next up previous contents
Next: Representation Up: Methodology and Design Issues Previous: Methodology

   
Self-Reproduction

Most of the work reported in this chapter has relied upon concepts of self-reproduction in one form or another. However, different studies have modelled self-reproduction in different ways and at different levels. Ray has insisted on the importance of self-reproduction in artificial life models of evolution (see Section 3.2.1), but von Neumann's analysis shows us that a variety of different issues are involved in this concept (e.g. the method by which reproduction is achieved, the robustness of the reproducers to mutations, the heritability of variations, etc.). There is a risk of confusion if one talks simply about self-reproduction, without referring to these more specific issues.

A great deal more work is required before we can properly understand the evolutionary consequences of design decisions relating to each of these more specific issues. Von Neumann's architecture was designed specifically so that the self-reproducing automata would have the potential of participating in an evolutionary process in which more and more complicated automata arose. I am not aware of such a detailed analysis of other sorts of reproduction, although it has been suggested that collectively-autocatalytic sets generally do not have great evolutionary potential (see, for example, [Maynard Smith & Szathmáry 95]). It has also been suggested that reproduction by self-inspection (discussed on p.[*]) has less evolutionary potential than genetic reproduction (e.g. [von Neumann 66] pp.121-123, [McMullin 92a] pp.191-193), although little has actually been proved.

For truly open-ended evolution, we might also want to consider how one sort of reproduction might evolve from another sort.3.29 I will return to discuss these issues in much more detail in Chapter 7, with the benefit of experience gained from running an extensive series of experiments with a Tierra-like platform (reported in Chapters 4-6). The main point I wish to make here is that the evolvability of the system clearly does depend upon the type of reproduction employed, and upon other issues involved in the concept of self-reproduction, even if we do not yet fully understand these dependencies.


next up previous contents
Next: Representation Up: Methodology and Design Issues Previous: Methodology
Tim Taylor
1999-05-29