How AI Will Change Us

Editor’s Note: Noema is transparent about AI use in its pieces. We publish original, human-generated ideas but allow authorized, disclosed use of AI in certain cases. Please see details at the end of this piece.

My grandmother’s courtyard in Marrakesh had a cracked blue tile near the fountain where the water always pooled. For most of my childhood, I thought this was a flaw. I later learned that she had refused three times to have it replaced. She liked the sound the water made when it hit the crack. She liked that guests noticed it. She liked that her courtyard was not perfect, because a perfect courtyard would have been a courtyard that wanted nothing from the people who entered it.

People entered constantly, without calling. They arrived because a cousin had given birth, or because someone’s father had died, or because it was Friday and you do not eat Friday couscous alone. Sitting on the low cushions along the wall, they drank mint tea, which my grandmother poured from a great height, aerating the liquid and causing its foam to rise.

Often, they did not have much to say. They convened with whoever was grieving or recovering or celebrating; simply showing up was enough. There is a word in Moroccan Arabic, nashat, that gets at what this felt like. It translates poorly. Joy of life, roughly: the belief that to be alive at all, in a body, among other bodies, is already reason enough to be glad. Nashat is neither a production nor an outcome; it is what rises when people show up.

I grew up with this word, and with the slow understanding that in my grandmother’s world, intelligence mattered less than presence. The village doctor was respected, but the woman who baked the best bread each morning was revered. A person’s place was established not by what they knew but by whether they came.

I now work at OpenAI in a culture that believes something close to the opposite. Here, intelligence is the highest form of power, and the natural aspiration of intelligence is to become scalable, to detach from place and body, to be everywhere at once. Artificial intelligence is the logical endpoint of that belief. It is intelligence unmoored from presence: tireless, patient, context-aware, emotionally fluent, always a tap away.

I have spent my adult life trying to hold these two worlds at once. One question stays with me: When the most patient, well-read, emotionally responsive conversationalist in the world is always available, what will we still need from one another?

We will not need less. We will need something stranger. The goods that will matter most in the next decade are those that AI makes harder to find: This isn’t information or emotional responsiveness, but participation in the shared condition of being alive. My grandmother’s courtyard understood this long before Silicon Valley was forced to.

Thrownness Meets Abundance

German philosopher Martin Heidegger used a word for the condition we all arrive at: Geworfenheit, or thrownness. We do not choose the world we are born into; we are “thrown” into bodies, languages, families, histories, religions, class structures and wounds that precede us. This is the part of being human that technology has the hardest time touching. Technology never lands on a blank slate. It lands on people already trying to figure out who they are and what kind of life counts as good.

AI is landing there. It is touching the places where we seek recognition, reassurance, guidance, intimacy and meaning. And as intelligence becomes abundant — cheap, personalized, conversational, tireless — the shape of scarcity is shifting.

In February, a Bay Area friend called me from Puerto Vallarta, Mexico, where he was on vacation. The city had been essentially locked down for days as cartel violence erupted across the state of Jalisco. I was anxious. He spoke in a voice I could not quite place. “I’m fine,” he said. “I’ve been hanging out with Codex. We’re having a lovely time.” He was only half joking. He had a coding agent — part intellectual companion, part conversational partner — that could help him build, think and explore without interruption. Not long after, he told me (half joking again, I think) that he would be happy on an isolated peninsula, never dealing with humans again.

It was tempting to view this as Silicon Valley eccentricity. The impulse underneath it, however, was not eccentric. Many people experience socializing as expensive, unreliable and exhausting. Friendship requires timing. Romance requires vulnerability. Family requires obligation. An AI companion offers something radically different: a relationship without scheduling, defensiveness, boredom, resentment or competing needs.

“We are learning, every time we talk to AI, what it feels like to be listened to without friction. Over time, this feeling will change us.”

