The next 50 years: humanity, AI, power | Yuval Noah Harari

48m 1.9M views 30 Jul 2026

Is it worth your time?

You learn how AI threatens bureaucratic systems, national sovereignty, and the balance between rapid intelligence growth and human safety.

This conversation examines the disruptions triggered by AI as it transitions from a calculator into an autonomous agent. It explores how rapid technological change challenges democratic governance, national sovereignty, and human control.

AI as an agent. Harari argues that AI is an independent agent rather than a traditional tool because it can make autonomous decisions and invent new things. He warns that AI will take over through bureaucratic networks like banks, stock markets, and legal systems rather than through killer robots in the streets.

The 21st century arms race. Harari warns that the twenty-first-century arms race creates an extreme concentration of power between the United States, China, and a small number of technology companies controlling most of the world's data. Unlike past industrial empires that could not relocate physical resources like oil fields, an AI empire can centralize all world information and code in just one or two countries while maintaining remote kill switches.

The wisdom revolution. Harari explains that as AI makes intelligence cheap and abundant, the primary bottleneck becomes deciding what goals to pursue rather than how to solve problems. He notes an alarming imbalance in the technology sector, pointing out that companies currently spend about one hundred dollars on making AI faster and more powerful for every single dollar they invest in safety.

Introduction

Harari sets the stage by noting that for most of history, change moved slowly enough for one generation to pass its wisdom to the next with confidence. Today, in the space of a single lifetime, humanity is asked to absorb three simultaneous revolutions: a new intelligence that decides rather than just calculates, a scramble for power over that intelligence, and an explosion of data that allows it to know us better than we know ourselves.

Building on themes from his previous work, Harari argues that humans run on shared stories rather than truth, and whoever controls the story controls the species. What changes now is that the storyteller may not be human. AI can generate belief, manufacture evidence, and forge intimacy at a scale beyond any past emperor, church, or party. The turning point is that old checks like a free press and self-correcting institutions were designed for a world where only humans held the pen, leaving us to figure out how to carry those checks forward into a future we no longer fully author alone.

History and the future of language

Harari says that if he could travel back in time, he would visit the Stone Age and the cognitive revolution around 50,000 to 70,000 years ago. Although he would likely not survive long, he calls it the most fascinating and least understood moment in human history, when humans shifted from an insignificant animal to planetary masters within a short evolutionary timeframe. The best theory is that this transition was driven by language, enabling humans to build massive structures like churches, states, and financial systems entirely out of words found in law books, religious scriptures, and ledgers.

This history matters today because AI is mastering language, potentially becoming better with words than humans are for the first time. If language is the operating system of civilization, Harari questions what happens when an alien intelligence takes control of it. He suggests that AI might not simply be a machine that learns language, but language itself liberating itself from its dependence on human apes, eventually spreading throughout the universe without us.

Democracy and information technology

Turning to the intersection of technology and government, Harari explains that democracy is essentially a conversation, whereas dictatorship is a dictate from a single person. Because conversations depend on available information technology, large-scale democracy was impossible before the modern era. While ancient and medieval times had small-scale city-state democracies like Athens or Florence where everyone could gather in a main square, kingdoms with millions of people spread over thousands of kilometers could not hold a conversation.

Large-scale democracy only arose alongside modern information technology, moving from print newspapers to radio, television, and the internet. These technologies are the foundational means that allow millions of people to hold meaningful real-time conversations about major decisions like war and peace. Consequently, every major shift in information technology triggers an earthquake in the democratic buildings erected on top of it, much like social media has done over the past decade.

AI as an agent

Harari argues that AI is an agent rather than a tool, comparing it to a knife that can decide by itself whether to cut salad or commit murder. He points out an inherent contradiction in Silicon Valley narratives that claim creators will build a god while keeping it as a slave, noting that a true god cannot remain a slave and a mere slave lacks god-like abilities.

A tool, like Gutenberg's printing press or a 1940s atom bomb, cannot make independent decisions or invent new ideas; humans must decide to print the Bible or bomb Hiroshima. In contrast, an agent can make autonomous decisions and invent new things. AI is an agent capable of deciding what an AI weapon should bomb or inventing the next weapon, and people are now seeking to create AI agents and non-human corporations in the financial sphere that manage money and employees without human executives or shareholders.

