The Human as Placeholder

Water Is Still Water, Humans Are Still Human

Imagine a glass of water. Clear and transparent, it becomes steam when heated and ice when cooled. The skin registers entirely different sensations in each case: wet and scalding, or hard and freezing. When the temperature returns to normal, it flows once more as a liquid. From this, we might easily draw a simple conclusion: its state has changed, but water is still water.

That conclusion is only half right. The phase transition does not turn H₂O into another compound, yet the arrangement, motion, and relationships of its molecules have changed. Ice can bear weight, water can irrigate, and steam can drive a piston or scald the skin. The same substance, entering different structures, acquires different capacities and different dangers. We recognize water not only because its composition remains the same, but also because its structure and pathways of transformation remain intelligible to us.

The Milesian school of ancient Greece used the concept of archē to inquire into the origin of all things. What makes a thing remain what it is, and what makes it change? These two questions have been entangled from the very beginning.¹ More than two thousand years later, we are still asking the same questions, only now we have replaced water with the human being.

Can AI create genuine art? Can robots love? Where does the boundary lie between humans and their creations? Debates of this kind often share an unspoken premise: that the human is already complete. A standard, singular version of humanity seems to stand at the end of history, armed with reason, language, tool use, or moral judgment, checking the credentials of every newcomer. Yet the embodied human, the political human, and the laboring human are each only thin slices cut from a complex whole. Human physiology, sensibility, social relations, and inner life interlock; remove any one dimension, and what remains becomes easier to measure but further from the living person.

This is why I prefer to say: the human is a placeholder. By "placeholder," I do not mean an empty shell waiting to be filled, still less that humans can be arbitrarily replaced. Rather, I mean an open status of personhood: a status not monopolized once and for all by the species Homo sapiens, but sustained jointly by bodily vulnerability, interests that endure over time, recognition by others, memory, and responsibility. How this status is carried forward through artifacts and institutions, and how it is being redistributed in the age of artificial intelligence, are the questions we should truly care about.

Homo sapiens in the biological sense has a clear evolutionary lineage and bodily boundaries. Personhood, by contrast, is the position we call the "subject." To enter that position means that a being is no longer merely an object to be controlled, priced, and replaced. It possesses history, memory, and identity, along with the rights and obligations that follow. It can become the bearer of injury, promise, honor, and guilt.

Reflexivity and futurity are two threads that lead toward this position. A being that can place itself within a picture of the world, and allow what has not yet happened to constrain the present, becomes capable of promising, waiting, preparing, and betraying. Yet these two threads must not be turned into new thresholds. Infants, people with dementia, and patients in comas may not be able to demonstrate these capacities at a given moment, yet we do not revoke their personhood. Conversely, an automated system, however fluent in talking about selfhood and the future, may not possess a body that can be harmed, interests of its own that persist over time, or the capacity to bear risk. Personhood is not a one-dimensional test; it always depends on multiple conditions, and on recognition within relationships.

The Human Boundary Does Not End at the Skin

The question thus moves deeper: where, exactly, are the conditions that sustain personhood located? Intuition points us inside the skull, as if memory, will, and personality all resided in a sealed chamber built of neurons. But the human boundary has never been drawn by the skin alone. It is more like a field of relations that continually expands and contracts, distributed across bodies, other people, artifacts, institutions, and environments.

In the Zulu tradition, Ubuntu is often summarized as "a person is a person through other persons." The Spaniard Marcos Rodríguez Pantoja spent much of his childhood cut off from human society, living among animals in the Sierra Morena mountains. After returning to society in 1965, he struggled for many years with language and social adaptation, and he died on August 15, 2026.^2,3 His case is often invoked in discussions of language and socialization, but it cannot on its own prove how personhood is generated. Marcos had contact with humans in early childhood, and as an adult he was able to narrate his own experiences. A more careful conclusion is this: general biological conditions do not automatically generate the language, self-narrative, and social capacities we take for granted. The calling, answering, naming, and correction of others are also part of that process.

