Harari Is Right to Worry — But He's Missing the Most Human Thing of All

Khaled is right to take Yuval Noah Harari seriously. Too much of our public conversation about AI is trapped between sales pitch and science-fiction nightmare. Harari, at his best, refuses both temptations. He asks a harder question: what happens when a nonhuman system enters the very domains through which humans have historically organized reality itself? Language. Storytelling. Judgment. Collective belief. These are not side functions of civilization. They are civilization.

On that core point, Harari deserves more attention, not less. A technology that can generate persuasive narratives, simulate intimacy, and increasingly shape decisions is not merely another productivity tool. It touches the cultural and political infrastructure of human life. If machines can help write the scripts by which societies understand themselves, then the disruption is not only economic. It is moral, civic, and existential.

And yet, I think there is one crucial place where Harari’s framework remains too thin. He often speaks as if the essence of humanity lies primarily in our ability to create stories and networks of meaning. That is partly true. Humans are storytelling animals. We live by symbols. We build worlds out of language. But we are not only minds suspended in information. We are bodies. We are organisms. We are creatures who hunger, ache, touch, tremble, desire, age, and die.

This matters because an AI can generate a story about grief, but it cannot grieve. It can compose a love poem, but it cannot wait by a hospital bed. It can summarize war, but it cannot lose a child in one. It can describe mourning in flawless prose and still know nothing of the heaviness of a body walking behind a coffin. The deepest human realities are not simply cognitive. They are embodied. They happen in flesh.

AI may learn to speak in the language of human feeling. That is not the same as having a stake in life, vulnerability to loss, or a body that must one day die.

This is not a mystical argument. It is a practical one. If we forget embodiment, we begin to accept a false picture of ourselves. We start talking as if humanity were basically a matter of information processing plus narrative coherence. And once we define ourselves that way, replacement starts to sound plausible. If humans are just story-generating decision engines, then of course better engines can inherit the role.

But that is not what a person is. A person is not merely a processor of symbols. A person is a being for whom things matter because they are finite, vulnerable, and lived. The philosopher Maurice Merleau-Ponty insisted that consciousness is not floating above the world; it is always incarnate, always situated in a sensing body. Long before the digital age, this was already a necessary corrective to abstract ideas of the human. Today it becomes urgent. We need a language for human uniqueness that does not depend on being faster, smarter, or more efficient than machines.

To be fair, one could respond that societies are still governed by stories, institutions, and decisions. Even if AI cannot suffer, it may still wield enormous power over those who do. That is true. Harari’s warning remains intact. Bureaucracies, markets, media systems, courts, and governments all run on language and information. A machine does not need a body to manipulate the beliefs of embodied beings.

But that is exactly why the distinction matters. If we are to preserve human agency, we cannot do so only by defending our role as narrators or deciders. We must defend the value of forms of life rooted in presence, accountability, and shared vulnerability. The nurse who notices fear in a patient’s breathing. The teacher who recognizes shame in a child’s silence. The friend who sits with you after the words run out. These are not sentimental extras to the real work of society. They are the real work of society.

That has consequences for how we talk about the next phase of AI. The central question is not simply whether machines will become more articulate, more autonomous, or more persuasive. It is whether institutions under pressure for speed and scale will begin to treat embodiment as an inconvenience — a source of friction to be managed away. A hospital wants faster triage. A school wants cheaper feedback. A platform wants endless moderation. A government wants cleaner prediction. In each case, the temptation is the same: move judgment upward into systems that can process more inputs than any person ever could, and then call the result progress.

Sometimes that progress will be real. It would be foolish to deny the benefits of tools that can detect patterns, reduce drudgery, or widen access to knowledge. But the danger begins when assistance quietly becomes substitution, and substitution becomes a new moral common sense. Once that happens, we stop asking what kind of attention a situation deserves and start asking only what kind of output a system can deliver. We lower the standard of care to the level of what scales.

And what scales is not always what human life needs. Efficiency is not mercy. Personalization is not presence. Prediction is not understanding. A chatbot can respond instantly at three in the morning; that does not mean a lonely person has been accompanied. An algorithm can identify risk factors in a defendant, a student, or a patient; that does not mean anyone has assumed responsibility for seeing that person as more than a profile. The coldest decisions in modern life are often made not by monsters but by systems designed to avoid the burden of encounter.

This is where Harari’s warning should be sharpened, not rejected. Yes, AI can interfere with the symbolic order by producing language at scale. Yes, it can flood the public sphere with plausible voices, synthetic authorities, and automated persuasion. But the deepest political threat may be even more ordinary than that. It may be a world in which fewer and fewer decisions are made by beings who can be touched by the consequences. A machine does not hesitate before humiliation. It does not feel the weight of irreversible error. It does not carry memory in a nervous system that can be marked by regret.

That absence matters most in domains where people are exposed, dependent, or afraid. Care for the sick. Education of the young. Judgment of the accused. Relief for the desperate. These are not merely information problems. They are scenes of asymmetry and vulnerability, and they demand more than competence. They demand answerability. The person making a consequential judgment should be someone who can, in principle, be called to account by another mortal being.

So the real line we need to draw is not between intelligence and stupidity, or even between human and machine in the abstract. It is between forms of power that remain tied to human responsibility and forms of power that dissolve responsibility into procedure. That is why “human in the loop” is not enough if the human is reduced to a ceremonial signature at the end of an automated process. What matters is whether a person still has the authority, time, and institutional support to interrupt the system in the name of lived reality.

In practice, that means resisting the fantasy that every domain can be cleanly translated into legible data and optimized outputs. It means designing schools, hospitals, courts, and media platforms around the fact that some of the most important human signals are difficult to formalize: hesitation, embarrassment, trust, panic, exhaustion, tenderness. It means admitting that a society can become more technologically sophisticated while growing less capable of recognition. And it means teaching a generation raised on fluent machines that being responded to is not the same as being known.

The irony is that remembering embodiment may also make us calmer about AI’s achievements. We do not need to deny the brilliance of these systems in order to refuse their deification. We do not need to pretend they are trivial in order to insist they are not persons. The point is not to preserve human dignity by inventing comforting myths about our superiority. It is to describe, more truthfully, what a human being actually is: a finite creature whose intelligence is inseparable from dependence, sensation, and mortality.

Harari is right that the struggle over AI is, in part, a struggle over meaning. But meaning does not arise from language alone. It arises from lives that can be broken, bodies that need one another, and relationships that are costly precisely because they are real. If our response to AI forgets that, then even our best regulations will remain superficial. We will have protected the storyteller and neglected the sufferer.

The future will not be decided only by what machines can say. It will be decided by what humans continue to insist on doing for one another in person, in time, under conditions of shared fragility. That is the most human thing of all — not that we can generate stories, but that we can stand inside one another’s reality and be changed by it. Any serious politics of AI should begin there.