about the mind
have been backed up by the previous chapters. It becomes clear that all of
his assertions about the nature of the mind have been thoroughly explored and
backed up.
As a first edition, this book has admirably few typos -- in fact the only one noticed was on pg 169, paragraph three, which should read `to some small degree'. Stan Franklin's informal yet educational writing style is conducive to explaining the modern complex of fields which comprise the emerging new paradigm of mind which is outlined in this book. It serves the dual purpose of exploring the subject matter, and arguing for a new view of the mind. The book has been written such that neither of these purposes interferes with the other. This book would make a useful introductory text for a university course in AI, as well as a useful aid to researchers in any area of Cognitive Science.
Brooks, R. A. (1991) Intelligence without Representation.
Artificial Intelligence, 47.
Chalmers, D. J. (1990) Syntactic Transformations on
Distributed Representations. Connection Science, 2: 53-62
Edelman, G. M. (1992) Bright Air, Brilliant Fire. New York,
Basic Books
Fodor, J. A. & Pylyshyn, A. (1988) Connectionism and Cognitive
Architecture: A Critical Analysis. Cognition, 28: 3-71
Jackson, J. V. (1987) Idea For A Mind. SIGGART Newsletter,
181: 23-26
Minsky, M. (1985) Society of Mind. New York, Simon &
Schuster
Rich,E. & Knight, K. (1991) Artificial Intelligence 2nd ed.
New York, McGraw-Hill