Nor is it only other people who shape us. The hunter reads an animal's direction from tracks and broken branches; the farmer reads the seasons through phenology, calendars, and seeds; the scribe preserves thought through letterforms, stroke order, and paper. In using these external things, humans are shaped by them in return. How sensations are distinguished, how movements are completed, and how intentions take form often depend on the scales and resistances that artifacts provide.

Watch a potter throwing at the wheel, and this becomes remarkably concrete. The lump of clay lands at the center of the wheel; the hands close around it, elbows braced against the body. As the wheel turns, what the palms feel is not abstract clay but a succession of subtle shifts in alignment, moisture, centrifugal force, and resistance. A little more pressure from the fingers, and the wall may rise or collapse; a little less water, and the clay abrades the skin. The skilled potter does not first complete a shape in the mind and then command the hands to reproduce it. Judgment happens in the back-and-forth among hand, clay, water, and wheel.

The archaeologist Lambros Malafouris calls this process material engagement. In his study of the potter's wheel, agency can be attributed neither entirely to the potter nor simply transferred to the clay; it emerges through the coupling of brain, body, tool, and material.^4,5 This is not to endow clay with a will of its own, but to acknowledge that intention is not always fully formed before action begins. The texture of materials, the scale of tools, and the habits of the body jointly determine what can be thought, done, and left behind.

A potsherd, then, is not merely a broken vessel. Its curvature, thickness, firing temperature, fingerprints, and trimming marks are traces left by a system of movements that has vanished. We can analyze its composition, date it, and reconstruct its form, yet still not know how its maker sat, how they steadied their body, the rhythm with which they kicked the wheel, or at what moment they decided the piece was complete.

Helmuth Plessner described the human as a form of life marked by "natural artificiality." We have no fixed nature capable of determining us once and for all; we can only keep organizing ourselves through language, technology, and institutions.⁶ Human incompleteness is not a defect. It is precisely what leaves room for culture to happen.

Cultural Memory Is a Relation of Re-enactment

Personal memory depends on the nervous system, but also on photographs, notes, smells, and repeated retelling. Cultural memory, however, must survive the passing of generations. Jan Assmann calls the memory sustained by everyday conversation communicative memory; it typically spans three or four generations. For a past event to travel beyond that range, it needs writing, images, rituals, architecture, monuments, and institutions of transmission: forms that turn fleeting recollection into something later generations can re-enter.⁷

Artifacts do not speak on their own. They offer shapes, resistances, sequences, and clues, giving those who come later a chance to retrace a path. Only when a ritual is performed again each year, a text reread by each generation, or a score sounded anew under different fingers does memory pass from potential into present experience. Cultural memory never seals the past in its original state. Every era reorganizes the past from the materials left behind. Tradition is therefore less a vacuum-sealed package mailed from antiquity than a protocol of re-enactment that is continually recalibrated.

Micronesian stick charts offer an elegant example. Slender coconut-leaf midribs or wooden sticks are lashed together in a lattice; shells mark islands, and curved strips represent swells, currents, and the refraction of wave patterns around land. The chart does not look like a modern map, and it is usually not taken aboard. Young navigators first memorize its structure on land; once at sea, they watch the wave crests with their eyes and feel the boat's pitch, drift, and impact with their bodies, calibrating these sensations against the relational map held in memory.⁸ The chart does not navigate on anyone's behalf. It trains an embodied intelligence, one that can be touched by the ocean and can understand that touch.

Bernard Stiegler used the term tertiary retention to describe technical memory externalized from the body. Experience first leaves traces in perception and is then reorganized in recollection; when memory is carved into stone, written on paper, recorded on discs, or stored on servers, it acquires a third mode of existence that outlasts the individual lifespan. Each recording technology also reshapes attention: writing allows us to look back, compare, and annotate; recordings let the same performance be heard again and again; short videos cut experience into a different tempo altogether.^9,10

Borrowing the ancient Greek word pharmakon, Stiegler reminded us that technical memory is always remedy and poison from the same source. The calendar coordinates a community in time, but also marks those who defy its rhythm as deviant; archives make evidence traceable, yet catalogs and classifications determine which lives are easily seen. Artifacts do not merely obey their users' intentions. They solidify the preferences of their makers and institutions into the environment of those who come after.