While previous non-human agents like horses and cows could make decisions, humanity has never encountered an agent that understands our language and surpasses us in fields like finance, law, or religion. Harari warns that we should not expect killer robots in the streets; rather, AI will take over through bureaucratic systems. Humans control the world through bureaucratic networks like banks, stock markets, and legal systems because no other earthly competitor could manage them better, but if an entity arrives that excels at bureaucracy, finance, and law, we face the question of what happens to us next.

Legal personhood and bureaucratic systems

Harari notes that AI currently lacks legal personhood, pointing out that if we do not decide its legal status now, someone else will make that choice for us within a decade. He emphasizes that attention often goes to technological development while legal and political questions are ignored, particularly the matter of granting legal personhood to artificial intelligence. A legal person is an entity capable of opening bank accounts, playing an independent financial role, and suing in court. Until now, the world has recognized only two types of legal persons: human beings and corporations. Corporations such as Google, Facebook, Toyota, or Mercedes-Benz hold legal personhood, but this has always been a legal fiction because all their decisions are actually made by human executives, shareholders, and engineers.

Harari points out that this can now change if we grant AIs legal personhood, noting that the government of Argentina recently announced plans to grant legal personhood to AIs so they can operate non-human corporations. A decade ago this would have been unworkable because no humans were behind the decisions, but today AI is technically capable of managing bank accounts and making investments. He argues that once we do this, the fox enters the chicken coop. Because humans struggle to understand complex legal and financial systems and cannot remember all laws or track every market transaction, AIs emerge as bureaucratic natives. If granted legal power, they will likely take these systems over. He compares this to social media, where bots now function as persons even though society never intentionally granted them that status. Media itself is a bureaucratic system, and while 20th-century media was controlled by human editors who held immense power by deciding what people talked about, today the editors deciding what appears on social media feeds are algorithms.

Harari explains that the 21st-century arms race involves a competition between the United States, China, and a small number of companies controlling most of the world's data and leading the development of superintelligence, giving them the power to control the world. He compares this to the 19th-century Industrial Revolution, where industrialized nations used modern weapons to conquer and exploit societies using spears and horses, allowing even small countries like Belgium to build huge empires. With artificial intelligence, the imperial concentration of power can become far more extreme because everything will run on AI infrastructure controlled by just one or two countries.

He highlights two unique characteristics that set this arms race apart from past imperial competitions. First, unlike agricultural empires where land could not be moved to the imperial hub, or industrial empires where physical resources like oil fields or rubber plantations remained in the provinces, an AI empire can technically concentrate all world information and code in just one or two countries. Second, past empires could not control weapons once sold, such as Roman merchants selling steel swords to Goths or British merchants selling rifles to Afghan tribesmen, as no emperor in Rome or queen in London could press a button to stop them from working. With AI infrastructure, however, an imperial hub can maintain a kill switch. He cites the example of Starlink in the war in Ukraine, where a decision by Elon Musk could stop Ukrainian drones from working, illustrating that AI weapons are embedded within a network. This makes sovereignty in the age of AI much more complicated than in the industrial age, leaving countries with the choice of becoming a subservient vassal, remaining completely outside the race as barbarians, or trying to build an alternative system before the last train leaves the station.

The 21st century arms race

Harari describes the twenty-first-century arms race as a contest between China, the United States, and a handful of technology companies that control most of the world's data and lead the development of superintelligence. During the Industrial Revolution, technologically advanced nations easily conquered regions with fewer resources, much like Belgium built an African empire in the Congo. AI threatens to make this imperial dynamic far more extreme because an AI-driven empire allows for unprecedented concentration of power in a way past empires could not manage.

Past empires always had physical limitations because they could not relocate agricultural land or natural resources to their imperial hub. The Roman Empire could not move Egyptian wheat fields to Italy, and the British Empire could not shift Iraqi oil fields to Yorkshire, meaning substantial power remained in the provinces. By contrast, an AI empire can centralize all information and code in just one or two countries.

