Biological Evolution and Information Acquisition

SubscribeSign in

Biological Evolution and Information Acquisition<br>Brian Potter<br>Jun 11, 2026

68

Share

A few weeks ago we looked at a simulation of technological evolution by economist Brian Arthur, in which he was able to start with simple building blocks (such as a NAND gate) and evolve surprisingly complex circuits (such as a 12-way AND gate or a 4-bit adder) by randomly combining increasingly useful existing components. We analyzed this as a way of simplifying a search problem: by using existing, working components as modules that can be combined, a few at a time, into more complex modules, and then combining those into even more complex modules, many unpromising and time-consuming branches of the search tree are screened off, and the simulation can find useful technologies amidst an enormous branching set of possibilities.<br>Real human technology is, of course, not generated by randomly combining components together and seeing if they do anything useful; the randomness in these simulations is just a way to see how easy or hard it is to create new technologies under different conditions. But biological technology — the huge panoply of lifeforms that exist on earth, from microscopic single-celled organisms to whales the size of a 737 — is also generated by randomness. Evolution builds biological technology bit by bit by harvesting the fruits of genetic variation, often caused by random mutation, preferentially selecting the most fit organisms to propagate their genes into the future. Over billions of years, this process can generate astoundingly complex biological systems.<br>What’s interesting is that biological evolution uses a very similar trick to Arthur’s circuit simulation. By leveraging modularity at the genetic level, populations of organisms can increase the rate that useful genetic variants spread through the population, effectively increasing their rate of information acquisition. Sexual reproduction, along with other ways of sharing genetic material like horizontal gene transfer, is essentially a mechanism for doing this. We can show this with some simple simulations.<br>Evolution and reproductive strategies

The simplest way for an organism to reproduce is asexual reproduction, where a parent produces a child that’s a genetic copy of itself. Simple single-celled organisms, for instance, reproduce by cellular fission, dividing into two or more “children” that each have the same genes as the original parent.<br>But children won’t necessarily be identical copies of their parents. Due to genetic mutation, some genes might get randomly altered during the fission process, producing children with slightly different genes. In some cases, these mutations might be useful, giving additional functionality such as antibiotic resistance and thus better odds of surviving and reproducing. Because of their contribution to the organism’s fitness, over time the useful mutations will become more and more common in the population.<br>We can demonstrate this with a simple simulation. In our simulation, we start with a population of 100 creatures, each of which has a genome of 200 individual genes. A gene can either be a 1 (the “good” version of the gene) or a 0 (the “bad” version of the gene). The initial population is random, with each creature having roughly a 50-50 mix of good and bad genes. Each iteration of the simulation, each creature produces two children. A child copies the genes of its parent, but due to mutation each gene has a 0.2% chance of being flipped, going from a 1 to a 0 or vice versa. The 100 most fit children (where fitness is just the sum of each gene value, since 1 is the “good” version of the gene in our simplified model) are selected to continue the next generation, and the cycle repeats. This is a simplification compared to how evolution actually functions — for one, it treats genes as contributing to fitness independently, ignoring the fact that the fitness value of one gene often depend on other genes — but it’s enough to show some of the dynamics at work.<br>When we run this simulation, the proportion of “good” genes in the population steadily rises over time as more-fit offspring outcompete less-fit offspring. Depending on the mutation rate, the population may eventually reach maximum possible fitness of 200, or plateau at some level below it.

The problem with this strategy — producing children that are noisy copies of a single parent, and relying purely on random mutation as a source of genetic variation — is that once you’re at above-average fitness, mutations are likely to be bad on average. If a genome has more 1s than 0s, a random change will be more likely to change a 1 to a 0 than a 0 to a 1. Thus for parents of above-average fitness, their children will on average have lower fitness.<br>Because mutation is random, there will nonetheless be variation, and some children will end up with higher fitness than their parents. And because selection...