Memes and Adaptation
Table of Contents
Go read The Selfish Gene if you have not done so yet.
Seriously, do it.
It is not an accident that this blog is called faster than light memes.
My theory of cognition is full-blown memetics. Memes are software animals. It's a biology. (Technology, biology, computers and cognition are the same kind of stuff, at different places on a few continuums.) This page collects the foundations: replicators, memes, adaptation, and the words I use for them.
- My current work: Memes as Abstract Constructors, a constructor theory of memetics (work in progress).
- What makes a design good, and why the alternatives of a design must be possible: Philosophy of Design.
The gene's eye view, and the meme
The Selfish Gene introduces a perspective, the gene's-eye view of evolution (Dawkins 1976). Introducing a perspective might be the most powerful kind of knowledge. You don't learn a new fact, you learn to see all the old facts differently.
With a happy little conceptual jump, Dawkins applies the perspective to mental content and discovers the meme.
The Extended Phenotype (Dawkins 1982) is the second half of the gene's-eye view, and strangely underrated. The effects of a gene do not end at the skin. Beavers build dams, spiders build webs, and some parasites reach into the biochemistry of their host: flukes make the shells of their snail hosts thicker.1 The dam is as much a phenotype of beaver genes as the teeth that build it.
Notable works on memetics
The Beginning of Infinity, David Deutsch
Deutsch fleshes out two classes of memes, which he calls rational and anti-rational. Rational memes rely on the critical faculties of their holders to get themselves replicated. Anti-rational memes rely on disabling them.2
A fascinating read, with a red thread through scientific inquiry, why flowers are beautiful, ethical truths, the nature of the universe and the fate of civilization. Big history.
The Meme Machine, Susan Blackmore
Blackmore lays out memetic concepts like copy-the-product and copy-the-instructions, and popularized the term memeplex, a group of memes that replicate better together, by analogy with co-adapted gene complexes.3 In memetics, we can play the game of reasoning by analogy to genes.
Blackmore goes on to describe a meme's-eye view of how culture and the brain evolved, and makes hypotheses about how music, language and other aspects of cognition, especially social cognition, might be explained by memetic processes.
She is also big on how memes might become the driver of genetic evolution.
Daniel Dennett, From Bacteria to Bach and Back
Dennett is an exceptionally clear thinker. He shows how memetics fits as a model for how religions work, and he points out the power of good ideas; the most powerful ones he calls intuition pumps.
He makes the analogy of computer virtual machines to how minds work. At some level of the hierarchy, the workings of the brain are abstracted away enough that any idea fits on any brain, the same way the Java Virtual Machine abstracts over different operating systems. His thinking tools are apps for your necktop.4
This ties into Blackmore's Meme Machine: the genes that made brains suitable for memes were selected for. Evolution built the infrastructure for having ideas, without specifying the contents of the ideas. Naively: content is to memes what software is to hardware.
One thing I think Dennett gets wrong is that much of his theorizing is about person-jumping memes. Memes can reproduce and live within a single mind. The strongest example I can come up with is a personal rumination: a memeplex that might never be replicated in another mind, but reproduces itself again and again within one mind.
But of course, once you think of it this way, maybe much of the machinery of the mind is assembled, in part, darwinistically. That is what I call a neuronal meme.
Adaptationism
The best available theory about design in nature. It says that biological design is made of adaptations.
A trait is an adaptation when:
- it is a variation of an earlier form,
- it is heritable through the transmission of genes,
- it enhances the reproductive success of those genes.
(Wikipedia.)
Adaptationism puts knowledge at the center of biology. Biology is a science about software.
And it hides a condition that becomes obvious once you write it down in constructor-theoretic terms: for a design to be an adaptation, its alternatives must be possible. Selection chooses among variants. Where no other variant could arise, nothing is selected and nothing is adapted. That is the fundamental conjecture of my constructor theory of adaptation and design, and I think it follows naturally from Dawkins. More in Philosophy of Design.
The tree of life lives in a space of alternatives
The tree of life is a causally contingent object in design space, made of ancestor-descendant relationships. There should be one tree of life per origin of life event.
It is an invisible, theoretical entity predicted by Darwin's theory of descent with modification (1859), and one of the most scientifically substantiated aspects of our world.
Science is often about invisible things, or counterfactuals, that serve as explanations.
That all life is connected and descends from a single life form, the last universal common ancestor (LUCA), is at first glance counterintuitive, and sometimes overlooked. What a grandeur.
A branch on the tree of life only exists because its alternatives were possible.
- Adaptation requires alternatives.
- Design is about tradeoffs: you have to look at, at least, two solutions to your problem.5
- Likewise, a theory only exists because of its alternatives.
