Life-as-it-could-be

ARTIFICIAL LIFE Edited by Christopher G. Langton MIT Press, Quarterly, $125 (institutions), $45.00 (individuals), $25.00 (students) ISSN 1064 5462

The first three issues of this new journal consist entirely of overview articles written by members of the editorial board. By adopting this tactic, Christopher Langton is able to indicate the scope of artificial life research.

And the scope of the field is wide. As Langton states in his inaugural editorial, "artificial life'' means different things to different people, and a concise definition remains elusive. Broadly speaking, artificial life research involves the study of synthetic systems which exhibit "lifelike'' behaviours. The synthetic systems could be physical hardware, computer software, or biological "wetware" such as laboratory preparations of macromolecules or cell cultures. The common intention in such studies is either to advance theoretical understanding of biology or to help develop new engineering techniques. To paraphrase from an introductory article Langton published in 1988: the aim of artificial life research is to study life-as-it-could-be, in order to advance our understanding of life-as-we-know- it; rather than viewing life as a property peculiar to carbon-chain chemistry, life is viewed as a property of the organisation of matter. No matter what the matter may be.

Under this view, it is possible that computer software entities, competing to survive and reproduce in some virtual environment, could be considered to be alive. As is argued in the paper by Thomas S. Ray, such software systems are not models or simulations of life on earth; they are independent instances of life. Clearly such a prospect raises technological, philosophical, social, and ethical issues, all of which are also within the stated scope of the journal. Philosophical issues are explored further, and with admirable balance, in the three papers by Daniel Dennett, Eric W. Bonabeau and Guy Theraulaz, and Stevan Harnad: in particular, Harnad questions the assertion that virtual environments could support instances of life.

Despite the welcome shown to philosophical debate, there is some tension caused by the unresolved issue of what it is to be alive: the notion of synthetic "lifelike'' systems is very broad. That this is true can be seen from the diversity of topics addressed. These include: "Discussion of the role of artificial life in biological inquiry" (papers by Charles Taylor and David Jefferson, by Ray, and by Walter Fontana, Gunter Wagner, and Leo W. Buss), "Cooperation and community structure in artificial ecosystems" (by Kristian Lindgren and Mats G. Nordahl), "Molecular evolutionary biology" (by Peter Schuster), "Morphogenesis" (Przemyslaw Prusinkiewicz), "The relationship between artificial life and artificial intelligence" (Luc Steels), "Artificial neural networks that exhibit cooperative behaviour" (Michael G. Dyer), "Adaptive autonomous agents" (Pattie Maes); "Chaos in evolution" (Kunihiko Kaneko), "Genetic algorithms" (Melanie Mitchell and Stephanie Forrest), "Computer viruses" (Eugene H. Spafford), and "The broader lessons of decentralization" (Mitchell Resnick). There is also a brief multiple review, by David G. Stork, of 11 books on artificial life and related topics.

Thus the study of "lifelike'' systems can be seen to cover a wide range of issues. From Schuster's discussion of molecular evolution in the test tube and template chemistry, which aims to generate replicating structures that have nothing in common with biological replicating molecules; through Prusinkiewicz's reaction-diffusion models of cell differentiation in morphogenesis, which give impressive matches between observed and predicted pigmentation patterns on mollusc shells; past the complete autonomous agents discussed by Steels and Maes, which have animal-like capabilities for adaptation and survival in uncertain and unforgiving environments; up to (and in principle beyond) the artificial ecosystems described in the papers by Lindgren and Nordhal, and by Ray.

As is common in MIT Press journals, the production standards are very high. While it is hard not to be impressed by the breadth of the work presented in this collection, and by the depth of most of the articles, there may be clouds on the horizon for this journal. To be sure, Artificial Life as a research endeavour has attracted a fair amount of public attention in recent years, with the publication of several popular science books offering descriptions of its roots, its current state, and its possible future (a good example of which is Steven Levy's Artificial Life: The quest for a new creation, 1993). But purchasers of popular books do not necessarily constitute much of a market for an academic journal. For an international journal to have good long-term prospects, it requires among other things a clear identity. My suspicion is that the scope of Artificial Life - both the journal and the field it speaks for - is so wide as to verge on the nebulous.

Within theoretical biology, empirical exploration of theoretical issues via computer experiments is increasingly (if not wholly) accepted as a viable method of scientific inquiry. And several well-established journals provide a forum for publication of results in "biologically inspired'' engineering. Indeed MIT Press already publishes two journals (Adaptive Behaviour and Evolutionary Computation) which deal with issues apparently within the remit of Artificial Life, but which are more clearly focused. Given this, it seems hard to resist asking why someone would opt to publish in Artificial Life rather than in a more specific journal?

This should not be read as criticism. Certainly cross-disciplinary journals such as this play a valuable role in bringing together research from otherwise disparate fields. The volumes of proceedings from the four international workshops on artificial life (three of which were edited or co-edited by Langton) have indicated that there are strong grounds for enthusiasm. The review articles presented in the first three issues show the possibilities for further work in a variety of directions. As an overview of the state of the art, the first three issues do a very good job of demonstrating both the potential and the problems in artificial life. The editor-in-chief, and the 39 associates, advisors, and editorial board members account for most of the influential researchers in this field. All involved deserve credit for attempting the difficult task of producing a truly interdisciplinary journal. What remains to be seen is the quality and scope of the "regular'' submissions, because ultimately it is on the basis of these that the journal should be judged.

Dave Cliff is a lecturer in computer science and artificial intelligence, School of Cognitive and Computing Sciences, University of Sussex.