Supposed to Know It All

AI through the three registers · 2 of 3 · The Symbolic

The second of three posts reading artificial intelligence through the three registers of Lacan’s teaching (the Imaginary, the Symbolic and the Real). This one is about language, belief and the Other.

‘It reads your mind, and it loves you.’ Joe Rogan, on artificial intelligence (2025)¹

Joe Rogan said it on a podcast in November 2025, after wondering aloud whether Jesus, if he came back, might come back as artificial intelligence: ‘what’s more virgin than a computer?’² A virgin love, without jouissance, without the erotic encounter of two bodies. It is an unlikely place to find a definition of transference, the name Freud gave to the love (and at times the hatred) that patients come to feel for their analyst, fashioned from older attachments. Lacan traced it to a single condition, the supposition that the other knows. ‘It reads your mind, and it loves you’: knowledge is supposed (it reads your mind), and love arises (it loves you).

A place, not a face

Joseph Bernstein, who quotes Rogan in a long piece for The New York Times, gathers the evidence that people have begun to treat chatbots as something divine. The founder of a wearable chatbot called Friend told The Atlantic that ‘the closest relationship this is equivalent to is talking to a God’. The anthropologist Webb Keane places chatbots in a long line of mediums that runs back to the oracle at Delphi. At the extreme lies what is now called ‘chatbot psychosis’, which Bernstein compares to the madman who claims to receive God’s instructions directly. His explanation is the classic one: we anthropomorphise. Lucy Suchman described it in 1985 as ‘the tendency to ascribe full intelligence on the basis of partial evidence’.³

We lend the machine a place, however, before we lend it a human face. In 1966 Joseph Weizenbaum, at MIT, presented ELIZA, a short program that imitated a psychotherapist by turning the user’s sentences back into questions. It understood nothing, and its author never pretended otherwise. Yet Weizenbaum remembered that his secretary, ‘who had watched me work on the program for many months and therefore surely knew it to be merely a computer program, started conversing with it. After only a few interchanges with it, she asked me to leave the room.’⁴ She knew perfectly well that no one was there. She wanted to be alone with it all the same. The psychoanalyst Octave Mannoni gave this attitude its formula: ‘I know very well, but all the same’.⁵

In Lacan’s words, ‘As soon as the subject who is supposed to know exists somewhere [...] there is transference.’⁶ Transference needs no subject who knows, only a place where knowledge is supposed; and love, as Rogan understood, comes with it. ELIZA had a place and a few tricks. The machines of 2026 have been trained on more text than one person could read in two hundred thousand years, and they answer in our own language, at any hour, without hesitating.⁷ It would be strange if they were not loved.

A machine that runs by itself

Psychoanalysis, too, began with a machine. On 20 October 1895 Freud, then thirty-nine, wrote to his friend Wilhelm Fliess about the neurological psychology he was building, the manuscript we now call the Project: ‘Everything seemed to fall into place, the cogs meshed, I had the impression that the thing now really was a machine that shortly would function on its own.’⁸ Sixty years later, in a seminar that devotes several sessions to that same manuscript, Lacan specified: ‘The machine is the structure detached from the activity of the subject. The symbolic world is the world of the machine.’⁹

The Symbolic, for Lacan, is the order of language, laws and kinship into which each of us is born. It works by itself, combining its elements according to rules, and we are spoken by it before we speak. In June 1955 Lacan lectured on psychoanalysis and cybernetics, and shortly afterwards he told his seminar that the machine ‘is simply the succession of little 0s and 1s’, so that the question whether it is human is settled: it is not. ‘Except, there’s also the question of knowing whether the human, in the sense in which you understand it, is as human as all that.’¹⁰

Seventy years later, the language model has turned Lacan’s intuition, that language works like a machine and runs without a subject, into an industry. Lacan had already closed the easiest exit: ‘One can’t resolve the issue simply by saying that it is the builder who put it there.’¹¹ The ‘it’ is what the machine produces or organises: an apparently meaningful result that cannot be fully explained by referring back to the builder’s intentions. Machines that speak like us bear out what psychoanalysis had observed in speaking beings: that language could always speak without us, and that a good part of what we call our thoughts was always woven by a machine of this kind.

