AI, Memory, and the Strange Illusion of Being Known
Opening
Over the last months, I have been exporting pieces of my digital life: conversations with AI systems, publication lists, drafts, CSV files, JSON files, and fragments of old projects.
It is a strange experience. A little like packing boxes before moving house. You discover forgotten things. You find traces of yourself. But you also notice something uncomfortable: even when the archive is large, it is still not the person.
An archive can remember what was said, but it cannot know why it mattered. AI can help us retrieve, organize, and sharpen our memories. But an archive is still only a trace. It is not the life that made it.
The archive contains words, dates, drafts, corrections, metadata, even emotion in written form. But it does not contain the full weight of a moment: fatigue, hesitation, grief, pride, shame, love, or silence.
I. The new private archive
Many of us are now creating personal archives without fully realizing it.
Emails, notes, voice messages, photos, AI conversations, Substack and WordPress drafts, LinkedIn and Facebook posts, browser histories, exported files.
We used to think of an archive as something for historians, governments, churches, or companies. Now ordinary people also carry archives. They are just hidden in devices and cloud accounts.
AI changes the meaning of this. For the first time, these archives can talk back.
That is fascinating. It is also dangerous if we forget the difference between retrieval and understanding.
II. The seduction of being recognized
AI can create a powerful illusion: it seems to “know” us. It remembers tone, notices recurring themes, summarizes our projects, and imitates our style.
But being recognizable is not the same as being known.
A system may identify that I write about dignity, governance, Thailand, Europe, grief, responsibility, and technological change. That is useful. But it does not mean it knows the lived path behind those words.
It can see the pattern, but it cannot carry the wound. Its usefulness depends on proper boundaries.
III. The archive has no body
An archive has no body. It does not age, and it does not wake up at night. It does not feel the humidity in Isaan, the winter in Alsace, or the pressure in a boardroom before a difficult decision.
It contains traces of experience, but not the body that passed through it.
It cannot see through “two lenses”: my Western lens organizes, classifies, exports, and stores. My other lens, from my lived experience in rural Thailand, reminds me that memory also lives in gestures, places, rituals, voices, and absences.
The archive is vertical: folders, files, dates, categories. It is rational.
Memory is more horizontal: a smell, a face, a song, a field after rain, a sentence one cannot forget.
IV. The danger of polished reconstruction
AI does not only retrieve; it also smooths.
This is helpful when a text is confused. But it is dangerous when the roughness contains truth.
A polished sentence may be less honest than the imperfect one. A clean summary may remove the hesitation or doubt that mattered. A balanced paragraph may hide the moral discomfort that gave the text its force.
This is where the essay becomes a reflection on authorship. The question is not whether AI may assist. It may, and it already does. The question is whether the human author still recognizes himself in the result.
V. Co-creation without surrender
A good AI assistant can help find what was forgotten, structure what is scattered, translate without flattening, and challenge the author without replacing him. It should help the text breathe, not seal it under glass.
But it becomes dangerous when it begins to sound too complete, too certain, too smooth.
So the real question is: how can AI help without slowly taking over my voice?
Co-creation without surrender begins with a simple rule: the human author must remain the final witness. AI may suggest, compare, structure, translate, and challenge. But it has not lived the life behind the words. It has not sat in the room. It has not carried the silence after a difficult sentence. It has not paid the price of the decision.
That means I must stay awake while using it.
I have to know what I am asking for. Am I asking AI to correct grammar, make a paragraph clearer, test an argument, find repetitions, or translate without flattening? These are different tasks. If I ask vaguely, I should not be surprised when the answer moves too far away from me.
I also need to protect the rough places in a text. Not every uneven sentence is a mistake. Sometimes the hesitation is the truth. Sometimes the short sentence carries more weight than the elegant one. Sometimes a small imperfection is the place where the human being is still visible.
Then I must read the AI version against my own inner ear. Does this still sound like me? Would I say this in a conversation? Is the sentence clearer, or only more polished? Has the text become more precise, or merely more impressive?
These questions matter more than style.
AI should be a mirror, not a mask. A mirror can show me what I missed: repetition, weak structure, missing links, or an argument that does not yet stand. A mask is different. A mask begins to speak in my place.
That is where the danger lies. AI becomes useful when it helps me return to what I wanted to say. It becomes dangerous when it tempts me to accept fluency as truth.
So I need a few practical habits.
I should keep my original draft. I should compare versions. I should mark sentences that feel too perfect. I should ask AI to simplify before asking it to beautify. I should ask what it removed, not only what it added. And when a sentence feels emotionally important, I should decide myself.
This is not resistance to AI. It is responsible use.
VI. Closing
The archive is useful. The archive matters. Without it, many things disappear. But the person remains more than the archive.
I will continue to use AI. I will continue to build my archive. I will continue to ask machines to help me find, compare, translate, and structure.
But I do not want to confuse the box of letters with the life that wrote them.
The archive can help me remember, but it cannot remember for me.
Robert F. Tjón, June 2026
For deeper reading:
Marshall McLuhan, Understanding Media: The Extensions of Man (1964). McLuhan’s phrase “the medium is the message” is often used to describe how communication technologies reshape perception and social life, not merely transmit content.
Alfred Korzybski, Science and Sanity (1933). The phrase “the map is not the territory” points to the difference between a representation of reality and reality itself.
Andy Clark and David Chalmers, “The Extended Mind” (1998). Their argument suggests that tools such as notebooks, and by extension some digital systems, can become part of human cognitive processes.
“Lifelogging” refers to the practice of digitally recording aspects of daily life. It is often discussed in relation to wearable devices, personal data archives, and quantified-self practices.
Paul Ricoeur’s work on narrative identity is relevant here, especially the idea that personal identity is shaped through time, memory, and narrative reconstruction.
Legend
AI — Artificial Intelligence.
CSV — Comma-Separated Values, a table-like data file.
JSON — JavaScript Object Notation, a structured data format often used for exported conversations or archives.
Content is subject to the ©️ CC BY-NC-ND 4.0 INTERNATIONAL license.
This piece first appeared on Substack. I republish it here voluntarily — not as repetition, but as a trace; a place where words can rest after their first flight.
👉🏻 rftjon.substack.com
