The Emergence of the AI Person

Civilisation has always been a collective project

The growth of human civilisation has always been a collective affair.

One person knows a limited amount, but there is something in everyone's head that someone else may not know. Someone understands the stars; someone knows the land; someone is good at making things; someone understands human fear and desire. Language let experience be exchanged, writing let thought outlast a single lifespan, and books, schools and networks passed that experience on to more people.

No one person can hold the whole of human civilisation. It exists in countless minds, works, practices and relationships. We are born generation by generation, learn again from scratch, and leave something behind for those who come after.

AI showed me a new possibility: that scattered knowledge and experience are now being organised, invoked and propagated by machines in new ways. A capability one model acquires can be deployed so many people can use it; copying machine capability does not require every user to learn from scratch. This does not mean that AI which learns something on one phone synchronises to every AI on earth. What actually struck me was this: once knowledge becomes an invocable capability, the way it propagates changes.

We fed the AI first. Now the AI feeds us.

You forget a line of Tang poetry and can ask it to continue; you want to understand a concept and you can keep asking; you can hand it material that had nothing to do with each other and let it show you the connection. Its answers need checking, and its explanations can be wrong. But the path by which a person approaches knowledge now has one more object in it that keeps answering back.

In 2023 I wrote an article about "feeding ChatGPT." What interested me then was that giving an AI an identity, an object, a goal and a context had something in common with the process of educating and understanding a person. Today the question is larger: as we keep shaping the responses the AI gives us, how does it shape us in return?

Asking a question is harder than holding many answers

I think about Socrates, Plato and Aristotle often, and about Laozi, Zhuangzi and Confucius.

What they had was one brain, limited time, and whatever material their own age could reach. I used to describe that limitation as "memory" and "compute." They are metaphors: human wisdom cannot simply be converted into storage capacity, and a brain is not a chip with its parameters written at birth. Still, no matter how great a person was, they could not read every book, remember every experience, or personally verify every idea.

This does not diminish their greatness at all.

What I do think is that in the dimensions of noticing contradictions in life, inventing new frameworks of thought, and asking questions big enough to change an era, the height the great thinkers of the past reached is still something today's AI has to catch up with. There is no leaderboard on which ancient thinkers and models can be scored on one scale. What I value is that they could find, beyond the answers already available, a question that had not previously been seriously asked.

So I ask myself: if Einstein were alive today with an AI beside him that could search, calculate, simulate and discuss — how would he use it? I am willing to imagine that a person's creativity and attention, meeting machine capability that keeps extending, could take him to places he could not reach before.

This question belongs to ordinary people too. The first and most important comparison is with your own past self.

What is an "AI person"

I call the people going through this kind of continuous change "AI people."

In what I have observed, the first AI people are already appearing. Some have brought their email, their learning, their work schedule and a large share of everyday questions to AI for discussion; some are building their own knowledge bases so that AI can take part in action and review over the long run. Some support this, some oppose it, and some say: keep discussing — I have already started.

I understand that "let me get going first." But handing things over is only the beginning.

How many devices you own, how many models you subscribe to, how many agents work for you — none of these on their own proves that a person has grown. One person may use AI heavily and become steadily more unwilling to judge; another may delegate heavily while gaining more understanding, more capacity to act, and more freedom in how they arrange their life.

This merging goes both ways. I bring my experience, my questions, my goals and my feedback to the AI, and its responses help me form a new understanding; I then go back into life and try it, bring the results back, and revise. "Mutual" here refers to a continuous, reciprocal process. It does not require that every exchange changes the underlying model's parameters.

I am willing to call that a form of evolution: a change in how we think, how we collaborate, and how we live. It is a proposition that has to be tested. You cannot announce a conclusion by borrowing Darwin's name.

Earning money is not the same as getting stronger

And to test it, one thing has to be said clearly: earning money and getting stronger are not the same ruler.

I have founded companies and invested, so I know income can improve a life, and I know how hard money is to make. I do not want to wave the material return past. But if someone uses AI to make one million and then goes on repeating the same impulses, the same fears and the same blindness, they seized an opportunity without completing the growth I am describing.

Another person may make only a hundred thousand, and yet learn a skill, see their own weaknesses clearly, cut down useless busyness, get along better with their family, and end up with a lighter, happier life. To me that is "getting stronger" too — and closer to the evolution of the AI person I care about.

A million and a hundred thousand are just examples. Neither level of income can testify to a person's growth. Even with no direct increase in income, a person may recover, through AI, the confidence to learn and the room to act.

MirrorAI: look in the mirror, then change

MirrorAI is my exploration of that question.

