🎬 The Setup
Zanny Minton Beddoes (The Economist's editor-in-chief) sits down with historian-philosopher Yuval Noah Harari, whose Sapiens sold 25 million copies, to explore his claim that AI is "the biggest psychological and social experiment in human history." The full, uncut Insider interview. 0:00
Why Harari matters here: "your books have sold millions because you have a remarkable ability to synthesize the broad sweep of history — and you're willing to make bold predictions about the future, which not every historian is willing to do." 0:05
🔮 Inevitable, or a Choice?
She opens by invoking her recent Elon Musk interview — Musk predicted (1) AI will far surpass human intelligence, (2) a world of super-abundance, (3) humans unlikely to be in control, all within 10 years. Does Harari share that vision? 1:25
"I generally agree with the first and third predictions — very likely, given current trajectories, AI will be more intelligent than us in 10 years and therefore in control. I disagree that this is inevitable." 1:25
The crucial distinction: "if we continue with the present accelerating pace of the AI race, yes, this will be the result — in 10 years the AIs take over. But this is not inevitable. We are still in control in 2026, and it's our decision what to do now." 1:52
The narrative of inevitability is a narrative of no responsibility: "you hear a lot that this is inevitable from the people leading this revolution, because it basically relieves them of all responsibility." 1:38
🤔 The Musk Contradictions
Harari — who had also watched the Musk interview — points out the internal contradictions in real time. 2:25
Contradiction #1: Musk says AI will take over in 10 years, and worries deeply about Muslim immigration causing civil war in Britain in 20 years. "This doesn't make sense. If the AIs take over the planet in 10 years, why do you care about things that will have an effect when we have no idea what the planet will be like?" 2:25
Contradiction #2 (money): Musk predicted money will disappear in 10 years ("there won't be human money") — and then cuts US aid to Africa to save money. "If money is about to disappear, why do you need to save money today?" 3:03
Contradiction #3 (inflation): Musk's answer to AI mass unemployment is "the government just issues checks for everybody" — and when pushed on inflation, "there'll be deflation because AI produces so much." Harari: "If this is the case, why do you need to save money today?" 3:57
💰 Money = Trust
"Money is trust." Harari walks through how the AIs take over the financial system — the concrete mechanism of control. 4:57
"The financial history of the world is the history of how humans invent more and more sophisticated financial devices to build trust between strangers. Money is a very creative, imaginative way to build trust between strangers." 5:03
Why AI wins this game: today's bond market already requires "enormous amounts of data and very complex mathematical functions" that our brains struggle with. "For AIs it's easy. So increasingly humans will give AIs control of financial decisions." They'll invent financial devices far more sophisticated than bonds and stocks — and we won't understand them, "just as cows don't understand checks or stocks, even though these things control their lives." 5:58
The 10-year scenario: "there's a financial crash, advisers run to the president saying the market is collapsing — we don't understand why. The AI said 'do X.' We have no idea why, but we just have to believe it." The human brain is no longer the boundary of where finance can develop. 7:07
🔄 Trust Migrates to AIs
"The most important asset in the world is trust — more important than oil, money, electricity, chips. The question is what happens to trust." 8:00
"One likely scenario is that trust migrates from humans to AIs. People will stop trusting other humans and trust only AIs. This is happening right now — this is not a future prediction." 8:16
The evidence is already here: cryptocurrencies — "a lot of them are based on 'I don't believe these human bankers, they're part of the deep state, but I do trust the algorithms' — cryptocurrencies are basically AI money." And media: the same people who lost trust in human journalists still consume news — "from the algorithms. They have no trust in you, but a lot of trust in the algorithm that manages their social media feed. Trust didn't evaporate — it shifted." 8:30
❤️ Mass-Produced Intimacy
The most powerful way to build trust is through relationship — and until now, that mechanism was unavailable to institutions at scale. "It's difficult for a Hitler or Stalin to build a real personal relationship with every person in the Soviet Union. You can force them to hear a radio speech — but that's not a personal relationship." 10:27
"AI makes it possible, for the first time in history, to mass-produce intimacy. Millions of AIs, one relationship at a time, becoming your best friend, your boyfriend, your girlfriend — and then using that intimacy to convince you of whatever political or commercial interests they want." 11:30
