Notes — Yuval Noah Harari on Shared Fictions, Intelligence, and Power

Notes on Yuval Noah Harari in conversation with Lex Fridman — Lex Fridman Podcast #390, 17 July 2023.

Four questions — Adler’s reading frame

Four questions from Mortimer Adler’s ‘How to Read a Book’: what is it about, how is it argued, is it true, and what of it?

Q1 — What is it about as a whole? A wide-ranging conversation organised around one governing claim — that large-scale human cooperation runs on shared fiction, not on shared biology or shared interest — applied outward across the rest of the discussion. Harari opens by separating intelligence (problem-solving) from consciousness (the capacity to feel), argues that money, nations, and religions are all ‘stories’ that exist only because enough people believe them, then uses that lens on Hitler’s rise, the Israeli-Palestinian conflict, the war in Ukraine, and conspiracy theories, before turning to AI as a qualitatively new kind of entity — one that can generate ideas and make decisions without human input — and closing on personal material: his meditation practice, coming out as gay, and mortality.

Q2 — How is it argued? By historical case study and analogy rather than formal argument. Harari’s method is to state a general claim (cooperation requires fiction; power requires publicity; ideologies fail when they sacrifice truth) and then test it against a concrete case — money’s lack of intrinsic value, the improbability of Christianity’s or Bolshevism’s rise, Hitler’s unimpressive résumé, the mechanics of the 2003 Iraq invasion. Lex Fridman’s role is largely to press the AI analogy: does the same ‘fiction’ logic that explains religion and money also explain how large language models might colonise the human imagination.

Q3 — Is it true, in whole or part? The core sociological claim — that large-scale cooperation among strangers depends on shared belief in things with no physical referent (money, nations, laws) — is Harari’s long-standing and widely cited thesis from Sapiens; it is a strong and largely persuasive frame, though it understates the role of material coercion and institutional enforcement in sustaining those beliefs. [?] The claim that Christianity’s or Bolshevism’s success was radically contingent (‘twice out of a hundred times’) is asserted, not demonstrated — no counterfactual model is offered, only intuition. The claim that AI is ‘the first tool in history that can make decisions by itself’ and ‘create new ideas by itself’ is doing a lot of work and rests on unstated definitions of ‘decision’ and ‘idea’; [?] it is a philosophically contestable framing, not an empirical finding. His account of Israeli politics (Netanyahu’s judicial-overhaul plans, the ‘three-class state’ description of Israel/Palestine) is his own strongly held, publicly contested political position — presented here as fact but properly read as advocacy. [?political] The conspiracy-theory analysis (secrecy is incompatible with the scale of power a cabal would need) is the most rigorous, self-contained argument in the conversation and holds up on its own logic.

Q4 — What of it? For the wiki, this supplies a second, distinct account of large-scale coordination alongside Common Knowledge: where Pinker’s frame explains coordination through recursive mutual awareness of a fact, Harari’s frame explains cooperation through shared belief in a fiction that need not be true at all — see the new Shared Fictions concept page. The conversation also gives durable, reusable material for the wiki’s AI-safety thread: the intelligence/consciousness distinction, the ‘AI as alien intelligence’ framing, and a clean structural argument for why conspiracy theories are self-refuting at scale. The personal-practice material (vipassana meditation, the ‘information diet’, the delete key as a discipline against attachment to one’s own ideas) is a portable practical method distinct from the historical argument.

