I made the grave error of judging a book, perhaps not by its cover but definitely by its title, when I first heard about The Origin of Wealth by Eric Beinhocker. In my mind, the word “wealth” was most strongly associated with lists in Forbes, and a treatise on how people make it onto those didn’t sound interesting.

But after seeing it come up again and again, mostly in Zack Kanter’s tweets, I picked it up, and it turned out to be the most interesting thing I’ve read in a long time.I suspect there are people who have not and will not read this book for the same reasons that made me hesitate. This post is an attempt to summarize the concept I found the most fascinating, both to force myself to grasp it properly, and hopefully to persuade someone else to pick it up. This concept is:

This is the concept that I want to explain in as simple a way as possible, without simplifying. To understand it we’ll first need a quick crash course in biological evolution, which we then can tie to economic evolution.

- Interactors: the living version created from a design (or schema). A human like you, for example! ‍

- Schemas: a codified description (or recipe) of an interactor. For humans, this is DNA. ‍

- Schema readers: something that can read the schema and create the thing it describes. For humans, this is a fertilized egg [1]. ‍

- Fitness: how well an interactor fares in an environment. We’ll often talk about it as a "fitness score” which usually refers to how well an interactor survives and reproduces. It is worth noting that evolution does not let us predict how fit something will be in the future, only how fit something has been up until now.

With these definitions in hand, let's dive into the concept itself.

Biological evolutionary components

What is a constantly evolving, roughly-correlated, fitness landscape?

Fitness Landscape

A fitness landscape is a concept from biology. In biology, it is a way to visualize how well an interactor fares in a particular environment. It is usually portrayed as a 3D plane, which looks something like this. Imagine that each point on the X & Y axes is a specific genetic combination representing an interactor (human) who has been created from a schema (DNA) by a schema reader (fertilized egg). Point [1,1] is one genetic combination, point [1,2] is an ever so slightly different one, and the one at point [5,10] is more different still. The third dimension - the height Z - is the fitness score of the interactor.

A fitness landscape

For illustrative purposes we draw this as a 3D plane, but it is actually a multidimensional plane where the coordinates represent potential genetic combinations. It is important to understand that this plane can, and does, contain every single potential combination of genes. This means that the plane is mind-bogglingly large. T-Rex, Tom Hanks, and your childhood dog are all somewhere on this massive plane. We call this massive, but finite, space of potential genetic combinations the design space. (Technically this is true for a specific point in time, as combinations of genes grow larger over time, which means the plane is growing over time, but we don’t have to dig into that to understand this concept).

Roughly Correlated

Now that we have the concept of the 3D plane we might wonder what it is shaped like? If the plane had a steady and constant slope we would call that “perfectly correlated”. In order to calculate the height of any point, you would only need to know the height of its surrounding points. In contrast, we would describe the plane as “random” if each point was independent. You wouldn’t be any better equipped to calculate the height of a given point if you had information about its surroundings. The biological fitness landscape turns out to be somewhere between - what we call roughly correlated. This means that the height of any given point is to some extent a function of the points surrounding it, but there is a random component to it as well. The correlation isn’t perfect, it’s rough.

This makes intuitive sense. Imagine that we find the point representing you on this plane. Our species is at the top of the food chain, you’re alive, and you have enough brain capacity to read this summary: you’re likely extremely fit (evolutionarily speaking). If you had been created ever so slightly differently - say your genes were different such that your nose was a bit straighter - your fitness would be unaffected. The point in the plane representing that genetic combination right next to the point your current self occupies, and is of the same height. But some changes - say one that radically increases your brain’s ability to solve complex problems - are likely to improve your fitness quite a bit. Conversely, another small change might affect how your lungs work and prevent you from breathing, which would be represented by a sudden and steep drop to a fitness score of 0. The points representing that genetic combination would be higher or lower than your current one, and look like big hills or holes in the ground, respectively.

Putting that together we’re left with a landscape that looks like the Swiss alps, if they were made of Swiss cheese (as Beinhocker so vividly puts it). Any one area is quite flat, but there are many hills to climb and quite a few holes to fall down into.

