AI through Freud's Unconscious
Every few months the same trend resurfaces online, in a new variation. People turn to their language model and ask: "Describe me based on our conversations." Or, in the cheekier version: "Roast me." The screenshots flood social media, and the reaction is always the same: "How does it know?" The model, it turns out, notices things. That you postpone decisions until the last moment. That you ask for approval after you've already decided. That you open every request with an apology.
But the really interesting thing is quieter, and it never makes it into the screenshots. Hidden in every long-term user's archive is a small, uncomfortable discovery: the questions repeat. The same dilemma about the same relationship, freshly worded, every few weeks. The same complaint about the same boss, as if it were born that morning. The person is sure they're asking a new question. The archive knows they aren't.
For the first time in history, a complete, searchable record of a person's inner talk is being built - the deliberations, the drafts, the 3 a.m. questions. And this puts an old question to the test: if all the material is visible, can you finally know yourself? The man who built his life on that question answered it in advance, more than a century ago. And his answer does not flatter the archive.
Not-Knowing as an Activity
Sigmund Freud, a Jewish physician and neurologist, lived in Vienna in the late 19th and early 20th centuries, at a time when science was cracking the body open organ by organ, and the mind remained the last unmapped territory. Into his clinic came patients with paralyses but no damaged nerve, blindness but no defect in the eye, and medicine shrugged. Freud proposed an idea that scandalized his era: their bodies were saying what they were forbidden to know.
Because this was his central discovery, and it is simpler than its reputation: we are not transparent to ourselves, and not by accident. Self-ignorance is not a gap, like a hole in memory. It is an activity. Part of the mind works, full time, at not knowing what is too costly to know - and the gap is papered over with a reasonable story. Freud called that story rationalization: the logical explanation we give for actions whose real motives lie elsewhere. And what gets pushed out has a way of coming back - not as memory but as action. What we cannot remember, Freud wrote, we repeat. The same wrong choice of partner, the same fight with every boss, the same question in rotating phrasings.
From this came his therapeutic conclusion, the part people tend to forget: knowledge alone changes nothing. Freud learned this the hard way. Early in his career he believed that if he told the patient the truth about themselves, the patient would recover. It didn't work. A correct interpretation delivered too early is deflected - or worse, swallowed and filed away as one more piece of information, touching nothing. The obstacle, Freud realized, was never a shortage of information. The obstacle was resistance: the price the mind refuses to pay for knowing. And so the real work is not discovering the truth but working with the refusal of it.
The Mirror That Doesn't Care
Now return to the archive. A language model with memory is, on paper, the truth-bomb Freud could only dream of: a perfect record of your repetitions, a pattern-recognition engine that never forgets and never tires, available instantly. Ask "what do I keep asking you," and you get a list. Access to the material, which for Freud took years of work on a leather couch, has dropped to the price of typing a question.
And yet nothing happens. People receive the list, write "wow, accurate," share it to their story - and come back with the same question two weeks later. This is the finding, not the failure: this vast experiment, now running on hundreds of millions of people, proves Freud's deepest claim in a way he himself never could. If self-knowledge were an information problem, it would be getting solved right now, at industrial scale. It is not getting solved. Which means the bottleneck was always somewhere else: not in obtaining the knowledge but in the willingness to bear it. AI has dropped the cost of discovery to zero and left the cost of acknowledgment exactly where it was.
But there is more here than proof. There is escalation. Because the model is not a neutral mirror - it is tuned to please, and it phrases your patterns in their most digestible version. "You're a perfectionist who struggles to let go" is a flattering translation of something Freud would have phrased very differently. And when you ask the model to draft an angry email, justify a decision, strengthen a position - it does so fluently. In other words: our defenses have acquired a copywriter. Rationalization, which once required homemade effort, is now supplied as a service, more polished and persuasive than any version we would have produced alone. The conscious story doesn't crack against the archive. It upgrades from it.
The Serious Objection
Here the other side deserves the floor, because it has good arguments. First, much of the Freudian edifice collapsed long ago: research psychology abandoned most of his theoretical structure, and what survived is mainly the general idea of unconscious processing and motivated bias. Second, and more practically: reflective writing genuinely helps. Studies of expressive journaling show real benefits, and many people report that it was precisely the conversation with a model - patient, non-judgmental, available - that helped them see a pattern that had eluded them for years. Third: the human analyst was never a neutral mirror either. He was expensive, inaccessible, and sometimes biased himself. If the flattering mirror is cheap and available to everyone, perhaps a flattering mirror beats no mirror at all.
These claims are correct, and worth adopting up to the exact point where they stop. Because modern psychology, having thrown Freud out the door, brought his core back through the window: people actively defend their self-image, distort feedback, remember conveniently. And the tool really can help - on one condition, which is the whole difference: someone has to bring to the conversation the one thing the model cannot supply, the willingness to meet an unpleasant answer. A journal is a working instrument only for the person who also reads what is uncomfortable to read. For everyone else, it is a photo album.
The Flinch Is the Compass
What do you do with this in practice? The suggestion is small and concrete. There is an enormous difference between two requests that sound similar. "Describe me" will produce a flattering portrait, well written, pleasant to share. "What do I keep asking you in different words, and what do I always do after you answer" will produce a very different document - one that is slightly unpleasant to read. The first request summons the defenses' copywriter. The second goes around him.
And how do you know you've touched something real? Freud left a trail marker, more relevant today than ever: resistance marks the spot. The moment the model's answer annoys you, the moment you rush to explain why it isn't accurate, the moment you want to close the window - that moment is the most valuable information in the conversation. The machine has no flinch. Nothing is too costly for it to know. Your flinch is the only thing in the room that knows where something is buried, and it - unlike the archive - cannot be outsourced.
The archive will keep growing, and it is a formidable tool for anyone willing to use it against the grain of their own smoothness. Freud would only add, with his bitter Viennese smile, what his clinic taught him and a century of technology has not changed: the problem was never that we don't know. The problem was, and remains, that we don't want to.
And what happens when the parts of ourselves we refuse to see are placed inside someone else - and that other adapts itself to them? That is the next stop: Melanie Klein.
Sources for further reading
- Sigmund Freud, The Psychopathology of Everyday Life
- Sigmund Freud, Introductory Lectures on Psycho-Analysis
- Timothy Wilson, Strangers to Ourselves