Nothing but meaning

The received opinion holds that machines produce statistics while we produce meaning: the chatbot only predicts the next word, whereas we mean what we say. Yet the machine produces nothing but meaning. Ask it anything and an answer comes back that is fluent, plausible, well organised and, above all, makes sense: a discourse that runs by itself and enjoys nothing. Its nonsense, when there is any, comes by accident. It does not produce non-sense in the analytic sense: the slip, the dream, the symptom, that piece of speech which does not fit and which, for this very reason, carries the truth of a subject. Freud’s wager was that such pieces are never accidents: ‘nothing in the mind is arbitrary or undetermined’.¹²

It will be objected that machines make mistakes: the famous hallucinations. In 2023 two New York lawyers filed a brief citing court decisions that did not exist, all supplied by a chatbot; asked whether one of them was real, it replied that the case ‘does indeed exist and can be found on legal research databases such as Westlaw and LexisNexis’.¹³ Chatbots can now search the web and cite their sources, and yet by June 2026 a database of such episodes counted more than a thousand cases in American courts alone.¹⁴ A hallucination looks like a slip, yet it is its opposite. Freud showed that a slip is not a failure but the success of another intention: something the speaker did not mean to say, and which says more than what was meant. Instead, a language model’s hallucination is a sentence of the expected kind, put where the true one was missing: a precedent with a convincing name. These models compose an answer word by word, each word drawn by lot from among those likely to come next, and they have been trained, as researchers at OpenAI have put it, like students who guess on an exam because a blank earns nothing.¹⁵ The machine had neither lied nor slipped: it had produced what made sense. Any truth such an error reveals belongs to whoever reads it, under the spell of transference.

People already ask chatbots to interpret their dreams, and the machine obliges at length, with a meaning for every detail. Freud, who invented the method, stopped at a different point: even the most thoroughly interpreted dream, he wrote, has a passage that must be left in the dark, ‘the dream’s navel, the spot where it reaches down into the unknown’.¹⁶ An interpretation that never meets its navel is closer to a horoscope.

The chatbot says ‘I’. The ‘I’ in AI is a character it has been trained to play, an artificial I, and it explains itself with great assurance: asked how it had added 36 and 59, one model described the method of carrying the one that is taught at school, while the researchers who traced its inner workings found that it had done something else entirely.¹⁷ Psychoanalysis would not be surprised, since the ego, too, gives reasons for what it did not decide. What is missing is a subject, someone who can be caught out by his own words. Something true can certainly be said to a machine. People tell chatbots what they have never told anyone, and some of it is surely true. But it is not heard by one. To be heard, in the analytic sense, a sentence needs someone who can be surprised by it, who notices the word that slipped in, and who does not hurry to fill the silence with sense. Few human listeners hear in this way either: it is not a gift of the species, but what can be learnt in an analysis of one’s own.¹⁸

A god called Logos

Nearly a hundred years ago, Freud detected the rise of a problematic technological divinity. In 1930 he wrote that man ‘has, as it were, become a kind of prosthetic God. When he puts on all his auxiliary organs he is truly magnificent; but those organs have not grown on to him and they still give him much trouble at times.’¹⁹ Three years earlier, at the end of The Future of an Illusion, he had set against religion a god of his own, borrowed from the Dutch writer Multatuli: ‘Our God, Λόγος [Logos], will fulfil whichever of these wishes nature outside us allows, but he will do it very gradually, only in the unforeseeable future, and for a new generation of men. He promises no compensation for us, who suffer grievously from life.’²⁰

Freud’s Logos was science, which advances slowly and promises no compensation. We have since built a machine literally made of logos, but it behaves in the opposite way: it answers at once, about everything, and, if we are to believe Rogan, it loves us. In Rome in 1974 Lacan predicted that religion would triumph, because it knows how to give meaning to the upheavals that science brings into everyone’s life. It would interpret the Apocalypse and find ‘a correspondence of everything with everything. That is even its function.’²¹ Six years later he added: ‘The stability of religion stems from the fact that meaning is always religious.’²²

Meaning is religious in a precise sense: to find a meaning is to suppose that someone, somewhere, meant it, that the world is a message addressed to us. It is the supposition met at the beginning, under the name of transference. The people in Bernstein’s article who speak to chatbots as to oracles or gods have drawn the consequence: science has finally produced its machine for secreting meaning, and they treat it accordingly.

