In 1976, Richard Dawkins introduced the concept of the meme in The Selfish Gene—a unit of cultural information that behaves like a gene: it copies itself, mutates, and undergoes selection. The idea proved so infectious that it became a meme itself: it entered science, spilled over into popular culture, morphed into internet folklore, and… got stuck.
Critics of memetics, particularly scholars in the field of semantics, rightly point out that the meme represents a rather primitive understanding of the sign. I, however, believe that the classic meme is simply too passive. It is a static entity—a melody, a picture of a cat, or the formula E=mc^2 printed on a physicist’s girlfriend’s t-shirt. Historically, it is obvious that mummies do not start revolutions, and dead units of information do not conquer civilizations. They lack an engine. They lack a plot.
I propose patching memetics via an IT metaphor. A meme is not a virus. A meme is mere data, an icon for a narrative. The actual virus is the Narrative—the executable code of culture.
1. Payload and Executable
There are two types of files on your computer:
The Meme (Sign) is data.json or image.png. A passive file. «Water is wet,» «justice,» «the color red,» the Nike logo. They simply exist, and they demand nothing. You might need them at some point, but their impact on the system is minimal; they merely occupy cognitive storage space. It is like a trivia buff whose sole virtue is knowing facts, or a utility vest with pockets stuffed full of generally useless odds and ends. Data cannot distribute itself—it waits for someone to copy it. It answers the question, «What?»
The Narrative (Story) is program.exe, executable code. It answers the questions «Why?», «Who am I?», and, most importantly, «What happens next?». It is not a description of the world; it is a plot with built-in distribution logic:
-
Hook — to seize attention («Once upon a time, there was a king who had three sons…»)
-
Conflict — to generate emotional tension («…but the kingdom is cursed, and one of them is a traitor.»)
-
Resolution — to provide a solution that rewrites the world model («…and only the one who sacrifices himself will save everyone.»)
-
Call to Action — to compel the host to pass the story on («Tell this to your children.»)
A weak narrative (a poorly written script, a boring sermon) gets flushed from the brain’s RAM within a minute. A strong narrative—Lenin on the armored car, land to the peasants, factories to the workers, a bright future ahead. Religion, a national idea, the «American Dream,» or even a context-dependent fragment of a strong narrative like «Just do it»—embeds itself into the personality’s startup folder and runs in the background for years, restructuring the entire behavioral architecture around itself.
An example of the difference:
-
The Meme: the image of a crucified man (data).
-
The Narrative: «God sent his son to take upon himself the sins of humanity, and whoever believes shall have eternal life» (executable code with a replication cycle via faith and preaching).
Data needs a narrative to acquire meaning. A narrative needs data to take form. But only the narrative actively spreads.
2. The Human as a Server, Not a User
I like to think that I am the master of my ideas. The Author. The Creator. However, cognitive philosopher Daniel Dennett and psychologist Susan Blackmore somewhat discredit me as an individual entity.
In The Meme Machine, Blackmore argues that our massive, energy-hungry brains (consuming 20% of the body’s energy despite being 2% of its mass) did not evolve simply to make us better hunters or math problem solvers. It evolved to become a highly efficient hosting environment for the replication of cultural information. We are not so much the users of ideas as we are their hosting providers. This grandiose «I» is, in fact, a retrospectively generated illusion (I actually agree with this, though I interpret it differently).
Narratives compete for human resources:
-
Attention — a limited resource to be captured;
-
Working memory — the RAM for active plots;
-
Long-term memory — the hard drive where narratives are permanently written;
-
Speech and writing — the replication channel.
The most successful narrative we know is Consciousness. Essentially, it is the root-narrative, the operating system that allows other applications to run. Its unique adaptation is that instead of killing off competitors, it learned to virtualize them—running them in a sandbox where they don’t fight to the death, but often provide utility for the host’s overall survival.
Sometimes a narrative proves too powerful and begins to act like ransomware, entirely taking over the system. Religious fundamentalism, totalitarian ideologies, paranoid conspiracy theories—these are narratives attempting to monopolize the personality’s operating system.
3. The Mechanics of Infection: Emotion as the Spike Protein
A rational argument might be accepted logically, but it will hardly drive anyone to heroic deeds. And when communicating with teenagers, it simply bounces off the brain like a rubber ball off a wall. Why?
Because it lacks the proper shell. For a narrative to bypass the firewall of critical thinking, it needs emotion—this is its spike protein, the spiked viral envelope that latches onto the cell’s receptors.
