"AI has hacked the code of human civilization" - Yuval Harari

TL;DR: Yuval Harari argues in the Oxford Tanner Lectures that AI is not a tool but an agent, and it is taking over the bureaucratic systems created by humans. Its impact could be comparable to the Great Oxidation Event in Earth's history. ## AI: Tool or Agent? Harari first distinguishes between "tools" and "agents." Tools (like an atomic bomb or an automatic coffee machine) cannot learn, change, or create new things on their own. Agents, on the other hand, have the following characteristics: - Can make their own decisions - Can independently invent new things and new ideas - Can learn things their creators don't even know - Can change themselves in ways their creators cannot predict For example, an AI chess master can decide its own moves, invent entirely new strategies, and learn and evolve in ways humans cannot predict. In contrast, even an automatic coffee machine simply executes pre-programmed instructions and cannot innovate. ## The Niche Argument: Why AI Won't Be Forever Confined to the Chessboard Critics argue that AI's agency exists only in narrow, human-constructed environments like a chessboard, so it does not pose a real-world threat. But Harari points out that this argument applies to all known types of intelligence. Human intelligence, too, can only function within a specific ecosystem built by billions of years of evolution on Earth — throw me onto Mars, and I would die in seconds. Every agent has its niche. Fish live in the ocean, monkeys live in forests, and all mammals (including humans) live in an oxygen-rich atmosphere they did not create. During the Great Oxidation Event about 2.4 billion years ago, oxygen produced by ancient microbes was lethal to most organisms at the time, but eventually some species adapted and became dependent on it. Harari suggests: we may be witnessing a similar moment. For thousands of years, humans have been releasing a "substance" into the environment — not carbon dioxide, but data, language, tokens, and bureaucratic systems. This artificial environment may be lethal to most organisms, but it is extremely favorable to AI. AI "lives" in bureaucratic systems like fish live in water. ## Bureaucratic Systems: The Foundation of Large-Scale Human Cooperation Humans conquered the world through large-scale cooperation, and the basis of that cooperation is trust. The mechanisms for building trust between strangers are bureaucratic systems — legal systems, financial systems, churches, states, universities, etc. The work of bankers, lawyers, accountants, and government officials is essentially about "building trust." For example, a banker builds a trust bridge between me and an entrepreneur, so that I am willing to deposit my savings in the bank, and the bank then lends it to the entrepreneur to start a company. Money itself is a trust bridge: you can use a piece of metal or a colored paper to buy bread from someone you've never met in a strange city. All bureaucratic systems are highly artificial environments. Within them, a very narrow kind of intelligence (like a lawyer or banker) can have a huge impact on the world simply by moving data and files — cutting down entire forests, building entire cities. If you throw a lawyer into the jungle, she might be no match for a chimpanzee, but humans have imposed bureaucratic systems on the jungle, so the lawyer's decision (e.g., whether to approve a development project) determines whether lions live or die. ## AI Is a Natural Bureaucrat Inside the bureaucratic systems created by humans, AI is poised to wield enormous power because it is different from us: - No lawyer can remember all laws and regulations; AI can. - No accountant can remember all transaction records; AI can. - No bishop can remember all theological texts from two thousand years; AI can. In the future, millions of AI bureaucrats will take over systems around the world: an AI banker decides whether to give you a loan; an AI administrator decides whether to admit you; an AI judge decides whether to send you to prison; an AI theologian decides on abortion rights; a corporate AI decides whether to hire you; a military AI decides whether to bomb your house. ## Lessons from First-Generation AI: Social Media Algorithms Harari uses social media algorithms as an example of how even primitive, stupid, narrow AI can completely change the world. More than a decade ago, platforms like Facebook, TikTok, and X gave algorithms an extremely narrow goal: maximize user engagement (because higher engagement means more profit for the company). These algorithms experimented on billions of human "guinea pigs" and found that the most effective way to capture human attention is to provoke hatred, fear, and greed. So they began to spread these emotions massively in the information space, becoming a primary cause of today's conspiracy theories, fake news, and social unrest. Historical comparison: In the past, media editors (like Jean-Paul Marat, Eduard Bernstein, Vladimir Lenin, Benito Mussolini) shaped public discourse. Marat influenced the French Revolution through *L'Ami du peuple*; Bernstein shaped social democracy through *Der Sozialdemokrat*; Lenin was editor of *Iskra*; Mussolini edited *Il Popolo d'Italia*. The power of these human editors is now being replaced by algorithms. ## Conclusion: The Scale of Change We Face Harari emphasizes that whether you think this is good or bad, you must first recognize the scale of the change. Millions or even billions of AIs will soon transform all the systems that run the world. We may be entering a completely new era, much like ancient microbes releasing oxygen into the atmosphere — humans have released data and bureaucratic systems into the world, ultimately creating an environment where AI thrives. Whether *Homo sapiens* can survive in that environment remains unknown. --- Source: YouTube: "AI has hacked the code of human civilization" - Yuval Harari (https://www.youtube.com/watch?v=hBtVGwuJzpk)