Kinematic Self-Replicating Machines
© 2004 Robert A. Freitas Jr. and Ralph C. Merkle. All Rights Reserved.
Robert A. Freitas Jr., Ralph C. Merkle, Kinematic Self-Replicating Machines, Landes Bioscience, Georgetown, TX, 2004.
2.1.5 Design for Nonevolvability
From the perhaps limited perspective of our desire to build “inherently safe” (Section 5.11) kinematic machine replicators that are incapable of evolving out of our control, Taylor  points out that von Neumann’s work on self-replication concerned the question of how machines might be able to evolve increased complication in order to perform increasingly complex tasks. This is why von Neumann’s design for a self-replicating machine had to be capable of universal construction, and why it could withstand some kinds of mutations. Von Neumann’s architecture was designed specifically to allow for a possible increase in complexity (Section 2.1.1) and efficiency of machines by evolution , and it is quite clear that he was primarily interested in self-reproduction that could lead to open-ended evolution . Noted von Neumann : “Anyone who looks at living organisms knows perfectly well that they can produce other organisms like themselves…. Furthermore, it’s equally evident that what goes on is actually one degree better than self-reproduction, for organisms appear to have gotten more elaborate in the course of time. Today’s organisms are phylogenetically descended from others which were vastly simpler than they are, so much simpler, in fact, that it’s inconceivable how any kind of description of the later, complex organism could have existed in the earlier one.”
As to the nature of the information encoded on a self-description tape, Taylor  observes that von Neumann suggested “it is better not to use a description of the pieces and how they fit together, but rather a description of the consecutive steps to be used in building the automaton” (von Neumann , p. 486). In other words, to optimize system auto-evolvability the information should be in the form of a developmental recipe rather than a blueprint (Section 5.1.9 (B5)). (The obvious corollary is that using blueprints rather than recipes can help to minimize auto-evolvability, which may be desirable in designing for public safety; Section 5.11.) The evolutionary advantages of this kind of genetic description have also been discussed by biologists such as Dawkins  (pp. 177-178, pp. 250-264) and Smith  (pp. 21-23), who note that from an evolutionary viewpoint, one of the most important features of the developmental approach is that it allows mutations on the genotype to have a wide range of magnitudes of phenotypic effect. For instance, mutations affecting the early developmental process can potentially lead to gross macroscopic changes in phenotype, whereas those affecting later stages can have the effect of “fine-tuning” particular structures .
Thus it seems axiomatic that desirable artificial nanomachines which are capable only of “inherently safe” replication without the possibility of surviving mutation or of undergoing evolution should have a simpler, less-capable architecture than those proposed by von Neumann and by others who are interested primarily in modeling living processes which need the more-capable architectures. That is, it should be possible to abandon complete logical or constructional “universality” and still retain the ability to self-replicate, but without the ability to self-evolve (Section 5.1.9 (L)). Drexler  has also noted the brittle organization of mechanical-style replicating systems as compared to organic-style replicating systems, and concludes: “It seems that building a self-replicating molecular system based on nanomachinery does not entail building a system capable of evolution. Indeed, it seems that the latter would be a distinct and challenging goal.”
Last updated on 1 August 2005