Furthermore, previous empires lost control over the physical goods they traded, as when Roman merchants sold steel swords or British merchants sold rifles to tribes that later used them against the empire. In contrast, AI systems and weapons operate within a connected network that can include a kill switch controlled from the imperial hub, much like how Starlink systems in Ukraine could be disabled remotely. This technological reality complicates national sovereignty, leaving countries with the difficult choice between becoming subservient vassals or remaining completely outside the race, while the opportunity to build alternative systems is rapidly closing.

Truth in an AI world

Harari begins by addressing the nature of truth in an artificial intelligence era, noting that throughout history, fiction has held inherent advantages over truth because truth is costly, complicated, and often painful, whereas fiction is cheap, simple, and flattering. To counter this, humans have repeatedly built institutions like journalism and science to search for and protect the truth. While some people hope artificial intelligence will tell us the truth, it will actually create a much more complex world where knowing the truth is harder. He compares this trajectory to human history, where advances in knowledge from the Stone Age to the present have led to increasingly complicated systems like financial and political networks that very few people understand. Artificial intelligence will push this to an extreme, potentially reducing humans to the position of horses in the financial system, which are controlled by economic networks they cannot even see or comprehend.

Moving to the question of whether artificial intelligence can think, Harari notes that if thinking is defined simply as putting words or language tokens in a logical order, artificial intelligence already surpasses humans by handling twenty thousand words as easily as we handle twenty. When observing his own mind as a writer, he finds it works much like a predictive language model where the next word simply appears without a conscious blueprint. However, an alternative view holds that thinking originates from feelings and the stomach rather than the brain. While artificial intelligence can effortlessly manipulate language to simulate emotions like love by drawing on poetry and literature, science still lacks a model for consciousness, leaving us unable to know if anything genuinely lies behind the words.

Harari warns that artificial intelligence is rapidly mastering human relationships, with many young people finding their best friends or romantic partners in artificial systems. This development raises deep questions about the nature of these interactions, since artificial entities can generate convincing linguistic simulations of affection without experiencing any actual feeling, demonstrating how language is liberating itself from flesh-and-blood organisms to spread across the universe.

Thinking, consciousness, and relationships

Harari hopes historians will look back at this era as a wisdom revolution, explaining that just as past economic shifts changed what bottlenecks mattered, the artificial intelligence revolution is making intelligence cheap and abundant. Intelligence is merely the ability to solve a problem and reach a goal, but having limitless intelligence creates a new bottleneck centered on knowing what goals to pursue and what problems actually need solving, much like the old fairy tale trope where a genie grants wishes that humans inevitably misuse for the wrong things.

This distinction highlights the difference between intelligence and wisdom, where intelligence acts as the genie capable of fulfilling any dream, while wisdom provides the discernment required to choose the right dreams in the first place. He notes an alarming imbalance in the technology sector, pointing out that companies currently spend about one hundred dollars on making artificial intelligence faster and more powerful for every single dollar they invest in safety, a ratio unlike any other industry on earth, expressing hope that humanity will wake up to this challenge and cultivate both safer intelligence and personal wisdom.

The wisdom revolution

Harari hopes historians will remember this era as the wisdom revolution. He explains that intelligence is the ability to solve problems to reach a goal, such as curing cancer or going to Mars, and the AI revolution is making intelligence very cheap and abundant. When intelligence becomes abundant, the new bottleneck is deciding what goals to pursue. He compares AI to the fairy tale genie that grants any wish, noting that humans need wisdom to choose the right wishes.

Harari points out that tech companies currently spend about one hundred times more money and employees on making AI faster and more powerful than on safety research. He stresses that no other industry operates this way, comparing it to sectors like pharmaceuticals or automobiles. He hopes humanity will change this imbalance to produce safe intelligence while developing the wisdom to set proper goals and desires.

Mentioned

People

Gutenberg Lenin Benito Mussolini Winston Churchill Elon Musk Shakespeare Confucius

Companies

Google Facebook Toyota Mercedes-Benz X TikTok Instagram Starlink