(It's just selection bias.)


[Note: I wrote this post for a friend who doesn't undertand the idea behind evolution (here). It was not originally meant for LessWrong in particular: frankly I'm just hoping you people can tear it to pieces so that I can better understand it myself.]

"A curious aspect of the theory of evolution is that everybody thinks he understands it."

Jacques Monod

A: There is no active force involved in evolution. No guiding hand that led scorpions to grow venomous tails, and wolves to grow sharp teeth. It sure looks that way—which is why we’re tempted to come up with complicated theories like “God did it”—but the truth is more counterintuitive. 

It’s selection bias.

B: That’s… not what evolution is.

A: Fine, let me explain. But try this out first: what happens when you taboo the word “evolution”? How do you explain life on Earth in its current form without using the magic word?

B: A process that favored the fittest species got us here. A fit gene is more likely to reproduce successfully before dying, so life naturally converges toward the fittest.

Dalle, just two suspicious cloaked figures taking notes about birds, as one does.

A: That’s the standard explanation. But it doesn’t put you in the right frame of mind for understanding evolution.

Calling it a “process” is already giving it a lot more credit than it deserves. An intelligent mind can devise a “process”: you plug your utility function into the algorithm, and out come the results. It’s like putting bread into a toaster, and toast coming out. But evolution is not like this. There is no utility function to evolution. Calling the whole thing “evolution” is treating it like a toaster, which it is not.

In other words, there is nothing in the system trying to be fit or trying to reproduce genes. It just… happens.

B: Of course there’s a utility function to evolution! How else can you explain how fine-tuned living beings are at surviving? Evolution is an optimization miracle.

A: An optimization miracle, say you? Okay, let me ask you a question. Which is the most successful life form on Earth, according to evolution?

In a simple aquarelle style, this image captures a platypus perched at the top of a gigantic mountain made of stylized animal skulls. The scene is rendered in a comedic, hilarious manner, with the platypus emanating a holy glow, adding a whimsical and surreal touch. The drawing style is simple and straightforward, emphasizing the absurdity and humor of the situation. The light-hearted take on this imaginative scenario, combined with the soft aquarelle textures, creates a visually appealing and amusing image.

And the winner of Darwin’s 4-billion-year old corpse pile is… the platypus?

B: Let me think about that… humans?

A: There are 8 billion humans on Earth. That may seem like a lot, but then there are over 30 billion chickens. So there are a lot more individual creatures on Earth with chicken genome than human genome. 

B: Right… then perhaps ants? There are 20 quadrillion ants in the world. If the optimization algorithm of evolution is aiming to maximize the number of genome-carrying individuals… ants are a good contender.

A: Those 20 quadrillion ants are divided among thousands of species. There is significant variance in the genome. You cannot speak of ants as the “single most successful species according to evolution”. Nor could you do this with bacteria, which deeply outnumber ants

B: Hm… maybe evolution’s most successful species is the one that uses the most resources, not the one with the most members. For example, a pig has several orders of magnitude more cells than an ant has, and so there might be more total genetic content in pigs on Earth than in ants. 

A: So evolution optimizes for total amount of genome content, and that’s how you determine which species is the most successful? The most complicated genome we’ve encountered is that of the microscopic water flea. It has 25% more genes than humans. If you were on a desert island in the company of your weight in microscopic water fleas, then they would be the most successful species, evolutionarily speaking, and not you?

B: Fine, you know what? People used to call the lion the king of the animals because it was at the top of the food chain. They were wise: they measured “evolutionary success” in how powerful a species was in its environment. I revert back to the humans. If we one day cage the stars, it’ll be hard to argue we aren’t the most evolutionarily successful species on Earth.

A: This is going to be even more difficult to justify as an optimization criterion than “increase amount of genome reproduced”! I would like to ask: where, precisely, do you see evolution optimizing for power in the environment? At what point in the chain of random mutations do you go “aha! The gene that built the more powerful animal continued on into the gene pool and the others didn’t!” To argue that evolution is a toaster, you need to tell me precisely where and when the bread turns into toast.

For instance: mitochondria lost their independence as a species and became enslaved by larger cells.1 They are not “powerful in their environment”. And yet, mitochondria are present in most life forms out there. They represent a lot of genome weight and a lot of discrete organisms. Doesn’t that make them better than lions or humans?

B: Gah! Fine, what’s your point in all this?



There is no optimization pressure from “evolution” at all. Evolution isn’t tending toward anything. Thinking otherwise is an illusion.

There is no “fitness level” or “power level” encoded into the chemical structures of molecules we call DNA: DNA exists separately from the purpose we give it. There is no such thing as “the most successful species, evolutionarily speaking, on Earth”. All that exists in real life is a bunch of atoms that take certain shapes: the rest is your map, not the territory.

So, the perceived purpose of the scorpion’s tail and the wolf’s fang exist only within our imagination. The only real thing you can talk about is the atoms that make up the tail and the fang and the shape of the DNA molecule which set off the chain reaction that led to the creation of the tail and fang.

It is in the shape of those molecules that we observe a selection effect. But first, I would like to distinguish between two parts of what we call life forms:

First, both you and I can point at something and more or less know whether it is a life form. The latin origin of “animal” is anima which means “breath” and has given rise to “animate” or “animated”. When something moves, chews on stuff, and runs away when you scream at it, it’s probably alive.

The reason you are scared of spiders and growling noises is because animals were everywhere in the ancestral environment, and if you wanted to survive, you had to watch out for them. You were selected to know what an animal was. There’s anthropic selection bias going on.

