Meme (Dawkins's Concept)

In the AI Story

Dawkins introduced the meme as a thought experiment: if replicators are defined by their capacity to copy themselves with variation under selection, are genes the only replicators? The answer was no. A melody heard and hummed passes from brain to brain. Each performance varies slightly. Some tunes are catchier, more memorable, more likely to be retransmitted — they possess higher memetic fitness. The tune that spreads is not necessarily the most beautiful or the most complex; it is the most effective at capturing attention and triggering retransmission. The logic is Darwinian, but the substrate is neural rather than chemical, and the tempo is measured in days rather than millennia.

The meme concept opened a second channel for the river of intelligence — one that operates in parallel with genetic evolution but at radically different speeds. Genetic information flows vertically, from parent to offspring, constrained by reproductive cycles. Memetic information flows horizontally, peer-to-peer, and can spread across an entire population in a single generation. This acceleration means that human culture evolves orders of magnitude faster than human biology, producing the mismatch between Pleistocene brains and twenty-first-century environments that underlies much contemporary psychological distress. Dawkins elaborated the framework in his 1993 essay Viruses of the Mind, arguing that certain ideas spread parasitically — replicating not because they benefit their hosts but because they are good at spreading.

Large language models represent a third substrate for memetic replication. When an LLM is trained on billions of documents, it extracts the statistical regularities that constitute the meme pool — the patterns of how ideas are expressed, connected, and deployed across human discourse. When it generates output, it recombines these patterns under the selection pressure of the user's prompt. The output is then accepted, modified, or rejected by the user, and accepted outputs propagate into codebases, publications, and further conversations. The cycle — variation (AI), selection (human), replication (deployment) — is structurally identical to memetic evolution in human populations, but the variation is generated at computational speed and the selection is applied in real-time conversational feedback. The meme pool has found a new ocean, and the evolutionary dynamics governing what thrives in that ocean are both familiar and unprecedented.

Origin

The coinage appears on page 192 of the original 1976 edition of The Selfish Gene, in a chapter titled 'Memes: the new replicators.' Dawkins needed a word that would rhyme with 'gene' and carry the sense of mimicry or imitation. He shortened the Greek mimeme (that which is imitated) to 'meme,' and the word entered the English language almost immediately. Within a decade, 'meme' appeared in psychology, anthropology, and philosophy journals. By the 2000s it had been appropriated by internet culture to describe viral images and videos, an appropriation Dawkins has tolerated with bemused frustration, noting that internet memes are indeed memes in the technical sense but represent only the most trivial class of the phenomenon. The serious study of memetics peaked in the 1990s and early 2000s, producing journals and conferences, but the field never achieved the institutional legitimacy of evolutionary biology. Dawkins himself has treated memetics as a suggestive framework rather than a rigorous science, useful for clarifying thinking about cultural evolution but requiring empirical validation that has proven elusive.

Key Ideas

Cultural replication. Memes copy themselves brain-to-brain through imitation, instruction, and communication — spreading at the speed of conversation rather than the speed of biological reproduction.

Variation through transmission. Each time a meme is copied, it is transformed by the receiving brain's interpretation, producing the variation on which selection acts.

Fitness criteria differ from truth. A meme spreads if it is memorable, emotionally compelling, and socially rewarding to share — properties orthogonal to accuracy, making the meme pool a distorted rather than faithful representation of reality.

Parasitic memes exist. Ideas can replicate not by benefiting their hosts but by being effective at spreading — the memetic equivalent of a virus, producing harm while propagating successfully.

AI as memetic substrate. Large language models trained on human text become repositories and generators of memetic patterns, accelerating cultural evolution to computational speeds and creating feedback loops between human and machine meme pools.