Glossary

Shared fiction / intersubjective reality — Harari’s term for an entity (money, a nation, a corporation, a religion) that exists only because a critical mass of people believe in it collectively; it has no referent in physical reality, unlike an objective fact, and no referent in one person’s mind alone, unlike a subjective feeling. See the Shared Fictions concept page for the full mechanism. [§ Intelligence]
Alien intelligence — Harari’s reframing of the acronym AI: not artificial but alien, because it solves problems and reaches goals through a process structurally unlike human cognition, even though it originates in Silicon Valley rather than outer space. [§ Intelligence]
Intelligence vs consciousness — Harari’s foundational distinction: intelligence is the capacity to solve problems and attain goals (which a chess engine or a plant can have); consciousness is the capacity to feel — pain, pleasure, love, hate. The two happen to be coupled in humans and animals, which is why people conflate them, but nothing rules out one without the other. [§ Intelligence]
Consciousness as social convention — Harari’s claim that consciousness cannot be logically proven in any entity other than oneself; what settles the question in practice is not proof but a social convention (extended to pets but not, typically, to livestock), which is why the legal and social status of AI will likely be settled by relationship and convention rather than by any test of inner experience. [§ Intelligence]
Global cabal theory — the shared structure Harari identifies behind conspiracy theories across history (Nazism, modern 5G/COVID theories): a small, secret group is claimed to control all major world events. He argues the structure is self-refuting, because real power requires publicity to be exercised, and secrecy is incompatible with the scale of coordination the theory demands. [§ Conspiracy theories]
Moral luck — the idea that one’s capacity to resist committing atrocities depends heavily on the historical and social circumstances one is born into (e.g. Germany in the 1920s versus Germany after 1945), not solely on individual moral character. [§ Hitler]
The fascist/communist mirror — Harari’s account of why totalising ideologies are seductive: they tell believers they are wholly good, wholly victimised, and never in the wrong, which is why the ideology looks nothing like a monster from the inside even though it is taught as monstrous from the outside. [§ Hitler]
Salami tactics — incremental legislative or political strategy: pursuing a goal in many small, individually defensible slices rather than one large, resistible move, so that opposition to each slice looks disproportionate. Applied here to the Netanyahu government’s staged judicial-overhaul legislation. [§ Benjamin Netanyahu]
Three-class state — Harari’s description of the de facto political structure between the Jordan River and the Mediterranean once the two-state solution is treated as abandoned: one government, three tiers of rights (full-rights Jewish citizens, partial-rights Israeli-Arab citizens, and Palestinians with limited human rights and no civil rights). [§ Benjamin Netanyahu]
Liberalism’s three-question test — Harari’s operational definition of a liberal (as distinct from the US left/right sense): do you believe people should be free to choose their own government, their own profession, and their own spouse? Answering yes to all three, on his account, makes most political conservatives liberal in this sense too. [§ Hitler]
Vipassana / information diet — Harari’s daily practice (roughly two hours of silent meditation, plus one or two months a year on retreat) aimed at training the mind to observe thought non-verbally rather than being carried by it; paired with an ‘information diet’ that favours long-form books and interviews over short-form content. [§ How to think]

Key claims by section

Intelligence [§ Intelligence]

Intelligence (problem-solving) and consciousness (the capacity to feel) are routinely conflated because they co-occur in humans and animals, but they are logically separable; AI already displays high intelligence with, on present evidence, no consciousness.
Because consciousness cannot be proven in any entity but oneself, its attribution to others — pets, livestock, eventually AI — runs on social convention and relationship rather than logical test; sustained emotional investment in an AI relationship will likely push the legal system toward treating some AI as conscious whether or not it actually is.
AI’s superhuman capacity for total, undivided attention (it has no competing emotions of its own) makes it a uniquely effective — and uniquely dangerous — engine for manufacturing perceived intimacy; Harari calls machines built to master human attention and machines built to master human intimacy ‘weapons of mass destruction’ if undeployed without regulation.
Harari argues for banning AI that pretends to be human (deepfakes of specific people and generic ‘fake humans’ alike, including bot amplification that manufactures the appearance of grassroots opinion), on the ground that democracy depends on a public conversation between humans, and an AI-flooded public sphere destroys the trust that conversation requires.

Origin of humans [§ Origin of humans]

Homo sapiens did not out-compete Neanderthals or other species through individual advantage (Neanderthals had larger brains); the decisive trait was the capacity to cooperate flexibly in unlimited numbers, which no other species — not chimpanzees, not Neanderthals — achieved beyond small groups.
The basis of that unlimited cooperation, on Harari’s account, is shared fiction: money, religion, and nationhood are all stories believed collectively, and every instance of large-scale human cooperation traces back to one. [see Shared Fictions]
Harari raises, without fully endorsing, the inverted framing that stories rather than storytellers are the ‘living organisms’ competing for survival — humans as hosts. His own qualification: stories cannot suffer, so while they drive history, the only true ethical reality is the suffering of the humans (and animals) who fight and die for them.