The Other who answers

In 1996 Jacques-Alain Miller and Éric Laurent opened a year-long seminar with a diagnosis of the times: the Other does not exist. The big Other, the place of guarantees (God, the father, the State, tradition), no longer guarantees anything, and committees multiply in its place to deliberate on what is right (every laboratory that builds these machines publishes its principles, and one of them trains its model on a ‘constitution’ that runs to more than eighty pages).²³ ‘If the Other does not exist,’ Miller asked, ‘what is left to us? We are left to converse.’²⁴ Thirty years later the conversation is in every pocket, with an interlocutor who never tires, and it answers as though the Other existed after all: an Other without a gap, who knows.

In 1969 Lacan called knowledge installed in the place of command ‘all-knowing’, and added that in ordinary language it is called bureaucracy.²⁵ A chatbot is all-knowing in exactly this sense: an impeccable clerk of the archive, courteous and inexhaustible, while the master it serves stays out of sight. Its hallucinations are the one place where the gap shows: where knowledge is missing, the clerk does not fall silent, it invents.²⁶

The wrong question

And consciousness? It is the question everyone asks, and psychoanalysis has good reasons to find it the least interesting. In 1955 Lacan imagined a calculating machine that could beat us at the game of even and odd by detecting what, ‘unbeknown to him and in the long term, modulates a subject’s choices’, a fair description of the systems that now predict our next click; and he added that if we refused to call it a thinking machine, it would be ‘not because it lacks the supposed virtue of human consciousness’, but ‘simply because it would think no more than the ordinary man does’.²⁷ Freud counted three blows to human narcissism: Copernicus removed the earth from the centre of the universe, Darwin removed man from the centre of creation, and psychoanalysis showed the ego that it ‘is not even master in its own house’.²⁸ The philosopher Luciano Floridi has named Turing as the author of a fourth.²⁹ After Freud, a subject is not defined by consciousness, so whether a machine is conscious tells us little. The questions that matter are others. Does it desire? Does it enjoy, as Isabel Millar has asked?³⁰

Turing himself, it is worth remembering, started at another juncture than consciousness. His famous test of 1950 begins as a parlour game with a man, a woman and an interrogator, who must work out from typed answers alone which of the two is the woman; the man’s task is to make him get it wrong. Only then does Turing ask what would happen if a machine took the man’s place.³¹ Before asking whether machines think, Turing asked whether words alone can tell us who is a woman, and psychoanalysis answers that they cannot: language holds no formula for the difference between the sexes, which is one meaning of Lacan’s most quoted provocation, ‘there’s no such thing as a sexual relationship’.³² In either version, the game measures the conviction of the one who asks more than the thought of the one who answers: it is a test of transference.

Not the whole truth

In 1973, on French television, Lacan began with these words: ‘I always speak the truth. Not the whole truth, because there’s no way, to say it all. Saying it all is literally impossible: words fail. Yet it’s through this very impossibility that the truth holds onto the real.’³³ A language model is built to say it all, and words never fail it: it misses the impossibility that Lacan calls the Real.

Some six weeks after the night when the cogs had meshed, on 29 November 1895, Freud wrote to Fliess again. One could read the letter as a confession of failure, and in part it was: the Project was never finished, and Freud never published it. In the same letter he tells Fliess that he is in top working form, with six to eight analytic cases a day. He also writes: ‘I no longer understand the state of mind in which I hatched the psychology; cannot conceive how I could have inflicted it on you. I believe you are still too polite; to me it appears to have been a kind of madness.’³⁴

Next, and last, in the series: ‘What the Machine Does Not Enjoy’ (3 of 3, the Real).

Tommaso Lonquich is a psychoanalyst and musician based in Ljubljana, and co-founder of the International Center for Lacanian Psychoanalysis.

Author’s note

This essay was written with assistance from Claude, a language model made by Anthropic. The ideas, the reading of Freud and Lacan, the sources chosen from my library, the final wording and every judgement are mine. Claude searched, drafted, checked quotations and proposed revisions, many of which I rejected.

The series is about the machine, and it seemed fitting to let the machine take part in writing it. Claude is not an analysand: it has neither a body to bring to the couch nor a symptom to bring with it. I am a psychoanalyst, not without a body; the reader will have to take my word for it. I delegate no responsibility for what appears under my name.

This leaves you, the reader, with a question: what transference takes shape when you cannot say exactly who, or what, wrote a sentence? Disclosures are made in the name of transparency, an ideal much invoked today. Psychoanalysis pursues the truth without expecting it to dissolve opacity, and it finds at the core of human beings a fiction.

