At the blog Construction Physics, which I also referred to here yesterday, author Brian Potter draws attention to research done by Brian Arthur demonstrating the evolution of complex circuits in a computer simulation. Arthur is an economist. In a paper, “The Evolution of Technology within a Simple Computer Model,” he models the development of technology. Nevertheless, he models it in a very evolutionary way.
His model begins with trivial components modeled on simple digital logic circuits. Arthur defines a list of “needs,” components that are considered useful. Different combinations of these components are tried at random until a useful component is found. Once a useful combination has been invented, it is added to the list of available parts that can be used to build further combinations. Through the random combining and recombining of parts, more sophisticated circuits are developed.
Evolving Technology
While Arthur models the development of technology, he refers to it as the “evolution” of technology. He chooses to model the development of technology as a process of randomly combining parts, a rather Darwinian, evolution-inspired approach. He describes it in a way that appears to grant agency to the circuits:
As the evolution proceeds more complicated circuits begin to construct themselves from simpler ones.
He draws a parallel between his model and biological evolution.
Our results mirror Lenski et al.’s, that complex features can be created in biological evolution only if simpler functions are first favored and act as stepping stones.
The reference to Lenski et al. is to the Nature paper “The Evolutionary Origin of Complex Features.”At the 2005 Dover trial, one of the authors, Robert Pennock, testified that the model from that paper, Avida, directly refuted the argument for intelligent design from irreducible complexity. I have previously published my own paper, “Digital Irreducible Complexity,” showing that this model and others did not in fact demonstrate the evolution of irreducible complexity.
Inventions by Chance?
An important point to be made is that technology does not, in fact, progress by randomly combining existing parts in the hopes of producing something useful. Inventions like airplanes, the printing press, the telephone, or the light bulb are not the result of numerous attempts at randomly combining components in the hopes of finding something useful. Rather, in each of these, intelligent agents, in this case humans, combined parts in deliberate fashion with the intention of producing a particular purpose.
The weaknesses of randomly combining parts are reflected in the kind of circuits that “evolve” in Arthur’s model. The circuits are either fairly simple, requiring only a small combination of parts, or highly repetitive, requiring the same combination repeated. The highly repetitive cases are able to “evolve” in his model, because each one is a relatively small addition on top of a slightly smaller version of the same. On the other hand, his model lacks more sophisticated circuits like multiplexers, comparators, encoders, decoders, population counters, or algorithmic logic units. This is a reflection of the fact that random combination of parts only gets you so far.
More Complex Versions
It is also noteworthy that, as Arthur points out, in order to “evolve” complex functions, his model must first favor simpler functions. In this case, Arthur was able to evolve more complex versions of a function by first rewarding evolving less complex ones as “stepping stones.” This certainly does work, but the crucial question is why we would expect such stepping stones to exist.
Arthur’s model seeks to demonstrate the process of “technological evolution.” But in so doing, it actually demonstrates the limitation of evolutionary thinking. It requires the introduction of “stepping stones” that reward simpler versions of the target functionality. Even so it is able to evolve circuits that are either simple or repetitive.