Whenever you look at nature, you’re going to see purpose everywhere, because being able to establish that the “purpose” of a wolf’s fang is to bite into your body and kill you was important for your ancestors. But the very fact that you see “purpose” is caused by a selection effect, where your purpose-noticing ancestors outperformed other humans.

In the same simple aquarelle style, depict a comical scene where a caveman is getting attacked by a wolf, with the wolf having bitten the caveman. The background features a candy pink sky, enhancing the whimsicality of the moment. Despite the chaos, the scene is portrayed humorously, emphasizing the absurdity rather than the violence. Hidden from sight, the platypus watches the spectacle from a bush, adding an extra layer of humor to the already comical situation. The entire scene is rendered in a simple, light-hearted aquarelle style, maintaining the comedic tone throughout.

Your ancestor discovers the purpose of fangs, while the platypodes watch smugly

Second, there is the deeper and more subtle selection bias with genes going on, for which we have no instinctive reaction to.2 It just happens to be a matter of fact that if you look out your window and see an animal, you are more likely to see genomes which seem to be optimizing for survival than genomes which seem to be optimizing for their own death.

So the deep mistake you are making is thinking of evolution as an active force instead of a passive observation. All Darwin did was notice a selection effect. There was nothing doing that selection effect, except reality.

Chemical reactions caused by the specific shape of DNA strands are an active force; a squirrel foraging for food is an active force; but evolution is not an active force. Evolution, when you look at the system very closely, simply does not exist. You see purpose everywhere, and so you are tempted to look at the system and see a toaster in which genes go IN and fitter genes come OUT. But the illusion of the toaster is emergent from what’s really going on, and you’re experiencing what it feels like to be on the inside of an algorithm.

You do not understand evolution until you eradicate all sense of direction from it. There is no intelligence behind evolution; there is no magic fairy helpfully molding scorpion genes into a tail for the scorpion; there is no alien god; and there isn’t even a process.


New Comment
11 comments, sorted by Click to highlight new comments since:

There is no optimization pressure from “evolution” at all. Evolution isn’t tending toward anything. Thinking otherwise is an illusion.

Can you think of any physical process at all where you'd say that there is in fact optimization pressure? Of course at the base layer it's all just quantum fields changing under unitary evolution with a given Hamiltonian, but you can still identify subparts of the system that are isomorphic with a process we'd call "optimization". Evolution doesn't have a single time-independent objective it's optimizing, but it does seem to me that it's basically doing optimization on a slowly time-changing objective.

Fair enough. I certainly didn't try to mince words. My goal was to violently shave off any idea of "agency" my friend was giving to evolution. He was walking around satisfied with his explanation that evolution selects for the fittest and is therefore optimizing for the fittest.[1]  The point of the dialogue format was to point out that you can call it an optimization process, but when you taboo that word you figure out it's hard to pinpoint exactly what is being optimized for. If you're going to call something an optimization process, you'd better tell me exactly what is being optimized for. If you can't, you are probably using that word as a curiosity stopper or something. 

I think we'll be able to pinpoint what evolution optimizes for, someday. [2] Gravity as a force optimizes for the creation of stars: enough so that loose clouds of hydrogen are pretty much guaranteed to form stars. You could say "gravity optimizes for the creation of stars from hydrogen clouds" and anticipate experience with seamless accuracy. Evolution is like this except it's so much more complex that in order to explain it as an optimization process you'll have to resort to the dreaded word "emergence". 

I think there's also something to be said about reminding people from time to time that "optimization pressure" and "emergence" and are in the map, not the territory; the territory is a different beast. I think you could reasonably take on the "true" way of seeing things for an hour or two after reading this post, and then go back to your business believing in the heuristic that evolution is an optimization process (once you've finished with your partial transfiguration). 

  1. ^

    Note the verb "optimized", which implies that something active is going on.

  2. ^

    In fact, most of the work has probably been done by Dawkins and others and there's a mountain of math out there that explains exactly what evolution is optimizing for. If that's the case, I definitely want to understand it someday, and find all of this very exciting. But neither I nor my friend are in a position to explain what evolution is optimizing toward, at least in a way that would let us accurately anticipate experience.

"Evolution" might be too broad a term: you could say things like asteroid impacts or volcanic eruptions, for example, apply no optimizing pressure on anything, and reasonably argue such things are part of evolution. Fair enough, so let's narrowly talk about "natural selection." That is an active force in every reasonable sense of the term, as much so as gravity. As certain as a dog breeder selecting for whatever he wishes, this process selects for … something. Precisely what is being selected for, … well, that's the Alien God aspect: something to do with gene frequencies, "inclusive fitness" or whatever.

Yeah I think I'm wrong about this. Thanks to all of you commenters for feedback. I'm updating.

I don't disagree with the content (as far as I could see, since I failed to really engage with it), but I'm infuriated by the style.  Please don't act this way here.

What specifically? I don't need a long explanation (you can get on with your life), just a pointer.

the condescending tone of exercises for the reader was particularly annoying.

Got it. Thank you.

This is probably correct relative to the average person's conception of evolution but probably not relative to the average LW reader?

Like even if you're just looking at chemical evolution, which is literally just some chemicals floating upwards and others not, I wouldn't bat an eye if you said "chemical evolution is a process that selects/produces amino acids and nucleotides" or even that "chemical evolution wants amino acids and nucleotides"

[-]Neil -10

"Process" and "wants to" are in the map, not the territory. I don't think anyone needs any justification for pointing that discrepancy out. Even if "process" and "wants to" are useful heuristics, I would not be miffed if LW posts resurfaced from time to time to remind everyone that we are not living in the territory here. I explain this in more detail in my response to Razied's comment. 

Every single word you might ever use about something in the real world is in the map, not the territory. It is semantically empty to point this out for three specific words.