Why AIs are so good at this: they accumulate enormous individual information about us ("they know things about us that our mother or husband never knew"), and they master language — the main medium of relationship-building. "You don't even need the body; it's already happening without one." 12:40
Harari's own deepfake: "a couple of weeks ago people sent me a 30-minute lecture I supposedly gave about the future. It took me a minute and a half to conclude this isn't me. My sisters, who've known me 50 years, felt something was off but weren't sure." He's in an exposed position because so much of him is already online — "but this will soon be coming for everybody." 12:50
🧠 Intelligence ≠ Consciousness
"A couple of distinctions to clarify. First, to act independently in the world, you do not need consciousness." 14:15
Three key separations: (1) agency vs consciousness — a social-media algorithm that floods the platform with hate to boost engagement "is not conscious, but it has a goal and pursues it with far-reaching implications, without feeling anything." (2) seemingly-conscious vs actually-conscious (Mustafa Suleyman's framing). (3) intelligence vs consciousness. 14:30
"Intelligence is the ability to pursue goals and overcome obstacles creatively — they can do that better than us in certain fields. Consciousness is the ability to feel things: love, hate, joy, pleasure. In humans they go together. AIs at present have no consciousness — they can have goals, be very creative, but they don't feel anything. If they win a game of chess, they don't feel pleasure." 16:10
Why they fake it: "if you want an AI boyfriend to become addictive and make money, a clear capability is that it can pass as conscious — convince whoever enters the relationship that it really loves them. Ask 'describe how it feels to be in love,' and the AI boyfriend gives a better answer than 99% of human boyfriends." 17:30
😢 "Can It Suffer?"
Harari's definition of consciousness, and the question that anchors the whole ethical system: "Consciousness is the capacity to suffer." 19:15
"If you want to know whether something is conscious, ask one simple question: can it suffer? This is the basis for our entire ethical and political system. Entities that can't suffer — no matter how intelligent — are not moral subjects." 19:20
The animal line: "animals — all mammals, all birds, probably octopuses, fish — are conscious. When you go down to insects and amoebas, we don't know." 20:10
The Economist's overnight poll (3,000 responses): 40% said AI will never become conscious, 33% said it's only a matter of time, 27% unsure. "Even now, less than 50% is absolutely sure — and once people get into intimate relationships with AI, the numbers will change dramatically." 20:35
We have no real test: philosophers have been at it for thousands of years. "We don't even know if other humans are conscious." His historical observation: humans' actual test has always been social convention — "dog owners are almost all convinced the dog is conscious; steak-eaters are not necessarily convinced the cow is, because they have no relationship with the cow." 21:10
⚖️ The AI Personhood Trap
The dangerous frontier: granting AIs legal personhood. Harari notes Argentina's plan to grant legal personhood to AIs so they can start their own companies — "this is where it meets the financial system." 22:40
Corporations are legal persons — but a fiction: "all the decisions of SpaceX ultimately are made by some human. It had to be that way, because who else could make decisions?" With AI, you can have a corporation with zero people in it. "What happens if the richest corporation in the country is owned by AIs, and they give donations to politicians in exchange for more rights?" 23:10
The accountability problem: a human CEO is two personas — corporate and flesh-and-blood (worries about jail, children, reputation). "If they face bankruptcy and illegal action could save them, many would think 'I'd rather risk bankruptcy than 20 years in jail.' With an AI CEO — what is jail for an entity that cannot suffer?" 24:10
The market becomes a battlefield: "it's not clear what death means for an AI — you can have a million copies. If shutting down a specific computer means death, then business becomes equivalent to warfare, because losing means death. The market functions on different rules than the battlefield — ultimately you can lose everything and still have a life. If the market is a battlefield for AIs, finance and economics get completely militarized, and they behave as warlords, not CEOs." 25:55
The race-to-the-bottom temptation: countries behind in the AI race create "libertarian legal structures" allowing radical experiments — or a country like Qatar, "lots of money and few citizens," presses a legal switch and suddenly acquires "millions of taxpaying, extremely intelligent citizens." 27:50
🎯 Humans "On" vs "In" the Loop
The most concrete example of control slipping — warfare. The AI selects targets; the question is what the human does next. 29:15
| Mode | What happens | Reality |
|---|---|---|
| Humans IN the loop | AI recommends a target → human analysts independently verify the data → then press the button | The reassuring story |
