6.12 Evolution as a creative constructor

The cellular-automaton lineage was partly motivated by biology. Von Neumann wanted to understand how a machine could construct a copy of itself while retaining the capacity for heritable change. Evolution supplies the natural example on an extraordinary scale: finite inherited descriptions participate in the development of organisms containing nested molecular circuits, organelles, cells, tissues, and organs, while the reproductive system transmits descriptions from which altered successors can develop.

It is tempting to say that evolution designed this architecture. That language must remain metaphorical. Dawkins’s phrase “the blind watchmaker” captures the essential qualification: natural selection is cumulative but has no foresight, plan, or final target [ Dawkins , 1986 ] . Organisms are not steps along a monotone path toward perfection. They are historically contingent solutions to local problems of survival and reproduction under changing environments, complete with compromises, redundancies, and inherited constraints.

The comparison with von Neumann becomes sharper when the description is used in two modes. A self-reproducing automaton must interpret its description to construct a successor and also copy that description without interpreting it. Biological inheritance exhibits an analogous division: genomic sequences are expressed through cellular machinery and are also replicated into descendant cells and organisms. DNA is therefore not a freestanding blueprint for a body. Development also depends on pre-existing cellular organization, regulatory and epigenetic state, intercellular signaling, spatial context, stochastic events, and the environment. The finite string is central, but the constructor is the larger living system in which that string is interpreted.

Schematically, a genotype \(g_k\) participates in a developmental map

At population scale, selection changes which resulting descriptions and developmental systems contribute to the next generation. The recursively important point is that \(g_{k+1}\) may change not only the organism but also parts of the developmental, reproductive, and variation-generating machinery that constructs later organisms. Evolution can thereby alter its own effective space of future possibilities—what evolutionary biology calls the evolution of evolvability.

For synthetic creativity, the resulting artifact is not simply one successful organism. It is a persistent constructor–description system, embedded in a population and environment, that can generate and test descendants. Its compositional hierarchy may be written suggestively as

The arrows are not simple one-way functions: feedback, context, and many-to-many realization occur at every level. Their value is to mark a tower of effective theories. A cell type, tissue boundary, or organ is a stable macroscopic object whose laws are not transparently stated in the nucleotide alphabet, even though the genome participates in its construction.

Boden’s three modes can all be recognized in this process. Recombination is literally combinational; mutation and selection explore an inherited developmental space; and major changes to regulation, reproduction, or levels of biological organization transform the space in which later variation occurs. Evolution thus offers a model of creativity whose product is not only a theory of a world, but an embodied constructor capable of continuing the creative process.

This analogy also anticipates recursive self-improvement. The biological loop does not certify itself: differential survival and reproduction under a changing environment supply external selective pressure, not epistemic admission. The narrower lesson for a synthetic discovery system is that it should revise its theory, probes, and constructor only while retaining evidential channels that it cannot silently redefine.

Contemporary research programs explicitly pursue recursive improvement in automated research, system design, and AI hardware [ Recursive Superintelligence , 2026 , Ricursive Intelligence , 2026 , Sakana AI , 2026 , Chen et al. , 2026 ] . They are signals of an active engineering direction, not evidence that an autonomous system can already construct and independently admit a new scientific theory.

Wilson brings an evolutionary and humanistic lens to the origins of language, art, and creative expression [ Wilson , 2018 ] . His account broadens the question from biological construction to human symbolic culture, but it is philosophical and conjectural rather than a computational mechanism or an empirical validation of DIAL. It therefore marks a phenomenon to be explained, not an evidential premise of the framework developed here.