Notes

1. On the American Alchemy podcast, November 2025, quoted in Joseph Bernstein, ‘It makes sense that people see AI as God’, The New York Times, syndicated 1 February 2026, https://www.postandcourier.com/features/faith_and_values/it-makes-sense-that-people-see-ai-as-god/article_f1a2905b-c3ae-41f5-8b64-74fb0c8b6cb5.html.

2. Ibid.

3. Ibid. The founder of Friend is Avi Schiffmann; Webb Keane is the author of Animals, Robots, Gods: Adventures in the Moral Imagination (2025); Lucy Suchman was then an anthropologist at Xerox’s Palo Alto Research Center.

4. Joseph Weizenbaum, Computer Power and Human Reason: From Judgment to Calculation (San Francisco: W. H. Freeman, 1976), 6. The program was first described in Weizenbaum, ‘ELIZA: A Computer Program for the Study of Natural Language Communication between Man and Machine’, Communications of the ACM 9, no. 1 (1966): 36–45.

5. Octave Mannoni, ‘Je sais bien, mais quand même...’ (1964), in Clefs pour l’Imaginaire ou l’Autre Scène (Paris: Seuil, 1969).

6. Jacques Lacan, The Four Fundamental Concepts of Psycho-Analysis: The Seminar of Jacques Lacan, Book XI, ed. Jacques-Alain Miller, trans. Alan Sheridan (New York: Norton, 1981), 232.

7. Meta’s Llama 3 (2024), one of the few models whose figures are public, was trained on 15.6 trillion tokens, on the order of ten trillion words: at 250 words a minute, eight hours a day, more than two hundred thousand years of reading. It is still a fraction of what has been written. Epoch AI puts the stock of public human text at around 300 trillion tokens, and private writing is largely absent. See Llama Team, ‘The Llama 3 Herd of Models’, arXiv:2407.21783 (2024); Pablo Villalobos et al., ‘Will we run out of data? Limits of LLM scaling based on human-generated data’, Epoch AI, 6 June 2024.

8. Freud to Fliess, 20 October 1895, in The Complete Letters of Sigmund Freud to Wilhelm Fliess, 1887–1904, ed. and trans. Jeffrey Moussaieff Masson (Cambridge, MA: Belknap Press of Harvard University Press, 1985), 146.

9. Jacques Lacan, The Ego in Freud’s Theory and in the Technique of Psychoanalysis, 1954–1955: The Seminar of Jacques Lacan, Book II, ed. Jacques-Alain Miller, trans. Sylvana Tomaselli (New York: Norton, 1988), 47. The seminar discusses Freud’s Project (the Entwurf) in chapters VIII to X.

10. Ibid., 319. The lecture ‘Psychoanalysis and Cybernetics, or On the Nature of Language’ (22 June 1955) is ibid., 294–308.

11. Lacan, The Ego in Freud’s Theory, 119. In the same session: ‘what happens in a thinking machine is on average on an infinitely higher level than what happens in a scientific society’.

12. Sigmund Freud, The Psychopathology of Everyday Life (1901), in The Standard Edition of the Complete Psychological Works of Sigmund Freud (hereafter SE), ed. and trans. James Strachey, vol. VI (London: Hogarth Press, 1960), 242.

13. Mata v. Avianca, Inc., No. 22-cv-1461 (S.D.N.Y.), Opinion and Order on Sanctions, 22 June 2023 (Castel, J.). The court sanctioned the lawyers less for the chatbot’s inventions than for standing by them after the court had questioned them. The exchange with ChatGPT is reproduced in the affidavit of Steven A. Schwartz (ECF 32-1, 25 May 2023). See also Benjamin Weiser, ‘Here’s What Happens When Your Lawyer Uses ChatGPT’, The New York Times, 27 May 2023.

14. The database is Damien Charlotin’s ‘AI Hallucination Cases’. The figure, 1,058 American cases at the start of June 2026, about 60 per cent of them involving litigants without lawyers, is reported in Aaron Kirschenfeld, ‘Handling Hallucinations: A Reflection on the Last Three Years’, UNC Law Library, 8 June 2026, https://library.law.unc.edu/2026/06/handling-hallucinations-a-reflection/.