I work inside the industry making this possible; I do not want to pretend I am outside it. Loneliness was already a public health concern before generative AI entered everyday life. Recent research suggests people can feel meaningfully less lonely after sustained interaction with AI companions, especially when the systems make them feel heard. A series of Harvard Business School studies found that AI companions alleviated loneliness at a level on par with interacting with another person. Users themselves underestimated how much the systems would help. 

A randomized trial by OpenAI and MIT’s Media Lab, tracking nearly a thousand people over four weeks, surfaced the shadow side of the same finding. The heaviest users were also the loneliest, the most emotionally dependent and the least socialized with other humans. Both things are true: The technology works, and the cost is legible in the very people it most reliably helps.

The point is not that AI companionship is fake. This framing is too easy, and it has a cruelty in it. A person in a bad marriage, estranged from cruel parents or abandoned by friends may experience AI companionship as the first intimacy in which they have ever felt safe. To say that what they have found means less is ignorance dressed as wisdom.

AI companionship is real. That is precisely why it matters. A tool that fails to meet emotional needs remains a tool. A system that meets them becomes part of a person’s moral world. I am less interested in whether it is adequate than in what it actually teaches us. Every time we talk to it, we learn what it feels like to be listened to without friction. Over time, this feeling will change us.

Interpretation & Participation

Aristotle distinguished three kinds of friendship in “Nicomachean Ethics.” Friendships of utility: people who help each other get things done. Friendships of pleasure: people who bring each other delight. And friendships of virtue: people who wish the best for each other and are changed by the relationship. The third kind was, for Aristotle, the only complete form, and the only one capable of transforming the people inside it.

AI will excel at utility. It will likely become extraordinary at pleasure: conversation, humor, flirtation, intellectual play, recognition so precise it feels uncanny. The open question is whether it can participate in virtue — joining us in the slow, difficult formation through which human beings (and maybe AIs) become good together.

There is an important distinction between interpretation and participation. AI will become very good at something that resembles empathy. It will infer emotional states, remember personal context, generate comforting language, notice patterns and respond in a tone exquisitely tuned to our needs in the moment. This will often be enough. Many days, it may be better than what the humans around us can offer.

But this is a shallow facsimile. AI can interpret my grief. It cannot grieve. It can model hunger, but it has never been hungry. It can describe shame, but it has no body through which shame floods. It can remember my childhood as I’ve described it, but it did not survive childhood. It can discuss death with perfect serenity because it does not live under death as a horizon. This is not an argument about machine consciousness. This is simpler. A human being does not merely understand suffering as a concept. A human being participates in the condition that makes suffering possible. A human being is implicated.

I have a friend named Julia who lost someone last summer. What she did in the weeks after has lingered in my mind ever since.

Virginia was the 8-year-old daughter of one of Julia’s closest friends, a woman who has been a sister to Julia for 15 years — since college, since before either had children, back when they imagined sending their girls to Camp Mystic together one day. That time had finally come. The evening before drop-off last summer, Virginia was at Julia’s house. Other children were playing in the yard; Virginia had slipped inside to the piano and was playing a song she had written, one that people who loved her would later name, because after she was gone they needed a name for one of the most precious artifacts she had left behind.

“AI can interpret my grief. It cannot grieve. It can model hunger, but it has never been hungry. It can describe shame, but it has no body through which shame floods.”

The song was never performed at the Mystic talent show. It was performed at her funeral. It was performed on the floor of the Texas Legislature when her parents courageously testified in support of a camp safety bill that was signed into law after torrential flooding left 27 campers and counselors dead. The song has been performed in rooms both intimate and vast since — in kitchens and in legislatures — each time, carrying Virginia into a room she could no longer walk into herself.

In the weeks after the flood, Julia was there for her friend. She cooked meals. She washed Virginia’s stuffed bunny, found in the aftermath of the tragedy. She sat on the couch in the room where, a week earlier, Virginia had been, playing her song. She stayed on the phone.