Swiss Swiss Alps

Constantly Evolving

An important quality of the landscape is that it is always changing. An illustrative example in biology is that the world’s atmosphere contained only 1% oxygen in the Proterozoic era, but contains 21% today. Lungs that work great today - and therefore have a high fitness score - worked very poorly back then, and vice versa. This change is largely due to interactors shaping the environment, which in turn shaped how the next generation evolves. We now have an idea of an evolving, roughly-correlated, fitness landscape, where each point represents a specific interactor and its fitness.

What is an evolutionary algorithm and why is it good at exploring the landscape?

We’ve mentioned that the plane - or design space - contains every potential combination of genes. However, the space is incredibly larger. Larger than the universe. In fact, there are more potential combinations in a single gene than there are atoms in the universe [2]. Most of the designs in the design space are completely useless; they won’t even be chemically stable to exist for more than a split second. The few that are interesting are extremely rare. As Beinhocker puts it: you’re more likely to find a specific drop in all the oceans of the earth than you are to find a point that represents anything remotely interesting, let alone something as complicated as yourself.

But here we are, a mere 3.8 billion years after life emerged, and we have you, and me, and every single living thing. How did evolution find us?

A few more definitions of biological evolution are needed here, most of which you’re likely familiar with. As you remember, interactors (humans) are created when a schema (DNA) is read by a schema reader (fertilized egg). We didn’t yet cover how new schemas are created through a combination of crossover and mutation. Humans, for example, reproduce by taking part of the two parents' DNA and combining them in a new combination. This is called crossover. This makes evolution recursive, as each generation uses the preceding generation as input. In addition to crossover, the schema reader (the fertilized egg) can make “mistakes” when reading the schema, leading to small unexpected changes in the genes. This is a mutation. This allows new variations of interactors to be created, representing previously unvisited points on the plane.

If these were the only dynamics in play, evolution wouldn’t be that effective. Sure, it moves through the design space, but it sounds like it’s just gradually expanding from wherever it is. To explain its ability to rush through the space and find homo sapiens, hippopotamuses, and hummingbirds, all before the sun explodes, two more ingredients are needed: selection and amplification.

Selection is survival in the environment. You need to survive until you pass your genes on. If we make the small modification we mentioned above, leading to your lungs not working properly, you would not survive, and your genes would not be passed on. Evolution 101. Amplification is the concept that the higher the fitness score, the more that interactor will reproduce. There will be more variations, created through mutation and crossover, using a fit interactor’s schema as the starting point.

Combining these four dynamics - crossover, mutation, selection, and amplification - turns out to be an extremely effective way of searching the design space. Crossovers are small steps around the plane, while mutations allow for larger jumps. Selection and amplification act as resource-allocators, discarding bad experiments and doubling down on what works. These forces are in effect as we speak for each living organism, allowing evolution to run billions of experiments in parallel.

We take it for granted but the effectiveness of evolution really is difficult to overstate. A mere 3.8 billion years to get from single cell organisms to +30 trillion celled humans is an incredible achievement when you consider how many forks in the road evolution have taken to get all the way to you and me. Bright as you might be, evolution is clearly cleverer [3,4].

How does a market economy use an evolutionary algorithm?

That was quite the biological detour, let’s get back to business. It turns out that evolution isn’t limited to genes and biology. Genes are essentially just information, and evolution can be applied to other types of information. To sound extra smart you can say that evolution is substrate neutral. All you need are the evolutionary components we see in biological evolution. Beinhocker maps these to economics in this way:

- Interactors: a business (not to be confused with a firm, which might have one or several businesses).

- Schemas: a business plan. These consist of combinations of physical inventions (a computer), and/or social inventions (contract law).

- Schema readers: a management team.

- Fitness: how well a business does in the marketplace - its profit.

Biological and economic evolutionary components

All of the concepts we’ve described for biology also applies to businesses. There is a fitness landscape where each point is a business plan, rather than a DNA combination. Most small changes to a business plan have a marginal impact on fitness - but some will make the score much lower, and a few will make it much higher. This landscape is constantly evolving, changing what plans are fit. A human is very fit today, but would die immediately in the 1% oxygen environment of 2.5 billion years ago. The Kardashians are worth billions today, but in a world without Instagram and television their business plan would have performed very differently.

Once a business plan is implemented, the market is the environment that determines fitness. A biologically unfit organism will die without passing on its genes, as will an unprofitable business. A successful organism will reproduce and amplify, as will a profitable business. First, the profits it generates can be used directly to explore new business plans with its current schema as a starting point. Second, other businesses will copy the successful components, amplifying them even further.