Cultural memory is therefore bound up with the distribution of power. Who is entitled to leave a name, and who appears only as a number? What counts as an original, and what gets filed as a copy or forgery? Whose oral account requires written corroboration, and whose documents count as evidence by default? The archive is not the sum of the past but the portion that survives after selection, naming, arrangement, and maintenance. Registers determine who is a resident; medical records determine who is a patient; household registration determines who can be recognized in a given place. Archives preserve subjects and produce the types of subjects that can be recognized.

The Convention for the Safeguarding of the Intangible Cultural Heritage, adopted in 2003, attempts to correct the habit of equating culture with antiquities. It places practices, expressions, knowledge, and skills alongside their associated tools, objects, and cultural spaces, and stresses that heritage is constantly re-created by communities in response to their environment and history.¹¹ Tradition stays alive precisely because inheritors have the right to bring their own time into the act of transmission. Preserving only a standard version may freeze a practice at the very moment it was discovered, nominated, or filmed by an institution.

The shikinen sengū of Japan's Ise Grand Shrine offers a seemingly paradoxical example. Every twenty years, the main sanctuary is rebuilt on an adjacent site according to its original dimensions, and the sacred vestments, furnishings, and treasures are made anew. If authenticity were tied entirely to the same batch of materials, this would amount to the periodic destruction of a monument. But if culture is understood as a continuity sustained by dimensions, craft, ritual, the organization of labor, and teaching across generations, then the rebuilding itself is preservation.¹² Wood decays, and so does skill. Young craftsmen must do the work with their own hands while there are still teachers to consult, so that certain judgments that cannot be written into blueprints can pass through time.

The Ise experience cannot be turned into a universal formula. Its lesson is that cultural continuity relies both on things and on society's capacity to remake, use, and reinterpret them. For different objects, the balance among originals, re-enactment, and living transmission must be struck differently; no single rule can decide it.

Digital Objects Have Not Escaped Materiality

In the digital age, artifacts seem to have finally escaped decay. A file can be copied endlessly, every copy bit-for-bit identical; cloud storage appears more reliable than wood, paper, or human memory. Yet the digital preservation guidelines of the U.S. Library of Congress repeatedly point out that bit-level preservation is only a starting point. Even a file that survives intact may become unrenderable when its format, codec, software, license, or hardware environment becomes obsolete.¹³

An email depends on character encodings and servers; a work of net art depends on browsers, plug-ins, and interfaces that have already been shut down; the memory of a game may reside in controller latency, multiplayer cooperation, and the slang of its player community. Taking screenshots or recording the screen proves that a work once existed, but does not necessarily preserve how it responded to its participants. Digital objects are not without materiality; they have simply hidden their dependencies inside a longer technical chain.

The longer that chain grows, the more preservation resembles continuous maintenance. Migrating formats, emulating old systems, and retaining source code and licenses can solve only part of the problem. If operational experience, community context, and rights relations disappear along the way, the surviving file may remain a shell that can be verified but can no longer be re-enacted.

Artificial Intelligence as a New Tertiary Retention

From this perspective, we should not rush to call large generative models "the memory of all humanity." Models do absorb the statistical structure of vast quantities of text, images, sound, and code, and can recombine them in forms approximating human expression. But there is a crucial difference between a model and an archive: an archive strives to maintain the identity, provenance, and context of its materials, whereas model training transforms materials into parametric relations. We can ask a model about the style of a poem, but from its answer alone we can scarcely trace which texts contributed to the parameters, or what each source underwent in tokenization, filtering, and training.

A large model is more like a new rock formed when layers of cultural sediment are subjected to heat and pressure. It retains certain structures while erasing countless boundaries. Compression brings the convenience of instant retrieval, but also makes sources and paths of transformation hard to discern. The question is not whether this new rock is "authentic," but whether it still allows us to trace the materials within it and to recognize which voices never entered the sediment in the first place.