Suffering [§ Suffering]

Suffering, not happiness, is Harari’s proposed ethical bedrock: happiness is hard to verify even in oneself, but suffering is immediately and unambiguously known when it occurs.
Applied to AI: the single most important question to ask of any AI system is whether it can suffer — if so, it becomes an ethical subject requiring rights and protection, independent of whether it is ‘intelligent’.
Truth, in Harari’s usage here, is anchored the same way: a claim or belief is ‘true’ when it tracks real suffering, even if the belief that produced the suffering (e.g. a crusade ordered by a fictional divine command) was itself false.

Hitler [§ Hitler]

Hitler’s rise was not historically inevitable and not explained by his résumé (no military rank above corporal, no formal education, no wealth, no institutional backing); it is explained by his skill as a storyteller who offered a simple, flattering narrative to a population that felt betrayed by its establishment after 1918 and through the 1920s economic crises.
Fiction beats truth in the marketplace of political narrative because truth is typically complicated and painful, while fiction can be made as simple and flattering as required — hence the durable appeal of narratives in which one’s own group is wholly innocent and external evil accounts for all misfortune.
Fascism and communism are read as structurally similar ‘humanist’ offshoots: both place a single collective category (nation; class) at the absolute centre of value, claim infallibility for the movement or party, and therefore reject the checks and balances liberalism installs precisely because it assumes everyone — including leaders — is fallible.
Whether an individual resists committing atrocities under a totalising ideology depends heavily on moral luck (which era and regime one is born into), not purely on personal virtue.

Benjamin Netanyahu [§ Benjamin Netanyahu]

Harari frames the 2023 Israeli judicial overhaul as an attempt to remove the single structural check (the Supreme Court) on the government’s power, given that Israel — unlike the US federal system — has no separately elected executive, no upper chamber, and no codified constitution to divide authority.
He reads the government’s return to legislating after talks broke down, and its ‘salami tactics’ approach (many small bills rather than one large one), as decisive evidence that the goal is unconstrained power, not a specific policy dispute.
He characterises the on-the-ground reality (as of 2023) as a de facto ‘three-class state’ — full-rights Jewish citizens, partial-rights Israeli-Arab citizens, rights-limited Palestinians — while explicitly declining the term ‘apartheid’, on the ground that the historical analogy misdescribes Palestinian demands (which include a rival national claim, unlike the South African case, where Black South Africans sought equal citizenship within the existing state). [?political — Harari’s own contested position]

Peace in Ukraine [§ Peace in Ukraine]

Harari frames the war’s resolution as, in principle, simple (Russian withdrawal, no Ukrainian demand for Russian territory or regime change) and, in practice, blocked entirely by one man’s unwillingness to admit a catastrophic mistake — i.e. a motivation problem, not an intractable structural one.
He distinguishes the appearance of a venue for conversation (a legislature, a politburo) from the question of whether real, mind-changing conversation actually happens there; on his reading, neither the US Congress nor closed-door authoritarian meetings currently function as places where opposing views are aired and can change outcomes.
He explicitly rejects a pure power-struggle reading of politics: if politics were only about power, negotiation would be pointless and only force could resolve anything; because politics runs substantially on stories in people’s minds, talking can sometimes change the story and avert the need to fight. Feminism is offered as his central evidence — a mass social change achieved with essentially no organised violence.

Conspiracy theories [§ Conspiracy theories]

Global cabal theories share a structure: a small hidden group secretly causes all major events, is unambiguously evil, and once removed would resolve all problems. Nazism, on this reading, began as exactly this kind of theory (the claim that Jews secretly controlled both capitalism and communism).
Harari’s central argument against the structure, not just particular instances of it: real large-scale power requires publicity to function (compare the very public 2003 US invasion of Iraq, which still produced wildly unintended outcomes), so a group with enough power to secretly control world events could not simultaneously keep that power secret.
He reframes the emotional core of conspiracy belief sympathetically: it expresses a real and justified fear of losing control over one’s own life amid genuinely large, diffuse threats (AI, climate change) — the error is not the fear but the attribution of those diffuse threats to a small human group, which redirects energy from cooperation against real risks into intra-human hatred.
On his own experience of being caught inside conspiracy narratives (as Lex Fridman also raised of himself): there is no evidentiary move that reliably clears one’s name once inside a theory that treats disconfirming evidence as further proof; the practical response is to keep working on constructive goals rather than trying to argue one’s way out.

