Macroevolution = Machines Creating Other Substantially Different Machines

By definition, biological macroevolution is the hypothetical emergence of higher taxa (new classes, phyla, kingdoms) from the ones that exist at any given point in time, by itself, i.e. without recourse to intelligent agency. Many authors maintain that combinations of law-like necessary factors (such as natural selection) and stochastic factors (mutations, genetic drift and recombination in the case of sexual reproduction) are causally sufficient to explain hypothetical macroevolution.

Note that while microevolutionary changes within lower taxa such as species or families can be observed and are well studied, emergence of new higher taxa cannot be directly observed and is hypothetical. From the cybernetics perspective, the hypothesis of macroevolution is equivalent to the generation of large enough amounts of functional information [Durston et al 2007]. Some authors believe that macroevolution is supported by the fossil record. While this question is outside of the scope of this note, it is worth noting that the fossil record is far from complete and is contradictory and it consequently cannot serve a solid case for the macroevolutionary hypothesis.

As far as the case in point, I believe that the capability of machines to create functionally different machines stands or falls with the macroevolutionary hypothesis and vice versa, because these two questions are equivalent. Indeed, certain aspects of the functioning of living organisms, according to John von Neumann, can be modeled as cellular automata which are analogous to computers. Therefore biosystems can be called biomachines with a substantial part of their functionality realised at nano-scale. Consequently, biomachines automatically creating other biomachines different from themselves by added functional complexity as part of macroevolution could be evidence of machines capable of creating substantially different other machines.

However, I think that macroevolution is not practically feasible for a number of reasons:

1. Every non-trivial function must be represented by already existing genetic and epigenetic code even before natural selection kicks in. Stochastic factors and law-like necessity are causally insufficient for the emergence of genetic/epigenetic code [Abel 2011], which is a template for the structure of the future organism and is an equivalent of functionally sufficient and complete sets of coherent instructions. In practice, code can only emerge as a result of decision making on the part of an agent (even though this code can be left alone afterwards under the action of stochastic factors and selection). At least, no means of generating enough functional information in the form of instructions other than intelligence is known today (for concrete metrics of how much is enough to exclude chance and law-like factors, see [Durston et al 2007]).
2. Most liberal upper bounds on terrestrial probabilistic resources measured based on the maximum possible number of quickest physico-chemical interactions over the entire life of the Earth, appear to be too low for the practical realisation of unguided search for stable protein structure supporting novel biofunction (see  here and here).

On the other hand, I believe that machines are incapable of creating other, substantially different machines, without intelligent agency. I think that this impossibility stems from undecidability of the halting problem. Note that machine replication is a different problem and is actually feasible (consider e.g. computer viruses). Also, the case of man creating machines does not apply because we are intelligent beings and, consequently, even though we have the biomachine aspect about ourselves, our intelligence serves as an oracle in terms of the Turing halting problem. This observation agrees with the main ID hypothesis which states that in practice only intelligence can generate large enough quantities of functional information.

References

1. Abel, D. L., The first gene, 2011.
2. Durston, K. K., Chiu, D. K., Abel, D. L. and Trevors, J. T., Measuring the functional sequence complexity of proteins, Theor Biol Med Model, 2007, 4: 47 Free on-line access at http://www.tbiomed.com/content/4/1/47.
3. Kirk Durston, A Scientific Case for Intelligent Design.
4. Kirk Durston, A Common Either-Or mistake both Darwinists and ID-theorists make, uncommondescent.com.