The Selfish Universe
or, who the genes are really working for
Here is a piece of biology most people have never been told. More than half of your DNA is parasitic.
It is not your DNA in any meaningful sense. It is sequences that exist only to copy themselves — transposable elements, retroviruses inserted into the genome a hundred million years ago, ancient molecular machinery that has nothing to do with making a human, only with making more of itself inside one. They sit inside your chromosomes. Your cellular machinery dutifully transcribes and replicates them every time a cell divides. They contribute nothing to your survival. In some cases they contribute actively to your disease. They are along for the ride, and the ride is you.
The fact that this is true — that the genome of every animal on Earth is something like a lightly-organized infestation — was not properly appreciated until well into the twentieth century. The genome had been imagined as a tidy blueprint for the organism. It turned out to be a jumble in which the organism’s genuine instructions are interspersed with vast quantities of self-replicating noise. Once you noticed it, the question was no longer "what is the genome for?" The question was: for whom?
The answer arrived in 1976 from a thirty-five-year-old Oxford zoologist with a paperback that fit in a coat pocket.
In The Selfish Gene, Richard Dawkins inverted the usual frame of evolutionary biology. The unit of selection, he argued, is not the organism. It is the gene. Organisms are temporary; they live a few decades at most and then dissolve into the dirt. Genes — particular sequences of DNA — persist for millions of years, copied from generation to generation, recombined, occasionally mutated, often unchanged. The organism, on the gene’s timescale, is a vehicle. An evolved-up apparatus the gene rides around in for a while, builds copies inside, and then discards.
From the gene’s point of view, the parasitic DNA in your chromosomes is not a bug. It is a feature. Genes that figured out how to get themselves replicated without contributing anything to the host are doing exactly what every other gene is doing — getting copied — just without the diplomatic gesture of helping the host first. They are the limiting case. Most other genes pay rent (they help build hearts and brains and immune systems and so on). The parasitic ones don’t. The mechanism is the same; only the contract is different.
This was a paradigm shift wrapped in a cheap paperback, and once you saw it you could not un-see it. Every biological feature you had ever wondered about — drone bees that die after one mating, lemmings that supposedly throw themselves off cliffs (a myth, but a myth that required a gene’s-eye refutation to lay properly to rest), male seahorses that incubate the young, peacocks dragging six-pound tails through tiger country — all of it became sharper when you stopped asking what was good for the animal and started asking what was good for the gene. The animal is what the gene built so that the gene could persist. The peacock’s tail is for the gene that codes for the tail. The drone bee’s death is irrelevant to the gene that got copied during the one mating that preceded it. (The peacock case has its own dynamics worth a separate treatment; see Sexual Selection for the full mechanism.)
The trick of the book is that selfish is meant in a strict, unsentimental sense. Genes do not have desires. They are short stretches of nucleotide. The phrase means only that natural selection works on them as if they were optimizing for their own propagation, because the ones that did propagate are the ones that are still here. Dawkins spent the rest of his career patiently explaining this to readers who had read the title and assumed he was endorsing greed. He was not. He was naming a mechanism.
And the mechanism turns out to apply to far more than DNA.
The last chapter of The Selfish Gene introduced a word that has since done extraordinary work: meme. A meme is a piece of culture — a tune, a phrase, a hairstyle, a religion, a recipe, a programming language, a particular way of holding a coffee cup — that propagates from mind to mind. Memes vary, replicate imperfectly, and undergo selection against the limited attention budget of human brains. They evolve. The ones that survive are not necessarily the ones that benefit their hosts; they are the ones that are good at being remembered and retold. Happy Birthday is not a particularly good song. It is, however, an excellent meme.
The deeper claim — the one Dawkins gestures at and that other thinkers (Susan Blackmore, Daniel Dennett, John Maynard Smith) have run with — is that evolution is not a fact about biology. It is a fact about any system that has three ingredients:
- Replication — things produce copies of themselves.
- Variation — the copies are not perfect.
- Selection — some variants persist or replicate more readily than others.
Anywhere those three are present, evolution happens. Not as metaphor. As mechanism. Genes have all three. Memes have all three. A market economy has all three (firms replicate their strategies into market niches, vary by experiment, get selected against by profit and against by bankruptcy). Software has all three (libraries are forked, modified, kept or abandoned). Spoken language has all three (children copy what they hear, imperfectly, and the more memorable forms persist). Your immune system has all three running every day inside your bone marrow, evolving antibody proteins on a one-week timescale to match the pathogens it has seen.
And this is where it gets strange. Patterns themselves have all three.