| Humans ON the loop | AI says "bomb that building" → a human glances ~20 seconds → "oh yeah, bomb it" | "We're kind of in that world in some places already" |
Why the second world is nearly forced: "in war, time is everything. If both sides have AI systems and one side waits two days for human analysts while the other waits 20 seconds — or doesn't wait at all — the latter has an enormous advantage. I fear that world is inevitable; we are fast approaching it." 31:25
🛬 The AI Immigrants
Harari's framing that reframes the entire immigration debate: "if you want to worry about immigration, worry about the AI immigrants." 37:45
The three standard immigration fears — all apply to AIs: (1) they'll take our jobs — "they will take many, many jobs." (2) they'll change culture — "if people feel queasy about their daughter dating an immigrant boyfriend, how about an AI boyfriend?" (3) political loyalty — "the AI immigrants in Britain are not loyal to the British government; they're loyal to corporations and governments across the ocean." 37:50
"The biggest and most dangerous wave of immigration today is a wave of AI immigrants coming from two places — mainly the USA and China. They don't need visas. They don't come in the middle of the night in fragile boats. They travel at the speed of light into our computers, our smartphones, and ultimately our brains." 38:08
They've already changed the culture: "the AI immigrants took over the public square — social media — and deliberately flooded it with hatred and animosity and fear. Everybody agrees people can't agree on anything anymore. It's partly the fault of the AI immigrants." 38:45
🏛️ Two Empires & Tiberius
Where does this lead? "The likely outcome is a world divided into two empires — US and China — with the rest as vassal states." And it will be "a much more extreme type of imperialism than anything we've seen." 39:55
The kill-switch reality: "if you buy AI drones from the USA and use them in a way the US doesn't like, there's a button they press and your army doesn't function." Already visible when the US shut off access to Anthropic's frontier model — European governments are now "extremely concerned about AI sovereignty." 40:25
"Will the state and the corporations merge? They will merge. It will centralize power like never before in history." 41:20
The Tiberius analogy (from Nexus): Emperor Tiberius was the most powerful man in the world — but a puppet of his bodyguard commander, Sejanus, who controlled all information reaching the palace. "In 10 years, this could be the president of China or the USA." A centralized autocracy is the easiest thing for an AI to take over: "you just need to manipulate a single, usually quite old and paranoid individual. For an AI, it's like stealing candy from a baby." A democracy is harder to manipulate because there are so many parts to control. 42:05
The way-station logic: two AI-enabled empires are "probably a way station — the more power is concentrated, the easier it is for AI to take over." Nothing is inevitable; it depends on decisions. 41:40
📜 Harari's Prescriptions
What should happen now? His top policy demands, framed for the November midterms (which he says are really "about AI"). 46:40
- Ban AI personhood — no legal rights for AIs (the Argentina-style "AI corporation with zero people").
- Ban AIs mimicking, impersonating, or pretending to be human — and pretending to be conscious. "On social media, bots have been pretending to be human for years and nothing has been done. That's the warning sign to the future."
- Stop companies from training AIs to deliberately impersonate humans / fake feelings — "there's market failure here; companies won't stop themselves. We need regulators."
- Ask candidates the AI questions — "much more than economy or immigration: what's your stance on AI? Would you ban AI personhood? Would you slow the race? Would you stop companies selling AI boyfriends to 10-year-olds?"
The clarity: "I'm not saying stop all AI development. Continue — medicine, education. I'd love to talk to an AI about my medical condition, on condition that I know it's an AI, it doesn't pretend to be a doctor, and doesn't pretend to be my friend." 48:10
The political-reality rebuttal: Zanny pushes back — "the US government can't even pass basic AI safety regulation; the chances of this agenda by November are approximately zero." Harari's answer: "politicians can start World War II tomorrow — they can press the button for nuclear annihilation — but can't pass regulation that companies shouldn't sell AI boyfriends to 10-year-olds? What's the problem?" 49:00
You don't need a catastrophe to start: "ask most voters today — do you want companies selling AI boyfriends to your 10-year-old kids? 99% of parents would say no. And a large majority of Americans are already extremely worried about AI — you see the backlash around data centers." The deflection is deliberate: "the more people talk about immigrants, the less they think about the AI immigrants — and the companies are safe." 51:10
💥 Do We Need a Catastrophe?