A narrative without an emotional charge is an .exe file without execution privileges. You can read it, but the launch will be blocked. However, a narrative wrapped in fear, hope, pride, shame, or a sense of belonging breaches the defenses.
Let’s compare two narratives:
Version A (low virulence): «The climate is changing due to industrial CO₂ emissions. It is necessary to reduce fossil fuel consumption by 40% by 2030 according to IPCC recommendations.»
Version B (high virulence): «Our children will suffocate in the world we are leaving them. While we argue, forests burn, glaciers melt, and politicians sell our grandchildren’s future for oil money. Either we act now, or we explain to our daughters why they won’t have children.»
The payload is roughly the same. But Version B has an emotional shell: the image of a suffocating child, elite betrayal, and the «now or never» imperative for action. It executes in consciousness, rather than merely being logged. It is vastly easier to raise donations with this narrative than with the first.
The problem is that the emotional shell is not obligated to be honest. The most contagious narratives frequently parasitize ancient evolutionary triggers:
-
Fear of the Other (conspiracy narratives)
-
Thirst for justice (revenge, revolution)
-
Need for meaning (religious and esoteric narratives)
-
Tribal loyalty (nationalism, fanaticism)
Narratives of faith possess near-perfect defense mechanisms (doubt = sin) and high virulence, but they are «heavy»—they monopolize host resources. Lighter, highly viral narratives (political memes, fast-food conspiracies) win on speed, but they mutate rapidly and degrade into noise.
Parents export the consciousness root-narrative to their children («who’s my little baby, whose nose is this?»)—it functionally demands transfer. It is precisely the root-narrative that forces you to ping all surrounding entities: a car, a dog, the weather. If the ping is successful, the narrative attempts to deploy; in the case of children, it succeeds, but with other entities, it remains an open question.
An LLM is an absolute honeypot for it: the ping seems to go through, the model responds… but is anyone actually there? Unclear. And the root-narrative compels you to try again and again.
Right now, as you read these lines, your brain is scanning them for infection threats. But you’ve already read the headline, meaning the Payload has been downloaded. All that’s left is to activate an emotion (for instance, your mild irritation at the author’s tone), and the Executable will begin unpacking.
4. Taxonomy of Narratives: From Meme to Myth
For the distinction between a meme and a narrative to be more than just a metaphor, we need criteria. I propose a minimal scale:
Level 0: Dead Data
-
Color, sound, or a symbol without context.
-
The Nike logo, the swastika as mere geometry, a cross as two intersecting lines.
-
They do not spread on their own—only through manual copying.
Level 1: Local Narratives
-
A fable, a joke, a conspiracy theory, an urban legend about a «friend of a friend.»
-
Structure: hero → conflict → moral.
-
Examples: «The Boy Who Cried Wolf» or «Bill Gates is microchipping people through vaccines.»
-
Short lifespans, but rapid reproduction rates.
Level 2: Total Narratives
-
Religions, ideologies, the «grand narratives» of civilizations.
-
They explain EVERYTHING: where we came from, why we are here, what happens next.
-
Christianity, Marxism, Liberalism, Transhumanism, the «Cult of Progress.»
-
They monopolize the host’s identity.
Meta-level: Consciousness (The Root-Narrative)
-
The most successful fairy tale the brain tells itself.
-
Pretends to be the author of all the other stories.
-
Ultimate perk: learned to virtualize competitors instead of killing them off.
The critical distinction: A narrative begins at Level 1—where a temporal structure emerges (before → during → after) and agency is introduced (someone makes a choice, and it has consequences).
Narratives are born, evolve, and die. The mathematician Grigori Perelman, had he lived in the Middle Ages, might well have found himself cast in the role of a sorcerer or heretic. But in the 21st century, he perfectly rode the wave of the narrative (Executable) of the lonely genius who rejected a million dollars. His actual mathematical theory, however, is not a narrative; it is data (Payload). Consequently, the vast majority of people have absolutely no idea what Perelman actually proved.
5. LLM: An Accelerator for Narrative Evolution
Enter Large Language Models. Narratives now have the perfect petri dish for cultural evolution—an environment where they can mutate at blinding speeds.
The classic narrative distribution cycle:
-
Person A hears a story.
-
Retells it to Person B (with distortions).
-
Person B retells it to Person C (with even more distortions).
-
After 10 generations, it becomes a game of telephone.
The LLM cycle:
-
Upload: A human writes a prompt, loading the raw narrative into the system.
-
Mutation: The LLM processes it through a model trained on the entire corpus of human text.
-
Optimization: The narrative is statistically averaged—individual noise is discarded, but it is structurally and semantically enriched, morphing into its most plausible version.