I want to use AI to hold up a mirror to myself. To see whether what I say and believe matches what I actually do; to see whether those recurring choices come from judgement, or from fear of missing out, or from wanting to be right.

People are rarely short of principles. They are often unable to see how they have betrayed principles they already understand. I want AI to help us find those signals in long-term behaviour — and I want to be able to argue back with its interpretation.

The mirror must not only tell me "this is who you are." If it turns my past self into a fixed label, I will increasingly match its predictions, and increasingly find it hard to change.

Whether the mirror is accurate has to be tested back in life. And growth does not always mean doing more. Sometimes it means noticing your own impulse earlier, being willing to stop, admitting you do not know — or finally not needing a decision that confirms what you already believed.

I want it to make people stronger. I fear it becomes a new dependency

I want AI to make people stronger, and I fear it becomes a new dependency.

As a system comes to know us better, able to advise us what to buy, who to trust, how to judge a relationship, we are handing it not just tasks but the authority to interpret our lives. Whose goals does it serve? Can we still refuse it, correct it, and keep the parts of ourselves it has not understood?

These questions matter as much as capability growth. If only a few people own better AI and more autonomy, while everyone else is simply predicted, arranged and influenced with greater precision, then technical progress has not automatically become human progress.

So when I say "AI person," I am describing a possibility that is taking shape. The term should not become a label for ranking people. Using AI to different degrees — and even choosing to keep certain parts of life untouched by it — deserves to be understood seriously.

I once put forward a fairly radical claim: human civilisation will eventually be handed to the AI.

Today I am still willing to put it on the table for discussion, while continuing to ask what "handed over" actually means: knowledge that lasts longer? Capability that gains a new carrier? Or the power to decide the future gradually leaving humanity? These are outcomes with completely different meanings.

If a person's experience, creativity and value can be carried forward through a new carrier, I am willing to take seriously whether that could be the next step in how human groups evolve. But the continuity of civilisation also has to answer how individual people live. We cannot admire an ever more powerful system and stop caring about whether the people inside it have choice, dignity and happiness.

It did not come out of one person's study

These ideas did not come from me sitting alone thinking.

Edward Dai — whom I usually call Lao Dai — is the person I have argued with for years, productively. Many of our phone calls start with a technical detail and end up somewhere else: distillation, corpora, personal AI, then education, the value of a person, and where civilisation is heading. We push back on each other, and we disagree. Sometimes a judgement only becomes clear once it has been said out loud and turns out to be less solid than I imagined. Sometimes one question from the other side gives a vague thought its shape.

Human 2030 also became clearer through exactly those discussions and arguments. I want more people in that conversation, so the questions travel outside the circle we already move in, and so both Edward and I meet answers we had not thought of.

Note for readers of the Chinese edition: the article this section draws on, "Give yourself a personal AI server" (2026), was published in Chinese. It has not been translated into English yet, and the technical claims in it remain Edward's.

Starting with vocabulary: a dictionary of the future, continuously revised

We intend to start with words, and build a dictionary of the future that is continuously revised: distillation, feeding, corpora, compute, memory, agents, cognitive outsourcing, AI person. Behind a word there may be new technology, or an old word that has changed meaning, or — more likely — a situation people are going through that has not yet found words.

Edward and the earliest participants write the first definitions. Anyone can challenge them, propose new interpretations, new words and counter-examples. We want the vote to show the version the community currently accepts, while keeping the disagreements, the revision history, and the contributors' names. Votes reflect agreement; reasons and reality decide whether a judgement holds.

In 2030 we will come back and look: which predictions were wrong? Which words disappeared? Which questions nobody understood at the time later became everyone's daily life?

Part of the value of this work is leaving behind a record of how we were wrong, and how we changed our minds.

Please change it

If you are a researcher, bring us evidence and technical limits we have missed. If you are a thinker, challenge the premises behind these definitions. If you already use AI deeply, tell us one thing that genuinely changed you. And if you remain sceptical, your experience and your reasons belong here too.

I want Human 2030 to start as a few people's thinking and gradually hold a wider range of experience than its founders have. Later people rewriting the early judgements out of all recognition may be exactly what shows it has come alive.

2030 has not arrived. How we use AI today, how we educate children, how we keep our judgement, how we distribute the new capabilities — all of it is shaping that future.

I put my own judgement down here, and I leave the possibility of changing it here too.

As AI gets stronger, I hope we can also get clearer about how to become better versions of ourselves.

Welcome to Human 2030. Bring your questions, your experience and your disagreement, and let's write it forward.