SITUACIÓN ACTUAL DE LAS ESTACIONES Y LOS SISTEMAS DE

The availability of a fundamental and uncontroversial description in terms of occupied and unoccupied cells and relationships between such cells in a single generation or over a series of generations cannot settle the taxonomic difficulties which arise in Life even though it can help in some cases. Spaceships are not picked out by reference to their microscopic properties, and the variety of spaceships suggests that any catalogue of microscopic properties of spaceships would be at least significantly disjunctive, and in any event parasitic for its construction on the broadly functional account or theory concerning what spaceships actually do. While it might be possible to lay down arbitrary criteria distinguishing spaceships from puffers it is far from clear that would give any benefit, since these classifications arose partly in the context of different streams of research into cellular automata including efforts to simulate computational processes, memory, communication and also to explore ‘life-physics’ (For example by trying to make the fastest possible space-ships, which it has been shown would travel at c/2.) and special evolutionary problems concerning certain structures.59 Since the terms need to be able to do the work required of them for each of these research programmes, and those programmes have as their objects, by and large, the macroscopic features of certain life-structures, such an imposed taxonomic order would tend to be counterproductive.

A symptom of this is the fact that the typical tools of the trade in Life research work with functional and macroscopic characteristics - search algorithms which scout large numbers of configurations for ones which fall into classes III or IV, further tools which generate and filter large numbers of ‘mutations’ of configurations already identified as interesting, etc.60

Any ‘cellular epistemologist’ is, by and large, not interested in the fundamental description despite its ready availability. In fact, a portion of the work on cellular automata goes on under the heading of ‘emergent’ computing (Forrest 1991, Langton 1991, Koza 1992) where, under some interpretation, information, structure and interactions are discernible at large scales or ‘high’ levels of the computing system, yet inaccessible (or in some cases at least prohibitively unwieldy) at the level of the basic constituents of the system.61 The point here is that even though it might be possible, in some cases, for initiates to argue over whether some configuration

59

For example the conjecture that any life-structure can be the result of a set of interactions between gliders. This conjecture has not been proven or disproven, although a number of structures have been successfully “reduced” to being the consequence of interactions between gliders.

60

See Koza (1992) for a discussion of genetic programming in a cellular automata-like simulation of the behaviour of a virtual ant colony.

of cells in Conway’s Life is a ‘spaceship’ or a ‘puffer train’, just as biologists can dispute whether a given creature is a member or one species or another, one thing which is not in doubt in the cellular automata case is that there is a fundamental description to which any other description will stand in at least a token-token relationship.

In a hypothetical case where the order of epistemological business is reversed, and our cellular epistemologist, or better a community of them, are given the surface behaviour of some automaton but not the rules, then the discovery of the rules would, of course, be very significant indeed. It is plausible to imagine that in the absence of knowledge of the underlying rules, any of a number of non-taxonomic controversies might rage over whether downwards causation was operating, whether a fundamental description was possible, whether the plausibility of such a description was affected by the taxonomic debates, and so forth. Even though the discovery of the rules would, hopefully, close such debates, it would not and could not answer or adjudicate on the taxonomic difficulties for the reasons already given.

Those working in these fields are entirely aware of this, and this recognition is a part of what is at stake in the use of the term emergent in some of the literature pertaining to cellular automata62_{. Forrest (1991:1), for} example, refers to emergent computing as having as its object ‘computational models in which the behaviour of the entire system is in some sense more than the sum of its parts’ adding that in these systems ‘interesting global behaviour emerges from many local interactions.’63 Nonetheless there is no suggestion of emergence in the sense involving downwards causation, hence Forrest notes that ‘the concept of emergent computation cannot contribute magical computational properties’ (1991:3). Rather the point is that under some interpretation cellular automata contain and process information not discernible at fundamental levels. Even if one accepts that in principle ‘emergent computations can be simulated by a Turing machine, interpreting the resulting patterns as computations is likely to be so difficult as to be infeasible’ (Forrest 1991: 3).

This means that despite the noted availability of fundamental descriptions for all cellular automata, it does not follow that such descriptions will be the most informative or useful, or that they will be able to adjudicate taxonomic or other disputes. In a sense this argument is a complement to that developed in Fodor (1974), to which I return below in specific connection with Dupré.

So, in conclusion, although it was noted earlier on that if the Completeness Thesis was false, then science could not possibly be unified in the ways Oppenheim and Putnam suggest, it does not follow that

62

‘Emergent’ computing is not restricted to cellular automata, and includes work on neural networks and other technologies.

63 _{Forrest refers to Hofstadter’s requirements for emergent computation, and notes that he ‘stresses that} information which is absent at lower levels can exist at the level of collective activities’ (1991: 2). Forrest also notes that an explicit practical advantage of emergent computations over traditional methods of programming using explicit instructions are that emergent computations are naturally more flexible, since their flexibility is not a function of any set of explicit instructions (1991: 3).

difficulties with unification pose any problem at all for completeness. The above argument does not, of course, establish that the visible variety in the world at non-microscopic levels is in fact the result of simple interactions between simple elements. What it does show, rather, is that no direct inference from macroscopic variety and even heterogeneity to the impossibility of a simple and essentially uniform fundamental level is sustainable. How the world is in fact remains an empirical question, and in due course I will consider the relative strength on empirical grounds of the view that physics is complete.

The present stand off should not be surprising, unless we expect to be able to extract metaphysical conclusions from analysis of epistemological practices and premises. Davies (1996: 5, 8-9) makes the point that such arguments cannot in general be decisive, and that they specifically fail to be decisive in Dupré’s case. It is an interesting question where the burden of proof might reasonably be expected to lie in this case. For most physicalists it seems clearly to lie with the defender of disorder, but physicalists by and large are people who accept the Completeness Thesis anyway. There is, though, a case to be made that considerations of minimal mutilation of our normal ways of dividing up and understanding the world place the burden squarely with the would be defender of the Completeness Thesis. For my own purposes the burden clearly lies in this side, for the simple reason that I am engaged in an attempt to see how well the Thesis can be defended against various lines of criticism. In due course I will argue that the correct way out of the present stand off is by means of a defence of fundamental physical laws against the criticisms of Cartwright. Before turning to that task, though, I need to complete the treatment of Dupré by dealing in more detail with the question of reductionism.