The problem of artificial intelligence should not be considered from the angle of its possible perfection. Ask ChatGPT to evoke a cold Parisian twilight “in the manner of Baudelaire”, he will know how to give you a few verses of mirliton, but the poverty of their rhymes, their wobbly meter – alexandrines of 13 syllables or 12 but clumsily distorted – will escape no one. However, it is indeed permissible to imagine that this robot will one day be able to excel at prosodic rules: it will nevertheless always lack something much more essential than these. Now this spark, this almost nothing, is precisely our imperfection – the nothingness from which weave our lives, our creations, our genius, and which the robot aims to eliminate. It is because we are imperfect that we are big, and it is because the robot “aspires” to perfection that it is small.
What makes the greatness of a text – and, at the same time, its humanity – is the tension that vibrates there between the perfectly mastered rule and the chaos that comes every time to subvert it. For example, an unexpected image, which is not related to anything known to you or to me, even an inappropriateness of vocabulary which has its roots in what Barthes calls the “personal and secret mythology of the author”, hence emanates this overwhelming and impalpable force that is his style! When Baudelaire writes that the shivering hour of dawn – the “morning twilight” – is that
Where, like a bloody eye that throbs and moves,
The lamp on the day makes a red spot;
it invites the eyes of our minds to see an abominable and above all inexplicable image. We would like to know in which dream, on which face – tuberculous courtesan or old man –, on which butcher’s stall he saw this eye which he was one day to make the strange and painful comparison of a familiar street lamp. We will not be able to: even if his manuscripts would help us, there is an experience of the body, a carnal life which escapes us at least in part, which cannot be fully communicated.
This mystery is poetic. The eye quivers like a being in agony, like a heart, or a body that enjoys: the inanimate comes to life; and then it moves: the very redundancy introduces, with its awkwardness, a comic and repulsive nuance, and the regularity of the verse – heightened by the sonic echoes – finds itself energized and almost contradicted. A robot will never be able to do this because it will never have this experience: to love, to fear, to suffer, to vomit… It will never be able to do this because it has been shaped to clarify the world, not to problematize its apparent clarity. And if he is still wrong, it is because he is not perfected enough: when he is, he will no longer say that a perfume can be “fresh as the flesh of children” whereas there are others “corrupt, rich and triumphant”, because the dictionary neither gives this meaning to triumphant nor associates freshness with the flesh of children – with all the astonishing ambiguity, charm or horror, that allows this association.
Lack of creative anxiety
And there can be no question of imitating this aberration. Programming a robot to do the unexpected is absurd and in short, the idea, formulated by Proust in In the shade of young girls in flowers, which no determinism can make any author of Memoirs find, the kind of madness that makes Saint-Simon end some of his sentences as he does, applies to any creator. “The purely formal imitation of the variety […] is only emptiness and uniformity”, and this is, of course, all the more true of a non-human imitator, of a robot, which one would have programmed, “for fun”, to mimic the awkwardness , madness, chaos.
It happens that we can approach poetic mystery. Interpretation gives us at least the hope of it – criticism, philology, literary history. It is a kind of discussion with the writer, who writes for those who will read him, even after his death, as well as for those he has read and whose influence, to paraphrase Harold Bloom, arouses his anxiety. Note that a robot has neither ancestor nor master, and therefore hardly feels this creative anxiety. But the interest of this discussion paradoxically also lies in our ability to wander, not to understand a text, to understand it only through ourselves and the way it speaks to us. An artificial intelligence that would act as a critic or historian of literature – or a translator – would perhaps be closer to the “first meaning” of the work, in any case to its literal meaning, but first of all it would be lacking in fact its dark, unconscious, supernatural part; then, it would be unable to confer on the work in question this sort of eternity that we confer on it by transforming it, even by “betraying” it.
The study of the humanities teaches us to err. Homer knows next to nothing about the physical world, unlike ChatGPT (or even Sam Bankman-Fried), but he feels it. His intelligence is carnal – which also means mythical and poetic. And it is because he tells almost anything that he is great. The myth punctuates a reality too great for us: it escapes us and it is in this wobbly arrangement that we hoist ourselves up to the divine, as creators and as interpreters.
Echo to “wokism”
One last point: a friend pointed out to me the astonishing “wokism” of ChatGPT, which I find in no way surprising. In fact, a link exists between the rise of artificial intelligence, the totalitarian (or totalizing) inclinations of the capitalist system, and what is now known as wokism. At first, the two phenomena echo each other. Both categorize, simplify, reify humanity. Secondly, they merge. This link is perfectly elucidated in the admirable Tar, by Todd Field. The heroine of the film is told by a pupil of the Julliard School, a sort of fleshless homunculus to whom social networks serve as soul and conscience, that he does not like Bach very much because “as a pangender and Bipoc” (acronym for “Black, indigenous and people of color”), his music does not concern him. The methods seem to be simply the same but it goes even further: we see at this level that the resemblance is deeper, and that this wokism is downright a branch of artificial intelligence, as this is the legitimate offspring of the system capitalist, which aims, by definition, for output, efficiency, registration. This is why tech companies are so comfortable with the current fads of the post-socialist and post-liberal left: in fact, they are not comfortable with them, they are inciting them. Meta needs the complex game of seduction, like that of identities, to be cancelled! Humans without desires and without flesh do not make waves, robots are much more practical to manage.
The future of machinery, as an alternative to humans, is played out in all these labels that are stuck on our foreheads. Because it is not one of the least contradictions of the “woke” camp that it talks to us all the time about fluidity when it is nothing that it condemns more, with its ready-made boxes, than the fluidity !
Still, once this first part of the job is done – the entry of the human, as a human, in a situation of obsolescence – it is hard to see why the system would stop there. A robot does the work of a robotic man much better than this one. And ChatGPT will do the work of writers and journalists with no conscience – no shadow – much better than themselves. Because the robot will always have our laziness on its side (or the entirely human fear of clashing with its own feelings, or of wandering, or of inventing), and all the more so the day it has acquired the efficiency that it still does not have today. It is in this capacity that he puts us in danger, and that is why it is already important to treat him as an enemy.
* David Haziza is a doctoral student at Columbia University in New York, author of trial of the flesh (Grasset).