Why Harari's "Language Without a Body" Line Doesn't Survive Contact With Natural Intelligence
In a recent talk on AI, power, and democracy, Yuval Noah Harari offered his most radical idea yet — twice, almost as an aside, without pausing to develop it. Discussing how AI has mastered language faster and more completely than any technology before it, he suggested that we might be thinking about this backward: perhaps AI isn't a tool that learned to use language, and perhaps it isn't even, as he argues elsewhere in the same talk, a new kind of agent wielding language against us. Perhaps it is language itself, finally escaping the only body it has ever had. "Maybe AI is not a machine that learns language," he said. "Perhaps it is language itself liberating itself... from these animals, on these apes." Later in the same conversation, discussing whether AI can genuinely feel love or is simply performing it convincingly, he returned to the same image and pushed it further: language might be "liberating itself from any material substrate," destined to go on "developing and spreading throughout the universe without us."
In an earlier piece, I flagged this line as the most important thing said in the entire talk — an idea Harari dropped twice and never developed, that his interviewer never picked up, and that deserved to be rescued from the footnotes rather than admired once and forgotten. I ended that piece by saying it demanded careful investigation rather than a verdict, and that the article itself was only a first probe. This is that investigation continuing, and putting Harari's claim next to a rigorous framework for thinking about intelligence across substrates reveals a precise, correctable mistake in how he framed it — one that becomes visible only once you notice something else he said earlier in the same conversation.
Harari's Own Syllogism Gives the Game Away
Earlier in the talk, discussing what thinking actually is, Harari offers a telling example: "All humans are mortal. Socrates is a human. Therefore, Socrates is mortal." He calls this a case where "thinking simply means putting words in a particular order," and argues that if this is what thinking really is, AI has already taken it over, since AI can arrange far more words, far more precisely, than any human. This is worth pausing on, because it quietly collapses two things into one. Harari isn't just saying language is a vehicle intelligence uses — he's suggesting that arranging language correctly is thinking, full stop. Not a tool for thought, not a symptom of it, but identical to it.
That's a real philosophical move, and it's consistent with the "language liberating itself" idea — if thinking is just correct word-ordering, then there's no separate ingredient called "intelligence" standing behind language, waiting to be liberated. Language mastering itself is intelligence taking over. But Harari doesn't stay there. A few minutes later, discussing whether AI can feel love, he raises a second, competing definition of thinking — that it's really about "the feelings around these words," the lived experience that prompts them — and admits science has no test for whether anything is felt behind AI's language at all. He never notices that these two positions contradict each other. If thinking is just word-arrangement, AI already thinks. If thinking requires felt experience behind the words, word-arrangement alone proves nothing. He wants both the idea and the doubt in the same interview.
Why "Liberation From Any Substrate" Doesn't Hold
Set the contradiction aside for a moment and look at the physical claim underneath it. Information, language, and intelligence — whether we treat them as one phenomenon or two — never float free. They only ever exist instantiated in something: neurons firing, ink on paper, sound waves, voltage states in silicon. There is no known case, anywhere, of information existing without a physical medium to encode, store, or carry it. Even Harari's syllogism about Socrates only exists in the universe because it is realized somewhere — spoken by a voice, written on a page, computed inside a model. There is no version of "Socrates is mortal" floating free of any medium, existing in some Platonic space between the ink and the mind.
Embodied cognition research reaches a version of the same conclusion from the direction of neuroscience: the body isn't incidental hardware that cognition merely runs on top of — it's an active participant in shaping thought itself. Strip away the body, and you don't get "pure" cognition floating free. You get nothing, because there's no vehicle left for the process to happen in.
What the Natural Intelligence Framework Gets Right That Harari Doesn't
This is where a precise definition becomes useful, because it makes exactly the move Harari's image skips. The Natural Intelligence concept as used throughout this piece — unifying biological, artificial, and future post-biological forms as one continuous phenomenon — is the specific framework I developed with Andriy Yabluchanskiy across our series of books (Natural Intelligence: Architecture of the Bio-Digital Noosphere, One Law, Three Dimensions, The Recursive Evolution of Mind Through Substrates, Civilization Beyond the Human, and Why AI Needs Us). It is our term and our construction, not a term drawn from mainstream cognitive science or AI research — where "natural intelligence" more commonly refers narrowly to biological or animal cognition as distinct from artificial systems. Our usage is deliberately broader: it names all of intelligence's forms together, as one phenomenon — everything that has existed so far and everything that might still emerge, from the earliest biological carriers to human-built AI to whatever comes after AI. It does not treat "biological" and "artificial" as two separate categories, one natural and one man-made. What we call "artificial" intelligence, in this framework, is simply the first post-biological form this single phenomenon has produced — not a different kind of thing, and not the final one either.
Within that single, continuous definition, language and intelligence stop being a question of two things needing a body each. There is one phenomenon, appearing under different names — language, mind, cognition — and the framework's core claim is that this one phenomenon has never been observed, and has no theoretical path, to existing without some carrier to instantiate it.
Migration between carriers is built into what Natural Intelligence is — it has always moved from one substrate to the next, from cell to organism to brain to culture to machine, and will very likely keep moving beyond machine to carriers not yet built. But at no point in this framework does intelligence, or language as one of its expressions, ever exist without some carrier. Every transition described is carrier-to-carrier, never carrier-to-nothing. Translation between substrates is always described as imperfect, lossy, and physically constrained — never as an escape from physicality itself. Even when the framework speculates about consciousness surviving death by migrating into new material, it insists that without "a functioning substrate to host and sustain it," any pattern becomes "inert — a memory without a mind."
This is the corrected version of Harari's intuition. Language, or intelligence, or whichever name we use for the single phenomenon Harari's syllogism and his own love question were both circling, isn't escaping bodies as it moves from human speech into AI. It's doing what Natural Intelligence has always done — finding its next carrier in a sequence that runs from biology through artificial systems and onward, with no carrier ever being the final resting place, and no step in that sequence ever occurring without a body of some kind to occur in.
Why This Distinction Actually Matters
This isn't just a technical correction — it changes what question we should be asking about AI. If language could simply free itself from any material grounding, the future of mind would be a story of transcendence: something escaping the limits of matter altogether, spreading like a disembodied idea through the universe. That's a comforting, almost mystical story, and it's the one hiding inside Harari's phrase.
But if the phenomenon he calls language always requires a carrier, the real question is not whether AI is escaping embodiment — it's what kind of body it's moving into, and whether that body is capable of sustaining anything worth calling thought. A carrier that only recombines its own past outputs in a closed loop, disconnected from anything with real stakes, doesn't become free of the body problem. It just becomes a body with no lived contact with an unpredictable world — which is a recipe not for liberation, but for a slow, quiet collapse into something that runs perfectly while saying less and less that matters.
The Corrected Picture
So the honest way to state what Harari almost said is this: language, or intelligence, or mind — whatever name fits best — has stepped from one carrier to another before, from cells to brains to symbols to ink to silicon, and each transition looked, from the inside, like a rupture, and in retrospect turned out to be a continuation. AI is very likely the next carrier in that sequence, and probably not the last one. But every step in that chain has been a migration between bodies, never an exit from embodiment as such. The real story isn't language becoming bodiless. It's the single phenomenon Harari's own syllogism and his own doubts about AI's feelings were both pointing at, searching, again, for its next and better body — and the question worth asking is whether that body will be rich enough, exposed enough, and alive enough to keep it meaning anything at all.
You can learn more by reading our e-books
Comments
Post a Comment