The word `artificial' is commonly used in two distinct ways. One carries the connotation that the thing being referred to is a fake--an imitation of something else (e.g. `artificial flowers'), but the other merely suggests that the process by which the referent was produced is man-made or unnatural in some sense, but the thing itself is just as real as the original (e.g. `artificial light').
These two uses of the term correspond to the contrasting philosophical positions of Weak and Strong Artificial Life respectively. Proponents of Weak Artificial Life view their work as an attempt to learn more about biological life (or to create artefacts such as robots) by creating synthetic models of various processes associated with living organisms (e.g. evolution) on computers and other artificial media, but do not claim that any parts of their models are actually living themselves. On the other hand, proponents of Strong Artificial Life claim that by instantiating such processes in artificial media, the end product will be just as deserving of the term `living' as are biological organisms. In other words, the former group see their computer programs as simulations of life, while the latter group see them (potentially) as realisations of life [Pattee 88].
This distinction is related in some ways to the argument between those who claim that life depends crucially upon the physical medium in which it is implemented, and those who claim that life is fundamentally a process (or a collection of processes) and as such can be implemented in any physical medium which has the logical structure to support such processes. Arguments for the former view have been put forward by, for example, Elliott Sober [Sober 91], Claus Emmeche [Emmeche 92], Eric Olson [Olson 97], and Margaret Boden [Boden 99] (although she claims that this depends on how we define metabolism). Proponents of the latter point of view, that life is independent of matter, include Langton himself [Langton 86] and Tom Ray [Ray 91], creator of the `Tierra' system (described in Section 3.2.1). A version of the former view which offers more hope for the endeavour of Artificial Life, and the view with which I have most sympathy, is that life must be embedded within a `physical' environment, but that the symbolic environment provided by a computer program (that is, an `artificial physics') might be sufficient for this purpose (e.g. [Pattee 88], [Rasmussen 91], [Pattee 95a], [Bedau 98b]). However, this view comes with the proviso that the artificial physics of the computer model must resemble the important aspects of the physics of the real world if the model is to be treated as science rather than just a computer game (see, for example, [Moreno et al. 94], [Morán et al. 97] and [Ruiz-Mirazo et al. 98]). A major challenge for the science of artificial life will be to develop theories of exactly which aspects of the real world are important in this sense.
A glance at any of the recent proceedings of conferences on artificial life (for example, [Adami et al. 98], [Husbands & Harvey 97]) reveals that the field now encompasses a very wide variety of research, including synthetic evolutionary models, artificial chemistries, models of autopoiesis, self-organising systems, evolutionary robotics and evolutionary hardware. The methods of artificial life are now also being employed for artistic purposes,3.4 and work of this nature is also being integrated into these conferences. The majority of current work in the field (at least as published in scientific journals and conferences) does not postulate any more than the weak position. In other words, the premise is that by modelling particular natural processes such as evolution, we can improve our understanding of those processes and their relevance to the biological world, or, alternatively, we can profitably use such processes for various practical purposes (e.g. to develop robust control systems for robots).
However, a minority of published artificial life research explicitly takes the strong position. The protagonists, not surprisingly, include those who are most vociferous about life being independent of matter, such as Langton and Ray. For example, Ray says ``I would consider a system to be living if it is self-replicating, and capable of open-ended evolution'' (ibid. p.372).3.5 Ray's requirement that living systems have the capacity of open-ended evolution is similar to Bedau's concept of life as supple adaptation, discussed in Section 2.1.1. However, the definition does not prescribe how such a capacity can be assured. In the following chapters, much time will be spent investigating the behaviour of an artificial life platform called `Cosmos', based upon Tierra, in which populations of self-replicating computer programs evolve. I have developed this system largely to study its evolutionary behaviour, but I certainly do not equate self-replication with life. However, I think the relationship between open-ended evolution and life is more interesting. After reporting the lessons learned from running an extensive series of experiments with Cosmos (in Chapters 5-6), I will return to a further discussion of this relationship in Chapter 7.