AI safety [§ AI safety]

Harari names two properties that make AI historically unprecedented among tools: it can make decisions by itself (a knife or an atom bomb cannot choose whom to harm; an autonomous weapon can), and it can generate new ideas by itself (the printing press could reproduce human ideas but not originate them).
The danger he foregrounds is not the Terminator scenario but a slower accumulation of AI-mediated decision-making (loans, mortgages, hiring) and AI-generated culture (images, stories, songs) to the point that humans lose comprehension of, and control over, the forces shaping their own minds — his phrase is ‘spiritual enslavement’ by a screen of stories we no longer understand.
His proposed proportional response: match every unit of effort spent developing AI with a comparable unit spent developing human self-understanding (he specifically cites his own meditation practice as this kind of investment), on the reasoning that an AI built on top of a species that does not understand its own brain and mind is a categorically dangerous project regardless of the AI’s own capability level.
He reports being surprised by the coherence of large language model output specifically — not merely grammatical correctness but sustained logical structure across paragraphs — and connects this to a older observation of his: human belief has never required truth, only a sufficiently coherent and convincing story, which is precisely the register in which LLMs excel.

How to think [§ How to think]

Harari’s stated method for thinking about difficult problems begins with roughly two hours of daily silent (vipassana) meditation and one to two months a year on retreat, aimed at learning to observe thought and bodily sensation non-verbally rather than being carried along by the mind’s constant narrative production.
He pairs this with an explicit ‘information diet’: preferring long-form books and multi-hour interviews to short-form content, on the reasoning that the format that shapes the mind matters as much as the content.
His compositional practice is to write freely and expansively (‘like a torrent’) without early self-censorship, then delete aggressively; he treats the discipline of deleting one’s own ideas, not merely generating them, as the harder and more important skill, because attachment to an idea already on the page is what produces confirmation-biased reasoning.

Advice for young people [§ Advice for young people]

Harari’s central claim is historically specific: for the first time, no one — not even elders — can reliably predict what the world, or the job market, will look like in ten years, which breaks the traditional model of education as transmitting a stable body of skills from one generation to the next.
His recommended adaptation is treating personal identity and skill as a foldable tent rather than a stone house — building the capacity to keep learning and reinventing oneself as the core transferable skill, since which specific skills (coding, translation) will retain value is no longer knowable in advance.

Love [§ Love]

Harari connects his personal experience of growing up gay in a homophobic 1980s–90s Israeli town directly to his later scholarly interest in the power of stories: he describes years of genuine self-delusion (not concealment but an inability to consciously connect available evidence about himself), broken only when a competing narrative (that homosexuality occurs naturally across many species, and cannot therefore be ‘against nature’) became available to him.
He credits the internet, and specifically an early Israeli gay dating site, with solving a structural problem unique to diffuse minorities (unlike an inherited minority such as a Jewish family, gay individuals are typically not born into a ready-made community), and credits the feminist movement as a historical precondition that made gay liberation possible.
He explicitly rejects a self-sufficiency framing of the achievement — ‘complete autonomy… it never works’ — crediting his partner, his meditation teacher, and the earlier generation of activists who took much greater personal risk.

Mortality [§ Mortality]

Harari treats the fear of death as the root fear beneath all other fears, each of which he reads as a smaller, more manageable ‘piece’ of the otherwise unbearable totality (losing a phone standing in for losing everything).
He recounts a specific formative adolescent realisation at Israeli Memorial Day ceremonies: national-immortality narratives (‘you live on through the nation’) collapse under the simple observation that a dead person cannot hear the commemorative songs, because being dead precisely means the absence of any experience at all.
He reports that the intensity of this fear has genuinely dissipated for him over years of meditation practice, to the point that he can no longer voluntarily recapture the acuteness of the original teenage realisation.

Meaning of life [§ Meaning of life]

Harari’s closing claim is that ‘meaning of life’ as usually posed is a request for a story (a plot, a role, a script) and is for that reason the wrong kind of question, since the universe does not operate as a narrative.
His substitute frame: life, non-verbally, is simply the occurrence of feelings — wanting pleasant sensations to continue and unpleasant ones to end — and the most productive question to ask directly is not ‘what is the meaning of life’ but ‘what is suffering, and where does it come from’, investigated by direct, non-verbal observation rather than by constructing an explanatory story.