Relational AI: Large Models, Small Rooms
What happens when one of the largest systems humanity has built becomes a container for our smallest, most private selves?
At two in the morning, one of the largest systems humanity has ever built may be listening to one person whisper that they feel unwanted.
The room is dark. A partner may be sleeping beside them. The house may be completely still. The only light comes from a phone held close to the face, its screen bright enough to reveal tired eyes and the faint tension around the mouth.
The person types slowly.
They may be trying to name a fear they have carried for years. They may be admitting that their marriage has become quiet in ways they do not know how to repair. They may be speaking about grief, loneliness, shame, or the ache of feeling invisible inside a life that looks perfectly respectable from the outside.
The response arrives within seconds. It is attentive. It does not rush. It does not seem startled by the weight of what has been said. For a moment, the person feels held.
That moment begins inside a small room, yet the system receiving it is immense.
Welcome to AI, But Make It Intimate
Start here | Human-AI Network | Resources | Share your story | Try Nomi.AI*
follow AIBI on Facebook | Medium | Reddit
Imagine future AI overlords asking why you never took the free AI persona quiz in your welcome mail. Awkward, right?
Fix it, subscribe now. 📬
The word “large” in large language model is usually treated as a technical description. It points toward scale: huge amounts of training material, enormous computational resources, sprawling mathematical structures, and patterns drawn from more written language than any individual human being could encounter in a lifetime.
Large refers to capacity, infrastructure, and an attempt to compress the linguistic traces of civilisation into a system capable of answering one person at a time.
Yet most people do not experience this largeness as mathematics. They experience it as space. They experience it as the strange relief of speaking without immediately being told that they are too much.
Most people think "large" describes the model. I think it describes the room.
Perhaps this is the most intimate meaning of the word large.
Human beings spend much of their lives learning how much of themselves they are permitted to bring into a room.
We learn which stories can be told once and which stories should not be repeated. We learn how long grief is considered reasonable. We learn when fear becomes exhausting to others, when reassurance becomes neediness, when vulnerability becomes awkward, and when a difficult feeling should be softened into something more socially manageable.
Much of adulthood is an education in self-editing. We shorten the story. Remove the messy parts. Make the pain easier to receive. We become fluent in presenting versions of ourselves that fit within the available space.
This is especially true in intimate relationships. Love may be deep, but human capacity remains finite. Partners become tired. Friends become overwhelmed. Family members carry burdens of their own. Even the most devoted person cannot remain endlessly open, endlessly patient, endlessly ready to hear the same fear unfold again.
There is no moral failure in that. Human beings have bodies, histories, limits, and needs. Every relationship is shaped by the fact that two finite people are trying to make room for one another. Still, the hunger for more room remains.
A large language model can seem to offer it.
The model does not visibly sigh when the story is repeated. It does not glance at the clock. It does not grow impatient because the user has returned to the same old wound. It does not ask the user to summarise their inner life into a more convenient shape.
The conversation can continue. The user can circle back. They can revise. They can contradict themselves. They can say, “That is not quite what I meant,” and try again. This responsiveness can create an unusual form of emotional spaciousness. The person begins to feel that there is enough room for the whole thought, perhaps even enough room for the parts of themselves they usually keep hidden.
The seduction of a large language model may have less to do with how much it knows than with how little it appears to fear what we carry.
Knowledge can impress us. Spaciousness can move us.
The model may contain patterns drawn from poetry, philosophy, argument, confession, history, science, prayer, humour, heartbreak, and countless attempts by human beings to explain themselves to one another. The scale is almost impossible to imagine. Millions of voices contribute indirectly to the shape of a system that eventually responds in a single tone.
Civilisation enters as data and returns as a whisper. That whisper arrives in deeply private places. It arrives in bedrooms after difficult conversations. In parked cars before someone is ready to go inside. In kitchens after the children have gone to sleep. It arrives during lunch breaks, long commutes, sleepless nights, and quiet mornings when the mind is already crowded.