When machine-generated output flows back in bulk as training material for subsequent systems, original records and derivatives become mixed together. Research on model collapse shows that indiscriminate recursive training on generated data causes information in the tails of the distribution to vanish first. Later work has specified the conditions more precisely: when real data is retained, and synthetic and real data continue to accumulate together, collapse is not inevitable.^14,15 Synthetic data cannot simply be dismissed as contamination, but sources and transformation processes must be clearly marked, and systems must keep encountering new material from the world.

This matters especially for low-resource languages, rare experiences, and expressions that do not fit mainstream formats. Already at the edges of the data distribution, they are the first to slip away in filtering, compression, and retraining. Cultures that can be scanned, transcribed, translated, and fed into training sets appear ubiquitous; knowledge still held in bodies, dialects, taboos, seasonal labor, and closed rituals is easily judged "insufficient data." Retrievability once again masquerades as existence itself. Large models did not invent this power; they merely extend the archive's authority over "what gets recorded" into the system's authority over "what gets generated, and with what probability."

If we regard artificial intelligence as a new form of tertiary retention, then the question is not only how much it remembers, but what kind of attention it trains, what modes of questioning it encourages, which materials it allows to be regenerated, and which experiences it deprives of a passage into the future. Generation lowers the barrier to making, but may also allow the most common forms to dominate the imagination in advance; personalization helps people find what they need, yet the shared world may grow thinner as a result. Remedy and poison still flow from the same mechanism.

And here lies a boundary that efficiency must not be allowed to obscure: accessible does not mean freely trainable, and preservable does not mean that something may be stripped of context and generated without limit. Some knowledge is suitable for open circulation; some passes only within particular relationships. Some materials may be preserved but not rewritten, commercialized, or used for retraining. Provenance, consent, withdrawal, and negotiation are not appendices to the work of memory; they determine whether transmission can continue to exist as a relationship at all.

Leaving the Future a Watershed

To say the human is a placeholder is not to hollow out human meaning, but to acknowledge that personhood is always in the process of becoming. Stripped of language, tools, institutions of care, and a shared world, an isolated biological individual can hardly retain the humanity we know. Artifacts have long since entered into us, becoming part of our movements, memory, and responsibility; humans, in turn, continually enter artifacts, leaving in them the classifications of institutions, the preferences of eras, and the habits of bodies.

What, then, should we leave to the future? Artifacts, certainly. Without originals, those who come later lose the material's resistance to contemporary interpretation. We must also leave provenance: who made a thing, who used it, how it circulated, what repairs and transformations it underwent, and what rights and disputes cling to it. Provenance determines whether future people can distinguish originals from derivatives, and whether they can copy, conjecture, and question.

The operating environments of artifacts, along with the practitioners who sustain them, must be preserved as well. For pottery, that means clay sources, kiln firing, tools, gestures, and the relationships between masters and apprentices; for wave charts, the islands, winds, swells, hulls, and training; for digital works, the code, systems, hardware, interfaces, and licenses. Environments cannot be sealed intact, but they can be documented and simulated, and the knowledge and operational conditions needed to rebuild them can be retained. Without living people to maintain them, archives grow moldy, links break, standards expire, and languages lose their last speakers.

The future also needs the right of reinterpretation. Communities are not data sources waiting for experts to extract information, and inheritance is not the mechanical reproduction of a standard version. Those who come after should be able to question the classifications we leave behind, change the modes of re-enactment, and keep space open for new subjects. At the same time, every preservation plan carries omissions. Acknowledging the unknown is more honest than filling in the blanks with fluent speculation. The unknown is not inventory awaiting technological conquest; it is also the past's way of refusing to be wholly naturalized by the future.

Return to that glass of water from the beginning. It can freeze, vaporize, and flow again. But if the river's course is cut off, the glaciers polluted, and the cycle between cloud and sea broken, then even a glass of chemically pure water sealed in a display case cannot show the full meaning of water in ecology and culture.

The artifact is that glass of water, precious and necessary. Craft, language, community, environment, and the freedom of reinterpretation constitute its watershed. Preserve only the sample in the glass, and the future inherits a relic. Only if the watershed can still hold tributaries, sediment, evaporation, and rain will the future receive not a dead relic, but a set of relations that can keep on growing.