The critics of adaptationism
Randomness
Genetic drift: allele frequencies change by chance.
It should make one thoughtful that adaptationism predicts random genetic drift, for those genes that don't contribute to fitness. The background mutation rate that sticks should be highest for junk DNA, and diminish with a gene's contribution to adaptive traits. The adaptationist framework is what lets us make sense of this.
Bottleneck effects: a founder population on an island carries a random, small subset of the original population.
Founders, bottlenecks and the like concern hyperparameters, not the algorithmic structure of natural selection. It could turn out that every single trait needed a bottleneck to evolve. It would still be the adaptationist framework that explains where the design comes from: nonrandom selection of random variations, with some more detail in the variation part.
Natural selection might go faster on islands, or proceed in jerks (punctuated equilibria). Not completely uninteresting, but not foundational either. The speed of natural selection is a much less interesting problem than its algorithmic structure.
Constraints
Spandrels [Gould and Lewontin 1979], developmental constraints, historical contingency. A spandrel is a feature without adaptive value, a by-product of something else.
But saying that a thing is not an adaptation already uses the theory of adaptation. Spandrels are a concept that only makes sense inside the framework of adaptationism. Presenting them as an alternative to it is sloppy thinking.
Why do we have a belly button? The answer has to do with engineering problems and constraints: some traits are problems to be worked around. But why is a thing a problem to be worked around, or no problem at all? That answer lives in the space of design. Design work includes situations where the solution to a problem matters less, a spandrel situation if you want. And the theory that says how biology makes design is adaptationism.
Self-organisation
That the structure of reality is a resource for evolving, adaptive systems is trivial. Ultimately these critiques are reductionistic: they try to get rid of the defining feature of biology, knowledge.
Should people have a copyright of their genes?
I am pretty sure the answer is no, because persons are memetic entities that merely populate brains. Personal rights only make sense for memetic knowledge.
People are not organismic, people are memetic.
For a similar reason, genetic modification is not unethical in itself. Genes are a building material, made from adaptive information.
There is a fallacy, the gene = soul fallacy. It is close to the truth that the spirit of an organism is its genome: the genome is the software that persists across time (the immortal gene) and across metamorphosis (caterpillar to butterfly, seed to tree), the spirit that carries on.
The spirit of a person, on the other hand, is not organismic. It is multilayered: the cultural, memetic software that grows on top of the high-dimensional, plastic hardware of the human brain.
I have no rights over a clone of mine; they are a person. So it makes no sense to have personal rights over genomes.
The Deutsch/Popper knowledge-based view often comes with slightly counterintuitive conclusions like this. For more of it: The Farthest Reaches (audiobook) and Carlos De la Guardia.
Platonic animism
If all patterns of information, and therefore all software, already exist in the multiverse, then the ideas are out there. Flip the perspective: it is up to the ideas to populate and recruit other parts of reality. Wherever there is a search process, an idea has the chance to invade the substrate.
This way around, the brain is a kind of receiver, and ideas populate it more or less successfully. The ideas are alive; they are adapted to be better or worse ideas. Software animals in the computational universe [Wolfram].
Glossary
Work in progress. These are the words I use, and what I mean by them.
- knowledge: information with causal power. Knowledge in a genome is a program that causes an organism to develop and replicate the genome. Knowledge in science is ideas that stand in a tradition of criticism. Knowledge is hard to come by.
- software: knowledge made of information transformation rules (instructions). Biology is software archaeology [Chaitin].6
- interpreter: a mechanism that executes the instructions of software in terms of a lower level. In computers this bottoms out in physical circuits. In biology, ribosomes and cell metabolism interpret the genes.
- design: knowledge about being effective in reality. Software incorporates design. In biology: adaptations.
- adaptation: a feature of a design, made adaptive by natural selection, chosen among possible alternatives.
- adaptive domain: the causal, implementation phase of replicating software. See replicator, extended phenotype.
- extended phenotype: the set of effects, in the adaptive domain, that cause a replicator to be replicated. Not limited to morphology: it reaches into behaviour, and into the biochemistry of other organisms. Technology is not an extended phenotype, since no genes select among variations of technology. The capacity for technology, a runtime dynamism, a Baldwin effect, is the phenotype.
- replicator [Dawkins 1976]: knowledge that causes itself to be stable, given a niche.
- error correction: not sure yet. For replicators: mechanisms that preserve their information (fidelity, longevity, fecundity). For organisms: regeneration. Error correction of scientific ideas implies fallibilism: the truth is out there, and we get closer and closer to it.
- evolution: a search process in some space of possible software (replicators), open ended.