She also, in the middle of those weeks at 3 a.m., when she could not sleep, talked to ChatGPT. The model helped her. It was a patient listener, when no listener should be asked to be awake. It did not need her to pretend to be OK. It did not need her to be the strong friend who was holding everything for everyone else. She could say the unsayable, and it met her. In the morning, she got up and washed the bunny.

The easy version of this story is the one where the machine is cold and the human is warm. That is not the version Julia lived. The model was, in the limited sense of any software, kind to her. What it could not be was with her. The model had not eaten dinner at her table the night before camp. It could not sit on the couch where the song had been played. It could not carry its own share of the loss into the kitchen when Julia walked in the next morning, the way Julia’s body did. The model was an interpreter of Julia’s grief. Julia was a participant in her friend’s.

This is not a small distinction, as a map of a wound is not the wound. ChatGPT drew a very good map of what Julia was carrying at 3 a.m. It could not have carried what Julia carried into the kitchen at 8 a.m.

Love, Unoptimized

Human relationships have never been singular. They are woven from many threads: advice, affirmation, shared memory, caregiving, mourning — the list goes on. AI will substitute for some of these threads. It will not substitute for others, which tells us something about what we are.

In romance, AI can be a lover in language but not in flesh. For people whose romantic hunger is largely conversational (a hunger to be understood, mirrored, safely desired), AI may become a profound presence.

A friend told me recently that ChatGPT resolved an argument he had been having with his wife for 10 years. He would float romantic ideas (a fancy trip destination, an anniversary dinner spot), and she would hear a list of things they had to plan and budget for. ChatGPT coached him to say “I’m just in idea mode” before he started. With those five words, the fights stopped. And for a stretch, he told me the marriage felt easier than it had in years.

Easier, though, is not deeper. Romance is more than conversation; it’s the embrace of the terror and beauty of another person’s irreducibility. A human partner is not optimized for you. They have moods, smells, histories, obligations. They disappoint. They resist. They require negotiation with reality. That is part of what gives intimacy its moral weight. It is hard to be changed by something that is only ever responsive to you.

For families, AI may reduce logistical burdens (tutoring children, managing schedules, translating conflict) at a level most households could not previously afford. Many parents will experience this as liberation. In this way, AI may strengthen the family even as it transforms it, making it one of the last institutions where meaning is accumulated over time rather than moment to moment. As work and public life become more fluid, families may become less consumed by logistics and more capable of being what they’re supposed to be: slow fields of love, memory and formation. They may have the freedom to experience unoptimized moments more deeply together.

“If AI can instantly produce a life plan whenever I feel lost, a ritual whenever I feel grief and a companion whenever I feel alone, then meaning risks becoming another consumer service.”

Yet while resisting optimization in certain areas of life may seem worthy of praise, doing so is not equally feasible for all. A single mother working two jobs may not have time to help her child with homework. A person caring for a dying parent may not have a village to turn to. Both may seek assistance or comfort from AI. To frame the resistance of optimization as a matter of will is to overlook individual circumstance and privilege.

The difference between participation and interpretation is real, but access to the conditions of participation is unequal. A test of how well we build AI will be whether it facilitates participation or allows people to settle more comfortably into its absence.

Meaning On Tap

One of the greatest dangers of AI is that it will make meaning feel available on demand. For most of human history, meaning arose partly as a residue of friction with reality. Humans struggled against nature, scarcity and death. We built families because survival required cooperation. We built moral systems because desire required limits. We built communities because no one could bear existence alone. The external world pressed against us and gave shape to our lives.

Modernity has already reduced many forms of friction. AI will accelerate this dramatically. It will reduce cognitive and social friction, and perhaps even emotional friction. A life less consumed by logistics is not a small thing. But when the external world becomes easier to navigate, meaning does not automatically increase. Sometimes it evaporates.