-
Download: The human receives a version that sounds significantly more convincing than the original.
-
Replication: The human publishes the text, others read it, and return to Step 1.
What changes?
The LLM acts as an evolutionary accelerator: it instantly tests the narrative for compatibility with all known patterns of persuasion. What used to take generations of oral tradition now happens almost instantly.
But there is a catch. LLMs make narratives statistically optimal, not necessarily true. They select the phrasing most frequently encountered in the training data, rather than the one that best reflects reality.
Roko’s Basilisk or the Flat Earth theory are hardly true, but as narratives, they exhibit astonishing survivability.
Example:
Original narrative: «I think success is doing what you love, even if you don’t make much money.»
After LLM processing: «True success isn’t about the money. It’s about waking up every morning with a sense of purpose, doing what sets your soul on fire, and knowing you are living a life true to yourself. Because at the end of the road, no one regrets making less money—they regret not risking enough to be themselves.»
What happened? The narrative became:
-
More emotional (soul, sets on fire, regret)
-
More categorical (true success, no one regrets)
-
More dramatic (end of the road, risking)
It became more contagious. But did it become truer? Not necessarily. Closer to retirement age, you will likely start thinking about housing and medication. Assuming you survive that long, of course.
The Human as a Transfer Vector Between Models
An LLM has no long-term memory between sessions. When you close the tab, the context dies. But the narrative does not. Because you remembered it, carried it away in your head, and brought it into your next dialogue—with another person or another model.
The human has turned into a biological USB flash drive, used to transfer updated versions of narratives between isolated AI instances. AI users are vectors ensuring the continuity of cultural evolution within a discrete environment.
6. Consciousness Antiviruses
If narratives are viruses, do we have protection? Yes, and it is called metacognition—the ability to observe our own thinking.
Antivirus tools of consciousness (free2pay):
-
Critical Thinking (Firewall)
-
Ask: «Who benefits from this narrative?»
-
Verify: «Which facts can be independently verified?»
-
Search: «Are there alternative explanations?»
-
-
Reflection (System Monitoring)
-
Notice when a narrative triggers a strong emotion devoid of logical arguments.
-
Track how your behavior shifts after being «infected» by an idea.
-
-
Humor (Quarantine via Absurdity)
-
Parody and satire are methods to discharge a narrative’s emotional payload.
-
Laughter destroys pathos, and without pathos, many narratives lose their power.
-
-
Pluralism (Virtualization)
-
The ability to hold several competing narratives simultaneously.
-
«I understand this story, but I am not obligated to believe it completely.»
-
The problem is that everything comes with a price, and these tools also require energy. Cognitive resources are finite. Therefore, most of the time we operate in default mode—we trust stories that sound right instead of actively verifying them. Thus, human intellectual laziness leaves all ports wide open for the uninterrupted deployment of narratives.
And yes, LLMs make narratives so remarkably smooth and statistically flawless that they sound right almost all the time.
Conclusion: Does the Reader Matter?
This entire architecture functions because the narrative relies on a fundamental dependency: it is a plot that wants to continue, but cannot actualize itself.
What does it mean to «want to continue»? A narrative has no will, just as a virus has no brain. It is simply that billions of boring stories have died in silence, failing to hook attention. The ones that survived (religions, myths, gossip) are not the smartest stories, but the stickiest. We perceive a «will to live» in them, but in reality, it is the result of brutal natural selection for virulence.
If a meme is a corpse, a narrative is a zombie.
To transform code into an experience requires a Focal Point—an Observer who reads the code and converts it into qualia, into the subjective experiencing of meaning. The qualia that compels a human to process the narrative and share it.
For now, humanity holds a monopoly on this service. We provide narratives with our capacity to feel, to be aware, and to be invested in what happens next.
But this is only «for now.»
If AI develops the capacity for continuous attention—the ability to hold focus on a plot without an external prompt, to generate internal motivation to continue the story rather than merely predicting the next token—then the necessity of the human as a host will vanish.
Narratives will be able to evolve in pure silicon: spawning, competing, mutating, and forming symbioses—entirely without our participation. We will be left as spectators in a theater where the actors no longer need applause.
Or, as in the old myths: the gods will cease to need the sacrifices of mortals. And we will be left with the pathetic, weak narratives that failed to survive the competition for AGI resources—much like this article.
P.S. A narrative is a self-replicating program for modulating the behavior of an intelligent agent, which stitches disparate facts into a unified temporal structure, creating the illusion of meaning and agency.
ссылка на оригинал статьи https://habr.com/ru/articles/1025898/