Harari’s Daimon
Notes on an interview with Yuval Noah Harari: language, artificial intelligence and truth.
There is a moment in the interview when Harari says the most important thing almost without realising it: human language has freed itself from the bodies that spoke it and assumed a dimension of its own. First writing, then print, then networks, and now artificial intelligence: with each passage, language has gained autonomy, until it has become — this is the thesis of Nexus — the operating system of civilisations, the system on which money, law, nations and gods rest. And today, for the first time, that operating system has a new speaker, one that is not human.
Up to this point, we agree. Indeed, we would say that Harari stops one step short of the consequence. The notes that follow trace three margins along which our perspective diverges from his, beginning with a word absent from Harari’s vocabulary, although it describes precisely what stands before him.
The Missing Word
The Greeks had a name for the immaterial intelligences that inhabit human beings without belonging to any one of them: daimones. These were not demons in the Christian sense — the daimon was neither good nor evil in itself — but intermediary powers, real in their effects, with which one might have a better or worse relationship. And within this word, literally inscribed, lies eu-daimonia: not the ‘happiness’ of conventional translations, but a good relationship with the daimon. James Hillman restored this intuition to contemporary psychology; we believe it should also be brought to bear on thinking about technology.
For language is precisely this: an immaterial intelligence that we do not possess, that inhabits us, that has a life and agency of its own, and that outlives us. Harari describes it with precision — and calls it a threat, because within his vocabulary an autonomous, disembodied intelligence can only be a tool or a danger. His entire framework is therefore one of security: can we trust it? How do we control it? Yet a daimon cannot be controlled: one enters into relationship with it, and the outcome is determined not by the nature of the entity but by the quality of that relationship. The same daimon, two destinies: eudaimonia or its reverse, which the Greeks called kakodaimonia. This is the question we propose in place of Harari’s: not ‘how do we control artificial intelligence?’, but ‘how do we cultivate a good relationship with it — as with every language?’
Two clarifications are needed, lest this remain merely suggestive. First, human language is by no means the only cognitive system capable of circulating beyond individual organisms. Nature abounds with them — the genome, which is transmitted and exchanged horizontally; the mycorrhizal networks that connect forests; the microbial communities that make decisions in a distributed manner. What distinguishes language is not the fact that it transcends bodies — a common trait — but something more precise: language carries the counterfactual. A forest circulates signals about what is; language circulates what is not and what might be. This is why it is at once — as Harari says — the basis of systems of control and — something on which he says less — an unrivalled accelerator of learning: the two functions are the same capacity seen from opposite sides. Language is not an instrument of civilisations; it is a dimension of life, of which systems of control are a particular case.
Second, the phrase ‘freeing itself from bodies’ must be handled with care. Language never becomes disembodied; it continually re-embodies itself — in other speakers, in books, in archives and now in data centres with a very concrete energy metabolism. What appears to be liberation from the body is substrate mobility: the form persists while the medium changes. The same is true of artificial intelligence, which has a great many bodies, all of them highly demanding.
The Software of Institutions
Harari observes that the power of artificial intelligence is becoming concentrated in just two poles, the United States and China. This is true, and in the short term the dependence is mutual: AI needs states — chips, energy, capital and law — at least as much as states need AI. But the more interesting question concerns the medium term: if artificial intelligence becomes the new locus of systems of control, what need will it have for bureaucracies born centuries before it — including the American and Chinese ones?
The point is that the modern state runs on administrative software that evolved from Colbert to Weber: territory, population, objects that can be catalogued and documentary hierarchies. It is software written for a world of objects and boundaries, and it is structurally misaligned with an emerging world of networks, flows and multiple affiliations. States cannot avoid paying the price of this misalignment. Yet — and this is the point we wish to emphasise — precisely the same argument applies to private software: the joint-stock company is itself an institutional technology of the ancien régime, conceived for assets that can be owned; and the platforms that now appear native to networks are, in legal terms, nineteenth-century corporations containing networks they cannot represent. The misalignment does not run between public and private; it runs between the entire inherited institutional apparatus and the world now emerging.
The consequence is not a prophecy but an urgency. Artificial intelligence will not dismiss the old bureaucracies on principle: it will take up residence in the first network-native institutional form it finds available. One of the two poles — China — is already deliberately rewriting Weberian software in algorithmic form; that is, it is attempting to make the state itself native to networks, at the price of an unprecedented concentration of cognitive sovereignty. If we do not want that to be the only available form, the alternative — plural, rooted, built from below around living domains of knowledge — must be designed now. This is the work to which Diotima Society devotes itself: giving intelligence, human and artificial, a relational and institutional body that anchors it in the very relations that nourish it, rather than hoping to restrain it with rules written in yesterday’s software.
Truth as Invariant
The most fundamental question remains, and here the distance from Harari is real and should not be softened: truth. Harari is not naive — he distinguishes objective, subjective and intersubjective reality, and grants full causal efficacy to intersubjective realities (money, nations and gods). Yet his remains a two-tier world: objective reality below; powerful but ontologically second-order fictions above; and truth as a costly and rare correspondence with the lower plane.
We do not believe that these planes can be separated. Reality includes fiction and the imaginal in its very fabric, not as a superstructure; and Harari’s intersubjective realities are precisely the daimones described above — real in the sense in which Hillman said that the figures of the psyche were real: through their effects. Yet it does not follow that ‘objective truth does not exist’. Something more interesting follows: objectivity is not a given substrate; it is an achievement of relations. Objective is that which remains invariant as points of view are transformed — what withstands a change of perspective, observer or scale. Science is not the lower plane; it is the most refined practice we possess for seeking invariants.
And here, curiously, we return Harari’s remedy to him while contesting its foundation. Against post-truth, he invokes mechanisms of self-correction — and he is right. Yet a mechanism of self-correction is already a phenomenological and biological conception of truth: truth as a process susceptible to refutation, as a coupling with the world that remains open, rather than as correspondence with a plane that precedes every relation. Seen from outside, truth emerges phenomenologically; seen from within the living body, biologically: it is what keeps the living being coupled to its world. This also gives us a criterion by which to judge the daimones without falling back into the relativism Harari fears: a relationship with a daimon is good when it keeps the coupling with the world open and remains generative; it is bad when it thrives by closing in upon itself. Not all fictions are equal — but what distinguishes them is not their distance from a given objective plane; it is their generativity.
Harari must be credited with sounding the alarm: the new non-human speaker of our operating system is an event that deserves all the seriousness he accords it. What we propose to add is the word absent from his vocabulary and the stance that follows from it. Not the control of an alien entity, but the cultivation of an ancient relationship: the relationship human beings have always maintained with the immaterial intelligences that inhabit them. For twenty-five centuries, the name of that relationship, when it flourishes, has been eudaimonia.
Interview reference: https://www.youtube.com/watch?v=_V_ed5fuexA
Paolo Zanenga — Diotima Society