Less a textbook, more of a jam session

This CD is the record of an interdisciplinary symposium on A-Life. Chaired by Daniel Dennett, it included Rodney Brooks, Murray Gell-Mann, David Haig, John Holland, Kevin Kelly, Pattie Maes, Bruce Mazlish, Marvin Minsky, Hans Moravec, Seymour Papert, Oliver Selfridge, Karl Sims, and Sherry Turkle. Clearly, then, the occasion provided both intellectual firepower and intellectual variety in good measure. I have no hesitation in saying that the face-to-face discussion must have been fun, and sometimes illuminating. And the students watching must have found it entertaining, and maybe instructive. But recommending people to buy the CD memoir to peruse alone at their desks is not so easy.

The discussion offers some provocative suggestions and occasional insights to those already acquainted with the field, and may encourage them to think about A-Life issues in a wider social and philosophical context as well as in relation to artificial intelligence. And the CD record of the event is enjoyable for users who know the individuals concerned. But it will be no more than a puzzling, though admittedly intriguing appetiser for others.

The puzzlement is not caused by jargon. There is relatively little of that, and a helpful glossary is provided. There is also a large bibliography, useful for people wishing to track the criss-crossing ideas. (The glossary, bibliography, and participants' biographies were prepared by Nick Jakobi. Such intellectual inputs would be acknowledged in a book, and in my view should normally be acknowledged in CD format also. OUP, as a respected academic publisher, could perhaps start a trend?) Rather, the puzzlement is aroused by the relatively informal manner of the discussion. On the one hand, more background knowledge is assumed than many users are likely to have. On the other hand, the discussion jumps around from one idea to another, inevitably giving short shrift to some of the topics concerned. The Cog and Cyc projects, for instance, are very poorly described, even though they are separately listed in the Section menu. Other topics on the Section menu include (for example) emergence, evolution, the Turing test, and the future. Some sections consist of only one subsection, others of four. (An irritating feature of the software design is that one cannot - or anyway, I could not see how to - go straight from subsection 1 to subsection 2 without having to go back to the full Section menu first.)

The project has both the strengths and the weaknesses of the symposium format, but in the CD version the weaknesses are more apparent. Many different issues are raised, and the existence of genuine, sometimes passionate theoretical disagreement between participants is made refreshingly clear. A "textbook" this is not. It is more like a jam session. And, as Dennett says: "Like any good jam session, if there aren't some flubs and some flat spots, then you haven't been bold enough." Flubs and flat spots there may be, but there are also many intriguing remarks, including the deliberately provocative "It's quite clear that we're shaping a new kind of humanity". The focus is more on the many unanswered questions in and around A-Life than on its actual achievements. But some thought-provoking achievements are described or illustrated: Karl Sims's work on the evolution of animated agents, for example, and Tom Ray's Tierra model of co-evolution.

In saying that Sims's agents are illustrated, I am putting my faith in the cover blurb and subsection headings, not my own experience. Both I and a highly CD-literate AI colleague tried to get the video and the sound to work, with no success. (We could not access the Introduction either, so I cannot say whether it provides a clear intellectual structure for newcomers to the field.) However, being familiar with Sims' videos already, I can assure readers that if they can get the CD to work they will be surprised, and very likely impressed, by his results. Multimedia offers exciting possibilities, but it is worth remarking that CDs can cause such difficulties. Almost anyone, anywhere, can turn the pages of a book.

What users of the CD certainly will not be impressed by is the discussion transcript. Non-sentences abound: "And so, you know, to say, well, you're putting a lot into that, but if we're to get - you know, we live in an amazingly rich world . . ." or "I want to, I want to switch our topic a bit because it's, we don't have that much time left in our, in our workshop and we were going to talk about the future" (this, by the way, at the beginning of a section). And the user is informed of LAUGHTER and APPLAUSE as well as the more defensible UNINTELLIGIBLE.

Such familiar speech-patterns are acceptable to the hearer - still more so, if the speakers' faces are shown in video. But the latter joy is not always available, as many sections of the script and soundtrack are accompanied only by a still photograph of a group of people sitting around a table. With or without full soundtrack or video, the choice to provide an exact transcript - with every "sort of" painstakingly and painfully recorded - was a mistake, as was the transcript-writer's "but they're really is something". Someone should have corrected the "they're", eliminated the "sort ofs," and pruned the screenful of effusive "thankyous" at the end. In sum, a few hours of amusement for the aficionado, and a source of scattered stimulation for the student - but an expensive and unsatisfying hors d'oeuvre for the novice.

Margaret Boden is a professor in the School of Cognitive and Computing Sciences, University of Sussex.