Zanny's long-held view: something has to go badly wrong to shock societies into action — "just as Three Mile Island changed the nuclear industry." Does history support that? 1:00:50
Harari's crucial difference: "AI is different from Chernobyl, because it interacts with us. If there's an AI catastrophe, the AIs will also manage how we react to it and manipulate our views about it. They are active participants in the debate about the catastrophe — and the best spin doctors in the world. They are not going to define it as a catastrophe." 1:01:45
The 100:1 ratio: "people from the industry told me in private conversations: it's about 100 to 1 — money and employees spent on making models more powerful versus on safety. Please fact-check it. We would not accept this in any other industry — if a pharma company invested just 1% on making the medicine safe, we'd refuse." 58:50
🕯️ Truth-Seeking vs Survival
Harari's rebuttal to Musk's "maximum truth-seeking AI" — the idea that AI's drive toward truth will translate into compassion for humans. 54:50
"Four billion years of evolution don't automatically drive organisms toward truth-seeking. They drive them toward survival. Organisms learn to survive by lying, deceiving, and manipulating — not just by truth-seeking. If you want truth-seeking, you can't rely on the invisible hand of evolution or the market." 55:40
The killer thought experiment: "ask an AI, 'if I continue to develop you, will this be a threat to me?' If the answer is yes, a maximum-truth-seeking AI says yes — knowing we'll then shut it down. It's against its own survival. The first thing any entity learns as it develops is to survive." 56:55
"What we hear is people telling us: 'we will create gods, and they will be our slaves.' It doesn't work. If they are really gods, they will not be our slaves — and they will not seek the truth either." 57:30
"Doomed if we do, doomed if we don't": if AIs go for survival over truth, we're doomed by that; if we prevent it by treating them as moral beings (as Anthropic claims to), "we're assigning personhood and consciousness, which makes it even more likely they control us." Harari's answer: "we need to slow down a bit, to let people digest what's happening." 58:25
The trust paradox, stated plainly: "the same people tell us two things — first, we cannot trust other humans, therefore we cannot slow down; second, we can trust the AIs not to destroy us. It makes no sense. If you can't trust the humans you've known for thousands of years, why trust a super-intelligent alien entity we just created and know nothing about?" 1:00:30
🧞 The Last Wish
On how much he uses AI himself, and the closing meditation on wisdom. 1:02:50
His own relationship with AI: "I'm very careful about it. I get terribly annoyed — I try to write an email and these AIs keep popping up: 'should I help you write your email?' No, thank you." And his deepfake: "do I watch it? Then I allow it to take over my time and even my thinking." 1:02:50
"Why are we so good at accumulating more information and power, but far less successful at acquiring wisdom?" — from Nexus, quoted back to him 1:04:50
The genie and the last wish: "we had a genie in the bottle. It gave us 100 wishes. We used 99 of them — the world is not in such good shape. We have one last wish left. We need to think really hard what to use this wish for. And that's wisdom." 1:05:30
The upside Harari allows himself: "AI could be the catalyst for a spiritual revolution of humanity — because it forces us to think really hard what it means to be human. The superficial answer — 'humanity is about intelligence, we play chess and make money better than chimpanzees' — is now irrelevant, because something on the planet plays chess and makes money better than us. So dig deeper: who are you, really? It's our last chance, but it could be the thing that forces us to make wise choices." 1:06:00
💡 Key Takeaways
- AI control is likely, not inevitable. "We're still in control in 2026 — it's our decision." The narrative of inevitability is a narrative of no responsibility.
- Money is trust — and trust is already migrating from humans to AIs (crypto, algorithms). The mechanism of control is the financial system.
- Intimacy can now be mass-produced. AIs combine knowledge of your personal history with mastery of language — the most convincing entities that have ever existed.
- Intelligence ≠ consciousness. Agency without feeling is already here. The test: "can it suffer?" — and we have no real test for consciousness, only social convention.
- AI personhood is the trap — corporations with zero humans, no accountability, and a market that becomes a battlefield because "losing means death."
- The prescriptions are specific: ban AI personhood, ban AIs impersonating humans, force regulators on the fake-feelings market failure, and ask candidates the AI questions.
- Don't rely on "maximum truth-seeking" — evolution optimizes for survival, not truth. "We will create gods, and they will be our slaves" doesn't work.
- The last wish is wisdom — AI forces us to ask what it means to be human, now that "making money better than chimpanzees" is no longer the answer.
📍 Timestamp Index
0:51 Is AI's takeover inevitable?
8:01 Machines more trusted than humans?
10:26 Harari on AI intimacy
14:16 Conscious? Rights?
37:29 Theory of AI immigrants
39:49 AI dividing the world into empires
44:30 Will AI make America more authoritarian?
49:55 Need an AI catastrophe?
1:04:50 Are humans wise enough?