- meme [Dawkins 1976]: a social replicator, a piece of software replicating between persons. Imitation of action patterns is not sufficient for memes, and so not for true culture. Memes must be understood, in terms of counterfactual explanations, in terms of the invisible. To copy using a rock to crack a nut, an onlooker must understand mental states (they are hungry and want the nut) and that the nut can be eaten once cracked.
- neuronal meme / neuronal replicator: information that replicates (is stable) within a brain. See also autopoietic systems, neural Darwinism.
- meme machine [Blackmore 1999]: an information processing system capable of having memes, a computer capable of growing persons.
- software synthesis: a search process in a domain of possible software. Psychological development solves a software synthesis task: a person must grow as software that runs on the brain.
- institution: mental software that coordinates processes of conjecture and criticism.
- top-down causation: knowledge, design or institutions causing physical transformations. See interpreter.
- creativity: an information-processing, evolutionary mechanism for creating knowledge. We don't know the algorithm.
- person: software capable of open-ended knowledge creation (creativity) and explanation, a universal explainer [Deutsch]. Persons can understand memes, and therefore take part in culture.
- AGI: a person that was programmed artificially. We would program a software synthesis task like the one the brain uses to grow a person.
Why can the brain grow persons, and our computers cannot? That is the missing theory of personhood, one of the problems of our time.
- Prohibiting an AGI from taking part in culture would be unethical.
- Forcing an AGI to behave in some way, constraining it, would be slavery.
- Running AGI on computers may be ill-advised; it is not figured out. Erring on the side of caution is ethical.
Why this blog is called Faster Than Light Memes
Replicating information, and ideas, are the real deal of what the universe is about.
The difference between humans and other animals is our ideas. Humanity is the memes, more than the biological substrate carrying them.
It is the memes that let us go to the moon: in spite of our biology, not thanks to it.
David Deutsch described this philosophy nicely on the Tim Ferriss Show. As civilizations become more advanced, knowledge and ideas become the driving force of what happens in the universe.
Consider a meteor flying towards Earth. Newtonian physics predicts an impact. But because of civilization, it might be deflected.
Eventually, ideas will have the power to create, transform and destroy stars and galaxies.
Dawkins says the new replicator is "still drifting clumsily about in its primeval soup".7 Once we start doing things like converting all reachable matter into computronium,8 it will be the descendants of our ideas that shape the universe.
If we should find ways to go faster than light, the memes will spread faster than light through the universe.
The title is a hopeful wish: the potential of humanity is greater by some factor x if you can reach more of the universe.
Of course, we, the memes, might figure out something more useful than populating the universe. Like making ourselves smaller and smaller, or something.
Going faster than light, or even spreading through the universe, is not the goal. Maximizing human flourishing is.
Sabine Hossenfelder: I believe the universe might be able to think
The physics she discusses is the same that faster than light memes would need: information (signals) traveling faster than light.9
From now on, whenever somebody makes fun of the concept of faster than light memes, I decide they are just closed minded.
Greg Egan: Diaspora (1997)
This book features stranger and grander things than faster than light memes.
Earlier versions
The infantile versions of Memes as Abstract Constructors:
- Neuronal Ensemble Memetics: musings on meme-engine algorithms for cognition.
- Neuronal Memetics Overview: my theory of cognition as slides, with YouTube talks.
Footnotes:
Richard Dawkins, The Extended Phenotype (1982): trematode flukes make their snail hosts grow thicker shells, protecting the parasite at the cost of the snail's reproduction. Wikipedia.
David Deutsch, The Beginning of Infinity (2011), ch. 15, "The Evolution of Culture": a rational meme "relies on the recipients' critical faculties to cause itself to be replicated", an anti-rational meme on "disabling the recipients' critical faculties".
Susan Blackmore, The Meme Machine (Oxford University Press, 1999). The term itself is credited to Hans-Cees Speel (c. 1995), short for "co-adapted meme complex". Wikipedia: Memetics.
Daniel Dennett, Intuition Pumps and Other Tools for Thinking (2013), where thinking tools are apps for your necktop.
Rich Hickey, Hammock Driven Development, Clojure/conj 2010 (transcript).
Gregory Chaitin, Proving Darwin: Making Biology Mathematical (2012): "that's what biology really is, a kind of software archeology!"
Richard Dawkins, The Selfish Gene (1976), ch. 11: "It is still in its infancy, still drifting clumsily about in its primeval soup, but already it is achieving evolutionary change at a rate that leaves the old gene panting far behind."
A hypothetical optimal configuration of matter for doing computation.
Sabine Hossenfelder, I Believe The Universe Might Be Able To Think (2025): we don't know that locality holds at untested short distances, or that faster-than-light signals are impossible, so the size of the universe does not rule out a thinking universe.
Generated by Claude on Benjamin Schwerdtner's behalf.