One difference you might point out is selection: biological evolution does not rely on designers or creators to come up with new schemas. The economy does. While this is theoretically true, most business plans draw a lot of inspiration from existing ones. Selling goods online? Offering a subscription? Providing 24/7 customer support? All of these are components of existing, fit, business plans. Mimicking them is economic crossover - even if it is due to an intelligent decision. As a business plan is interpreted by a management team, mutations are given an opportunity to sneak in. Sometimes such mistakes lead to lower profits, but sometimes it can lead to breakthrough insights that were not present in the original plan [5].

Biological evolution was able to dive through a haystack many times larger than the size of the universe to find the needle that is humans. By combining these economic components in the modern market economy, we can use that same force to find better and better business plans to solve people’s problems and make their lives better.

Neat. Does it matter?

I sure think so! Perhaps the most interesting implications come from the ways in which economic evolution differs from biological.

One difference is the size of the jumps we can make on the plane. An important constraint in evolution is path dependence, that the path you’ve taken so far determines where you can go next. Genes can’t change all that much in one generation. Economic evolution isn’t an exception, but the restrictions are likely less significant. Biological evolution has come up with some incredibly fit designs: from the flying abilities of an eagle to the brain capacity of humans. But now that it has those two designs, they can’t merge. Biological changes are smaller, and each generation has to be fit enough to pass genes on. Fully functional wings might on a human might be a great point to end up at, but because we need to muddle through generations of half-baked non-functioning wings to get there, we might never make that journey. In the economy, this looks quite different. Two ideas as far away from each other as “wings” and “big brains” in biology can merge in the head of a creative entrepreneur and become reality without any intermediate steps, as long as she’s only relying on existing inventions (again, the Kardashian business plan doesn’t do as well without social media).

A second difference is the impact of interactors on the environment, where the difference between the two is mostly the speed. While it’s impressive that oxygen levels have increased 21x in the past 2.5 billion years, the economic environment went from an “economy” consisting of a few variations of hand-made axes traded through barter to billions of unique products that arrive on your doorstep simply by typing 16 digits into a phone, all in a mere couple of millennia. Each step of the way, new social and physical inventions - from the rule of law to the creation of the microwave - shaped the environment. New business plans are created on top of those inventions, in an ever repeating cycle. Once again, this same dynamic is present in biology, but the speed at which it’s allowed to run its course in the economy is why you’re reading this on a phone instead of a clay tablet.

Through this lens, I’ve put a higher value on people attempting to create new things, efforts aimed at helping more people get off the ground, and inventions of new products, services, or ways of organizing. Each and every one of us can be an explorer in the fitness landscape, searching for a new peak. Every once in a while, an effort will hit the environment like an earthquake, enabling new peaks of fitness to emerge. Movements like no-code, services that help individuals improve their own skills, and tools to run your own business are all even more interesting against this backdrop. The design space is humongous, and we’ve only just scratched the surface, so let’s keep exploring!

Want more?

First off, buy the book. If you want to keep digging, the rabbit hole is pretty deep and I've only poked at it a bit myself. A few suggestions:

- Podcast: Complexity. Published by The Santa Fe Institute, which probably deserves more credit than any other single organization for supporting research in this area.

- Book: Complexity: The Emerging Science at the Edge of Order and Chaos. Written by Mitchell Waldrop, another interesting book exploring adjacent concepts.

- Tool: HASH. An open-source tool to build agent-based simulations, a technique that is featured heavily in the book and the field of complexity at large.

Sources:

[1]: Fun fact: A 20 week year old female fetus (which is growing according to the schema) is already developing ovaries (schema readers of human DNA). Absolutely crazy.

[2]: Putting a firm number on the potential number of gene combinations isn’t quite that straight forward, but an estimate is 4 to the power of 300 combinations for one gene which humans have 25,000. This is compared to the 10 to the power of 80 atoms in the universe.

[3]: It’s hard to put a firm number on a count of human cells as well, but even at lower estimates evolution’s effectiveness is quite clear. https://www.tandfonline.com/doi/abs/10.3109/03014460.2013.807878

[4]: This is a paraphrase of Orgel’s Second Rule: https://en.wikipedia.org/wiki/Orgel%27s_rules

[5]: Post-its are a famous example https://en.wikipedia.org/wiki/Post-it_Note