Преимущества, предоставляемые использованием неймановского пассивного описания клеточного автомата. Интересно, что автор утверждает, что Нейман нигде не доказывает исключительную важность описания при воспроизведении; Нейман лишь продемонстрировал, что копирование системы с использованием её описания имеет вот эти самые преимущества, о которых идет речь.
Crucial for the remainder of our arguments is that there is a duality between the effective component, which in biological terms could be called the “soma” and the quiescent description, called the “germ-line.” In this duality the germ-line plays a dual role: both as a source of programmatic information and as a passive source of data for copying. The description represents, and instructs the creation of, the machines that copy and express it. I assert that living information flows are almost always representational (if not self-representational) and therefore always consist of a duality between a mediated signal and a machine that must transduce the information in that signal.
To my knowledge, the aforementioned Scientific American article to the contrary, neither Von Neumann nor anyone else ever proved that this duality was necessary for self-reproduction, that is whether genetic descriptions are a requirement for self-reproduction. An organism could also use itself directly as the source of information necessary to construct itself. Yet Von Neumann and others make some compelling arguments about the advantages of genetic descriptions for self-reproduction and other powers:
Homogeneity of components: the information has its basis in a small number of like components. Thus the problem of replication (the reproduction of information) can be simplified tremendously.
Defined state: the whole concept of an active, dynamic entity directly using itself as a source of information for reproduction is complicated by the fact that the stimulation associated with observation and necessary for copying may change the state of other components of the automaton. At best this would reduce the fidelity of self-reproduction and at worst could lead to undefined conditions. This shows clearly why Von Neumann emphasized “quiescence” as an aspect of the description.
Consistency: Von Neumann constructs self-reproduction in the stronger context of universal construction. He suggests that it is unlikely that a machine with the power of universal construction can be programmed to obtain descriptions of itself directly from itself without entering logical fallacy.
Completeness: By containing an independently copied representation of itself, the automaton avoids representing its own representation ad infinitum to maintain this representation for its offspring.
Generalizability: Von Neumann constructs his scheme for self-reproduction in a context that is immediately generalizable to the reproduction of a more general automaton, itself plus a generalized component, and this is demonstrated through the use of the description. In this way Von Neumann argued that quiescent descriptions allow self-reproducing automata to construct automata more complex than themselves, something machines generally cannot do.
Evolvability: Variations in quiescent descriptions can occur without altering the integrity of the automaton itself, facilitating the generation of variation to be acted on by natural selection.
These properties are strikingly reminiscent of design principles of programs and programming languages.