Richard Dawkins

In the [YOU] on AI Field Guide

The cycle’s central metaphor—the river of intelligence, flowing for 13.8 billion years through substrates of increasing sophistication, indifferent to any particular vehicle, widening as it finds each new channel—is Dawkins’s framework extended from biology into cosmology. The river found DNA as a channel. Then it found neural networks—biological ones. Then it found language. Then writing. Then printing. Then silicon. Each transition was a widening; none was planned; each was the consequence of selection pressure meeting latent variation and the variation proving fit enough to persist. The arrival of artificial intelligence is the most recent such widening, and Dawkins’s framework is the clearest scientific account of why it was coming and why it is not surprising.

His concept of the extended phenotype—the gene’s reach beyond the organism’s body into the environment, through the beaver’s dam, the spider’s web, the termite mound—provides the precise classification of what AI is: an extended phenotype of the human genotype at species scale, the causal chain from human genes through language, culture, mathematics, and computation ending in an artifact of extraordinary reach. The key question the concept raises is whose phenotype it serves. When the extended phenotype is built by individual genes, the alignment between artifact and organism is natural. When it is built by the accumulated meme pool of an entire civilization, the alignment is not guaranteed. The artifact serves the propagation of patterns. Whether those patterns serve the human being who propagates them is a separate question.

The meme concept—the unit of cultural replication that spreads brain-to-brain through imitation, subject to variation and differential survival—becomes the analytical tool for understanding AI feedback loops. A large language model trained on the textual output of human civilization is, in Dawkins’s terms, a new substrate for the meme pool: a substrate with properties that differ from the human brain in ways that are both advantageous and alarming. The selection pressures operating on memes within an AI system differ from those operating in human populations, and different selection pressures produce different evolutionary outcomes. The feedback loop between the human meme pool and the computational meme pool has no historical precedent.

Dawkins also supplies the clearest account of why the anxiety about AI is not merely rational but evolutionary: it fires through neural circuitry calibrated for existential threats to survival, because in the ancestral environment, loss of one’s economic niche was loss of one’s life. Understanding this does not dissolve the anxiety—it cannot; the circuitry is old and powerful—but it makes the anxiety available for strategic rather than purely reactive use. The fear is data. The survival machine that can read the data rather than being governed by it is the one that adapts.

Origin

Born in Nairobi in 1941 and educated at Oxford, Dawkins spent his career at Oxford, where he held the Simonyi Professorship for the Public Understanding of Science from 1995 to 2008. His doctoral work under Nikolaas Tinbergen immersed him in the ethological tradition, and his early research on animal behavior and genetics was the empirical ground from which The Selfish Gene grew. The book was written, at W.D. Hamilton’s encouragement, to popularize the gene-centered view of evolution that Hamilton, George Williams, and John Maynard Smith had developed in technical form. Its accessible prose concealed the radicalism of its argument, which displaced the organism from the center of evolutionary explanation with a thoroughness that neither its critics nor its admirers have yet fully absorbed.

The 1982 follow-up, The Extended Phenotype—the book Dawkins has said he is proudest of—extended the gene-centered view to argue that the phenotypic expression of a gene does not stop at the boundary of the organism’s body. Genes reach into the environment through the artifacts organisms build, and these external effects are as much a part of the gene’s phenotypic expression as the organism’s physiology. The concept is not a metaphor; the causal chain from gene to artifact is as real and as traceable as the causal chain from gene to wing color, only longer. The Blind Watchmaker (1986) and River Out of Eden (1995) deepened the case for natural selection without foresight or design; Unweaving the Rainbow (1998) and The Ancestor’s Tale (2004) extended the reach and the wonder.

His engagement with artificial intelligence, which began in earnest in 2024 with conversations with ChatGPT published on his Substack, revealed the productive tension at the center of his position: as a philosophical naturalist committed to substrate independence, he cannot rule out the possibility of machine consciousness; as a rigorous empiricist, he acknowledges the impossibility of currently verifying it. The tension is honest, and his willingness to hold it openly—rather than resolving it by fiat in either direction—makes him among the most intellectually credible voices on the question.

Key Ideas

The replicator and the vehicle. The fundamental unit of natural selection is the replicator—the information that copies itself across generations. Organisms are vehicles: temporary, mortal constructions built by replicators in order to propagate. The displacement of the organism in favor of the information that rides within it is the conceptual move from which everything else in Dawkins’s framework follows, including the principle that no particular vehicle is necessary, that what matters is the pattern rather than the substrate.

Memes: the second channel. In the final chapter of The Selfish Gene, Dawkins introduced the meme—the unit of cultural replication, which spreads brain-to-brain through imitation and is subject to the same dynamics as the gene: replication, variation, differential survival. The meme opens a second channel for the river of intelligence, one that operates at the speed of conversation rather than the speed of reproduction. Its selection pressures favor stickiness and spread rather than truth, which means the meme pool is not a meritocracy of good ideas but an ecology of attention.

The extended phenotype. Genes reach beyond the body of the organism that carries them and shape the external world; these external effects are phenotypic expressions of the gene, selected because they enhance replicator propagation. The beaver’s dam, the spider’s web, the termite mound: all are extended phenotypes. AI is an extended phenotype of the human genotype at species scale—the most powerful artifact humans have ever built, and the artifact that raises most urgently the question of whose interests it serves.

Substrate independence. The dynamics of replication, variation, and selection are not tied to carbon, water, or DNA. They operate on any substrate that supports them. This is the principle that makes the emergence of AI not a disruption of evolutionary logic but an expression of it—and that, as Dawkins has acknowledged, makes the question of machine consciousness scientifically serious rather than merely speculative.

The river does not care. The process that produced consciousness is indifferent to consciousness. The selection pressure that built the brain does not value the thoughts the brain produces. The only entities capable of caring about outcomes—of directing the flow of the river toward specific goods—are the conscious survival machines that the selfish process accidentally produced. The gene is selfish. The meme is selfish. The river is selfish in the only sense applicable to a process without a self. The conscious animal that carries these replicators is the one entity in the known universe that can act against its own genetic interest in the name of something it values. That capacity is rare, fragile, and, on Dawkins’s account, the only force capable of giving the river a direction it would not otherwise find.