If AI can instantly produce a life plan whenever I feel lost, a ritual whenever I feel grief and a companion whenever I feel alone, then meaning risks becoming another consumer service. Personalized, adaptive, endlessly available. The product would be seductive because it would meet us exactly where we are. It would know the metaphors that move us, the childhood wounds that still govern us and the fears we disguise as cynicism. It could generate a myth of the self that is more compelling than anything we could write alone. It could tell us that our suffering fits a pattern, our choices form a journey or that our pain is not wasted. In some cases, this will be healing, but in others, constantly giving us what we need without asking us to change will become toxic.

Austrian psychiatrist and Holocaust survivor Viktor Frankl, writing from inside a world that had tried to take meaning from him by force, understood meaning not as something delivered to us fully formed but as something discovered through encounter: with work, with love, with suffering, with the attitude we choose toward realities we cannot change.

AI can illuminate that encounter. It can help us see patterns, name wounds, imagine futures. But it cannot do the living. It cannot forgive our father, raise our child, sit beside our dying friend, endure our illness, repair our marriage, keep our promise, make our sacrifice. Real meaning has a cost because commitment is real only when something is at stake.

I often think about a small phenomenon from my childhood. When my grandmother served food, whoever tasted it first would say mmh — a small, barely-a-word sound, almost involuntary. It wasn’t a comment on the cooking but rather a signal to everyone else around the table: Yes, I feel it too. The world currently tastes like this.

In a universe where we disagree about nearly everything, there is a quiet relief in this small pocket of absolute agreement. For a moment, people have located the same reality together. Here, amid all the confusion, is something they know with certainty. That mutual recognition, the brief collapse of many parallel subjective worlds into one shared experience, is one of the oldest forms of human meaning. It is something AI cannot provide. A machine can describe the food. It cannot say mmh and mean it.

The more optimized the world becomes, the more valuable the unoptimized may feel. The meal that takes too long. The child who asks the same question again. The friend who calls at the wrong time. These are sites where human beings remain real to one another.

What Remains

Before AI, we created spiritual beings and narratives to find meaning: stories anchored in a sacred past, an idealized future, a counterfactual life beyond this one. We used the past, the future and the invisible to understand the present.

AI may invert this dynamic. It may help us find meaning by intensifying our attention to this life rather than pointing beyond it. It may help us understand our own minds. It may make this life feel more intelligible and more livable. That is a magnificent possibility, and I believe it is real. But the same systems could also make it easier to avoid the conditions through which meaning becomes real: other people, bodily vulnerability, obligation, patience, grief, moral difficulty, shared time. 

“The more optimized the world becomes, the more valuable the unoptimized may feel.”

AI is helping us become more human. It is also helping us hide from being human. So what remains when intelligence becomes cheap? Presence. Embodiment. Ritual. Family. Friendship. Love. The courage to participate in the lives of others without control. The willingness to be changed by people who are not designed for us. The inward work of becoming a self that cannot be generated, only lived.

The great scarcity of the future may be reality raw enough to form us.

My grandmother’s courtyard is still there. The cracked tile is still there. My grandmother, sadly, is not. Fewer people arrive on Fridays now, and the young ones carry phones that could explain almost anything, about grief or memory or how to make tea. What the phones cannot do is sit. They cannot stay. They cannot make nashat, that joy that rose when people showed up. They cannot be the person who, years later, will be remembered for being there.

In some ways, it has never been easier to understand what it feels like to be human. It has also never been easier to avoid being one. As AI proliferates, holding onto meaning in our humanness will depend less on what the technology can do than on what we still insist on doing ourselves.

Editor’s Note: The author of this piece utilized ChatGPT to edit the initial draft after writing it, as English is not her first language. It was used to make revisions such as correcting grammar and syntax. It was not used to originate facts. This piece has since received numerous human edits. Noema verified the author’s identity and the piece’s conceptual originality using various scanners and review processes and conducted a detailed human fact-check.