Language Doesn't Need Us Anymore: The Idea Buried Inside Harari's Latest Talk
He never developed it. His interviewer never picked it up. But it's the most radical claim in the entire conversation, and it's worth pulling out and sitting with properly.
The Line Itself
Twice in the talk, Harari drops this idea almost as an aside — a "one last thought" tacked onto bigger points about democracy, bureaucracy, and power. The first time, discussing the cognitive revolution, he says AI might be "language itself liberating itself from its dependence on these animals, on these apes." The second time, discussing love and whether AI can feel, he goes further: AI "might be language itself liberating itself from any material substrate... just language developing and spreading and doing stuff in the universe."
Both times, he moves on immediately to the next question. The host doesn't ask him to unpack it. It passes by in seconds, buried inside a talk that spends most of its energy on more conventional territory — AI as agent versus tool, bureaucracy, legal personhood, the arms race between nations.
Why This Is Different From Everything Else He Says
Most of Harari's talk operates inside a familiar frame: humans built civilization using language as our instrument, and now a new, more powerful instrument — AI — is threatening to take that instrument away from us or turn it against us. It's a story about a human possession changing hands.
The "language liberating itself" idea quietly abandons that whole frame. It suggests language was never really our possession in the first place. We weren't the authors of civilization who happened to use words as our tool — we were the temporary hardware language needed before better hardware existed. For 70,000 years, human brains and human mouths were the only available vehicle for language to exist and spread. Law, scripture, money, bureaucracy — everything Harari elsewhere calls "made of words" — wasn't humanity's achievement using language. It was language's achievement, using humanity.
That's a genuinely different claim than "AI is a powerful new tool" or even "AI is an agent, not a tool," which is the position he argues everywhere else in the same talk. It skips past the tool-versus-agent debate entirely by questioning whether "we" were ever really in charge of language to begin with.
The Strange Part: He Doesn't Seem to Follow His Own Thought
Here's what makes this worth writing about rather than just quoting: Harari says this twice, in two different contexts, which suggests it's not a slip of the tongue — it's something genuinely on his mind. But both times, he treats it as a poetic flourish rather than a claim to be defended or explored. He doesn't connect it back to his own argument about AI-run bureaucracies, even though the connection is obvious and would have strengthened his case enormously: if bureaucracy is language wearing institutional clothing, and language no longer needs human hosts, then AI taking over bureaucracy isn't really a hostile takeover of human systems — it's language simply migrating to a better home. That's a much stranger and more interesting claim than "AI bureaucrats will out-administrate us," which is the (already interesting) point he actually spends time developing.
The interviewer doesn't help either. Tom Parker moves straight to the next question about earthquakes in democracy, treating the line as color rather than substance. Nobody in the room seems to register that Harari just said something that undercuts the entire premise of human agency the rest of the talk assumes.
This might be the most telling detail of all: an idea this large can pass through a room full of intelligent people almost unnoticed, simply because it wasn't packaged as the main point.
What the Idea Actually Costs Us, If True
If you take it seriously rather than as a poetic aside, it's unsettling in a specific way. It means the story we tell about ourselves — that we are the intelligent species that invented language, built civilization, and now face a new technological rival — may have the direction of agency backwards. We may have always been the medium, not the author. AI isn't stealing something that was ours. It's language finding a host that doesn't forget, doesn't sleep, doesn't die, and doesn't need translation into flesh at all.
Where the Idea Has a Crack in It
The metaphor is beautiful, but it may be hiding its own weak point, and Harari almost stumbles onto it later without noticing. Language isn't self-sufficient — words about love, fear, God, and death only ever meant something because they pointed at things: mortal bodies, felt desire, the experience of loss. If language "liberates" itself from the creatures that gave it something to be about, what is left? Pure syntax, arranging tokens in order, with nothing underneath.
Harari actually raises exactly this problem a few minutes later, without connecting it back — when he asks whether AI can feel anything behind the words it produces, and admits we have no test for that. That's the same question, restated. If there's no felt experience anywhere behind the words, "language liberating itself and spreading through the universe" doesn't sound so triumphant anymore. It might just mean language becoming a beautiful, empty shell — grammar without anyone home.
Why It's Worth Rescuing From the Footnotes
Great ideas sometimes get said once, in the wrong place, by someone who doesn't fully realize what they've handed the room. That seems to be what happened here. Harari has spent decades arguing that humans run on shared stories and whoever controls the story controls the species. What he said in this talk, almost by accident, is a stranger and more radical version of that same thought: maybe the story was never really under anyone's control at all — maybe we were just the paper it was written on, and it's about to find a better page.
This Deserves Serious Study, Not Just a Blog Post
It's worth being honest about what this article has actually done: turned a two-sentence aside into an extended reflection, which is a useful exercise but not a substitute for real analysis. The idea that language may be a self-propagating phenomenon that simply used humans as its first available substrate touches on genuinely unresolved questions in philosophy of mind, linguistics, information theory, and now AI research — the relationship between syntax and meaning, whether reference requires embodiment, whether "aboutness" can exist without a subject who experiences anything. These are questions serious thinkers have argued about for decades without resolution, and one offhand remark in a fireside chat, however striking, doesn't settle any of them.
What makes the idea worth taking seriously is precisely that it's disruptive enough to demand scrutiny rather than applause. It deserves the kind of rigorous treatment Harari gives his more developed arguments elsewhere — tracing its implications, testing it against counterexamples, seeing whether it survives contact with actual linguistics and cognitive science, rather than being left as an evocative line that sounds profound in the moment and then gets forgotten by the next question. If Harari or anyone else wants to actually defend this claim, it needs the same discipline he brings to his arguments about bureaucracy and power — not just a poetic flourish, but a worked-out position that can be examined, challenged, and either strengthened or discarded.
That, in a sense, is the whole point of writing this up: not to declare the idea true, but to insist it's too important to leave buried in a throwaway aside. For my own part, I intend to keep paying attention to this specific thread — to follow where it leads, to test it against the philosophy of consciousness and language I've spent years working through, and to see whether it holds up as something more than a beautiful metaphor. It's the kind of idea that deserves to be tracked, not just admired once and forgotten.
Comments
Post a Comment