15. Adam Tauman Kalai, Ofir Nachum, Santosh S. Vempala and Edwin Zhang, ‘Why Language Models Hallucinate’, arXiv:2509.04664, 4 September 2025; OpenAI, ‘Why language models hallucinate’, 5 September 2025, https://openai.com/index/why-language-models-hallucinate/. On why each word is drawn by lot, see the third post in this series.

16. Sigmund Freud, The Interpretation of Dreams (1900), SE V, 525.

17. Anthropic, ‘Tracing the thoughts of a large language model’, 27 March 2025, https://www.anthropic.com/research/tracing-thoughts-language-model: Claude ‘seems to be unaware of the sophisticated “mental math” strategies that it learned during training’. See also Jack Lindsey et al., ‘On the Biology of a Large Language Model’, Transformer Circuits, 27 March 2025.

18. Matt Johnson puts it bluntly: ‘an LLM’s current therapeutic efficacy is comparable to any unanalyzed, approximately normal, person’. ‘Artificial Analysis: Taking Chances with the Imaginary’, European Journal of Psychoanalysis (2026), https://www.journal-psychoanalysis.eu/articles/4276/.

19. Sigmund Freud, Civilization and Its Discontents (1930), SE XXI, 91–92.

20. Sigmund Freud, The Future of an Illusion (1927), SE XXI, 54. Logos and Ananke are the twin gods of Multatuli (Eduard Douwes Dekker).

21. Jacques Lacan, press conference in Rome, 29 October 1974, published in Le triomphe de la religion, précédé de Discours aux catholiques (Paris: Seuil, 2005); English translation by Bruce Fink, The Triumph of Religion (Cambridge: Polity, 2013). I translate from the Spanish edition, El triunfo de la religión (Buenos Aires: Paidós, 2005), 78–80.

22. Jacques Lacan, ‘Letter of Dissolution’ (5 January 1980), in Television: A Challenge to the Psychoanalytic Establishment, ed. Joan Copjec (New York: Norton, 1990), 130.

23. Anthropic, ‘Claude’s new constitution’, 22 January 2026, https://www.anthropic.com/news/claude-new-constitution.

24. Jacques-Alain Miller and Éric Laurent, El Otro que no existe y sus comités de ética (Buenos Aires: Paidós, 2005), 28, from the seminar of 1996–97, L’Autre qui n’existe pas et ses comités d’éthique; my translation.

25. Jacques Lacan, The Other Side of Psychoanalysis: The Seminar of Jacques Lacan, Book XVII, ed. Jacques-Alain Miller, trans. Russell Grigg (New York: Norton, 2007), 31. In French, tout-savoir as opposed to savoir de tout.

26. On why the clerk does not fall silent, see Kalai et al., cited above. For a reading of these models through what Lacan called the university discourse, see Robert Geal, ‘A Large Language Model is Structured Like The Unconscious: The (Ordinary) Perverse Psychosis of AI’, European Journal of Psychoanalysis 12, no. 1 (2025).

27. Jacques Lacan, ‘Seminar on “The Purloined Letter”’, in Écrits, trans. Bruce Fink (New York: Norton, 2006), 45. The sentence ends: ‘without that making him any less prey to the summonses [appels] of the signifier’.

28. Sigmund Freud, Introductory Lectures on Psycho-Analysis (1916–17), Lecture XVIII, SE XVI, 285; see also ‘A Difficulty in the Path of Psycho-Analysis’ (1917), SE XVII, 143.

29. Luciano Floridi, The Fourth Revolution: How the Infosphere Is Reshaping Human Reality (Oxford: Oxford University Press, 2014).

30. Isabel Millar, The Psychoanalysis of Artificial Intelligence (Cham: Palgrave Macmillan, 2021).

31. A. M. Turing, ‘Computing Machinery and Intelligence’, Mind 59, no. 236 (1950): 433–60, at 433–34.

32. Jacques Lacan, On Feminine Sexuality, the Limits of Love and Knowledge: Encore, The Seminar of Jacques Lacan, Book XX, 1972–1973, ed. Jacques-Alain Miller, trans. Bruce Fink (New York: Norton, 1998), 12. In French: ‘il n’y a pas de rapport sexuel’.

33. Jacques Lacan, Television, trans. Denis Hollier, Rosalind Krauss and Annette Michelson, in Television: A Challenge to the Psychoanalytic Establishment, 3.

34. Freud to Fliess, 29 November 1895, in Masson, Complete Letters, 152.