The attractor of a chaotic weather system is replicating itself in a sense: the trajectory keeps coming back to the same region of state space, day after day, because the region is mathematically stable even though no specific point inside it is. A flocking pattern in a starling murmuration is replicating itself: it propagates from one moment to the next because it is an attractor in the dynamics, and small perturbations are pulled back into it. A synchronized rhythm in a population of fireflies is replicating itself: once the population locks into a rhythm, the rhythm sustains the rhythm. None of these has DNA. None has neurons. But each is a thing that persists by reproducing the conditions of its own persistence, which is functionally what a gene does. They are selfish in the same precise, unsentimental sense Dawkins meant.
This is, I think, why the universe looks the way it does. The patterns that exist now are the ones that turned out to be good at persisting. The ones that were not are gone. We do not see the unstable orbits, the disordered flocks, the dissolved sync, the molecules that were almost but not quite self-catalyzing. They were there for a moment and then they were not. The universe is not a parade of all possible patterns; it is a graveyard with a few survivors still moving. The survivors are selfish, in the strict technical sense .. they are the ones that managed to keep being themselves.
This is also, I would argue, the only way the universe could possibly look. Whatever physical laws had been in place at the beginning, the patterns we would see at this moment in time would be the ones that were selected for, because any system run for long enough surfaces its attractors. Different laws would yield different attractors. The flavor of the survivors would change. But the fact of survivors, of repeated patterns, of the universe being populated by exactly the things that are good at being themselves — that fact would not change.
There is a hypothesis at the cosmological end of this argument, due to the physicist Lee Smolin, which I will only mention here because it deserves a page of its own. Smolin proposes that black holes give birth to new universes with slightly different fundamental constants. Universes that produce more black holes — which happen, suspiciously, to be the same universes whose physics support stable matter, complex chemistry, and life — have more “offspring.” On this view our universe is the result of selection on the cosmos itself. We live in the kind of universe that is good at making more universes. The fine-tuning argument inverts. We are not lucky. We are the inheritors of something old.
I will write that page another day.
The Experiment
What follows is a small program that Richard Dawkins wrote in BASIC in 1986, on a Macintosh, after a long flight to Australia. He called the creatures it produces biomorphs. Each biomorph is a tree-like shape generated from nine integers — nine “genes” — that determine how a recursive branching procedure unfolds. Eight genes control the angles at successive depths. The ninth controls how deep the recursion goes. Mutations are simply: pick a gene, change it by one.
You will be the selection pressure. You see a parent biomorph at the top, and below it nine offspring, each of which differs from the parent by a single mutation in a single gene. Click the offspring you find most interesting, and it becomes the new parent. Its nine offspring are then mutations of it. And so on.
What Dawkins discovered, and what surprised him, is how quickly random + select reaches strikingly specific shapes. Within thirty generations you can breed a tree, a face, a letter of the alphabet, a recognizable insect. Within a hundred you can breed shapes that are clearly what you intended, even though you started from random noise. You did not design the result. The mutation operator did not design it. Selection did, by repeatedly choosing — out of all possible variations — the ones you happened to like.
Things to try:
Just click around for a few minutes. Notice how unrecognizable shapes resolve into something organic-looking after only a dozen or so generations. Branches turn into limbs; limbs turn into tendrils; tendrils turn into things that look strangely intentional.
Pick a specific target — a tree, a stick figure, a letter, an insect — and try to breed toward it. You will sometimes arrive in twenty generations. You will sometimes spend forty and end up nowhere near. Selection on a fitness landscape with local maxima is hard, even when the only fitness function is your taste.
Hit Random restart to begin from a new starting biomorph. Try to evolve toward the same target you just got. Notice that two runs almost never converge on the same lineage even when you steer toward similar shapes. The space of possible biomorphs is vast, and the path you take through it depends on every single one of your choices.
Hit Undo if you take a step you regret. Your selection is the only thing creating direction; the variation has no preferences of its own.
Watch the gene values printed below the parent. Each gene is a small lever in a much larger configuration space. A single integer changing produces visibly different offspring. Notice that some genes seem to matter more than others — depth changes the entire structure; angle changes only modify it. There is no a priori way to know which gene controls what; you discover it by mutating.
The biomorph experiment was originally written to make one specific point: the appearance of design does not require a designer. Mutation is undirected. Selection is a filter, not an architect. And yet within minutes, the shapes look like things someone might have drawn on purpose. The watchmaker, in the title of Dawkins’s follow-up book, is blind. He cannot see what he is making. He is making it anyway.
Hold onto that. Most of what looks designed in the universe was made the same way.