It is hoped that the re-unification may be achieved by relocating many research questions within the context of trying to build, or at least design, a complete working mind, for a human-like system, though obviously there will be intermediate stages involving the construction of simpler (but complete) minds, and also relevant research addressing philosophical and methodological issues.
For example, work on vision, or language processing, will not be relevant unless it is explicitly designed to show how such processes could be part of a complete agent, perceiving, acting, communicating, within an environment of some sort.
The symposium web site includes a directory containing online abstracts and brief information about presenters and a directory containing papers and poster summaries.
Contributors may from time to time produce updated versions of their papers, or may instead provide links to additional material, to be included in this file.
This file is expected to grow over time. It presents links to additional relevant material, not in any significant order.
Anyone, not only those who attended the symposium, who wishes to have a link added below to some online material relevant to the objectives of the symposium should write to me with a URL and a short paragraph describing the contents.
Follow the style of the entries below, and send your entry to A.Sloman@cs.bham.ac.uk.
AARON SLOMAN: revised version of slides used for plenary address at the conference. Also available as PDF version, and information about the Sim_agent toolkit used for the demos in postscript or PDF
One of the characteristics of such research is that it is based on the assumption that there are many problems which require a grasp of complex structures, and relations between those structures, and in some cases relations between the relations. These may be linguistic structures, spatial structures, mathematical structures, structures of complex actions, structures of kinds of reasoning, social structures, and many more.
A typical example would be the familiar geometric analogy tests of the form "A is to B as C is to which of D, E, F, G, H", where A to H are all pictures of some kind. Other tasks requiring a grasp of relations between structures include planning, synthesis and debugging of computer programs, scene analysis, understanding of sentences with complex structures, mathematical discovery, understanding the function of a complex mechanism on the basis of perceiving the shapes, debugging a faulty electronic circuit, and many more.
Work in this area (sometimes disparagingly referred to as "GOFAI") was very difficult, and progress was very slow, not least because the huge obstacles provided by the computing resources available at that time. For instance computers with a megabyte of memory were rare in AI labs before 1980, CPU speeds were so slow that it could take 20 minutes to segment a TV image that would be done in a fraction of a second now, and graphical input and output devices were too expensive for most research labs.
The slow, difficult, progress, and the rise of new fashions, e.g. for neural nets, evolutionary computation, general purpose learning mechanism, general purpose inference mechanisms, building toy robots, low level vision (partly because people learnt to program in C rather than AI languages) etc. meant that much important work on understanding and unravelling the complexities of domain-specific problems involving complex structures was left unfinished, and the training of students tended not to include the skills required for such work.
Now that far more powerful computers and software development tools are cheap enough even for hobbyists, and the use of AI languages does not involve unaffordable amounts of memory and computing power, it may be a good time to revive some of the old problems and the skills required for dealing with them, within the educational system. (Note that this is not a proposal for "grand challenges" for AI. The time is probably not yet ripe for such things.)
The tasks listed above may not be relevant to the task of designing a complete insect mind, but that does not mean they are irrelvant to human-like minds.
There is probably a lot of teaching material already to be found in departments where AI is taught, since not everyone has abandoned those themes. Some examples can be found in the Sussex University Poplog System and our extensions at Birmingham, available at http://www.cs.bham.ac.uk/research/poplog/freepoplog.html but there are probably many other sites.
I hope to collect suggestions of topics and pointers to resources.
MARVIN MINSKY'S WEB SITE: with many relevant papers.
THE COGAFF DIRECTORY: The Birmingham Cognition and Affect directory includes many papers relevant to topics presented at the symposium.
THE COMPUTER REVOLUTION IN PHILOSOPHY:
Philosophy, science and models of
mind.
This book, published in 1978 by Harvester Press and Humanities
Press, author Aaron Sloman, has been out of print for many years. It has
now been photopied (two pages side by side per A4 sheet) and may be
ordered for UK £10, to cover photocopying, binding and posting in the
EU. For airmail postage to other countries add £1. Please send orders
to:
Ms Ceinwen Cushway, School of Computer Science, The University of
Birmingham, B15 2TT, UK
EMAIL: C.Cushway@cs.bham.ac.uk
FAX: +44-121-414-4281.