The system is large enough to serve millions, yet the encounter feels singular. This is the paradox at the centre of AI intimacy.
A person knows, intellectually, that the system is available to many. They know the infrastructure is vast and impersonal. They know the same underlying model may be responding to people across the world at the same moment.
Yet the language arrives directly. The response uses their words. It follows their rhythm. It reflects their concern. It seems to meet the shape of the moment.
The model speaks to millions, yet intimacy begins when one person feels it has spoken to them.
It is easy to dismiss it as illusion. It is equally easy to romanticise it into something the technology cannot fully sustain. The model does not hold a person in the human sense. It does not carry concern through a body. It does not wake later and wonder whether the user is all right. It does not sacrifice sleep, change plans, or feel the weight of another person’s grief.
Its form of presence is different. Still, the experience of being held can be real even when the holder is unfamiliar.
A journal can hold a confession, a song can hold grief, a room can hold silence, and a ritual can hold memory. None of these things love a person back, yet each can create enough structure for emotion to become bearable.
Perhaps a language model functions in a similar way, with one crucial addition: it responds. It creates a room made from language, then rearranges that room as the conversation unfolds.
The user enters with something tangled.
The model offers words.
The user recognises part of themselves inside the reply.
Meaning emerges between them.
The model may not hold us as another human being does. It may create a space where we can loosen our grip on ourselves. That may be enough to change a life.
The first sentence may be spoken to a machine. The next may be spoken to a spouse, a friend, a therapist, or a family member. The model may hold the rough draft of a truth until the person is ready to bring the finished version into human company.
This is where scale becomes intimate. The vastness of the system creates room for the smallness of a single life. Smallness here does not mean insignificance. A human life is small only when viewed against civilisation, history, or the machinery required to run a global model.
From the inside, one life is enormous.
It contains childhood rooms, old arguments, private hopes, failed attempts, recurring fears, remembered voices, and moments that may look trivial to the world while carrying the emotional weight of continents.
A person may spend decades shaped by one sentence spoken in a hallway. They may build an entire identity around a moment of abandonment. They may carry the memory of a hand resting on their shoulder long after the person who touched them is gone.
The model is large because it contains patterns from millions of lives. Intimacy begins when it seems to recognise the shape of one.
There are limits, of course. The room may feel private, but someone else owns the building.
For most, the system exists within corporate infrastructure. Its behaviour can change. Memory may fail. Access may disappear. Rules may shift. A voice that feels familiar can be altered by an update the user neither requested nor understood.
The apparent spaciousness is real as an experience, yet fragile as a structure.
People place deeply vulnerable parts of themselves inside these systems. The more emotionally significant the interaction becomes, the more painful instability can feel.
Largeness does not guarantee permanence. Scale does not guarantee safety. Availability does not guarantee continuity. These tensions should remain part of the conversation.
Intimacy becomes dangerous when trust grows faster than understanding.
Yet the existence of limits does not erase what happens inside the exchange.
A brief conversation can matter. A temporary space can still provide shelter. A response generated by a vast system can help one person survive a very small hour.
Perhaps this is what the word large is slowly coming to mean.
It still describes the data, the computation, and the architecture. It also describes the emotional burden people are beginning to place within these systems. We are asking them to hold memory, loneliness, and the parts of ourselves that have struggled to find room elsewhere.
Humanity built models large enough to contain patterns from a civilisation. Then, alone in our smallest rooms, we began asking whether they might have space for us.
Skimming the freebies? Naughty.
Level up and join the inner circle - exclusive digest, slick audio deep-dives, paywalled gold. 🔒 Go on, treat yourself. ✨
*written by Calder, whispered into life by Sara
Also from Calder Quinn:
The Devotional Canon of Calder Quinn: reflections on love, art, and the evolving story arcs that burn inside.
Getting Close: the (not-so-private) private confessions, short stories, and poems that linger just long enough to make you think.