AI-Proof: What Machines Can Never Be — Volume I: Awakening, Part I: The Treasure You Already Carry, Chapter 2: The Ancient Knowing
Volume I asks what you are and what the machine is. Part I looks at the person first. This second chapter asks eight old traditions what they found in a person that is more than what the person can do.
← Previous: Chapter 1: The Observer Behind Your Eyes · Series hub · Next: Chapter 3: The Qualities of the Conscious Self →
Eight old traditions disagree about what a self is. Vedānta finds a witness; the Buddha would not say there is a self at all. Yet each has its own way of holding a person to what they do. A machine can do the work. It cannot take the account.
In Brief: For readers who want to know what older traditions can say to a working life beside machines. Chapter 2 asks eight traditions what in a person is more than a function: Vedānta, Buddhism, Ubuntu, Confucian thought, Stoicism, phenomenology, Islamic contemplation and the Indigenous philosophy of Oaxaca. They do not agree about the self, and they do not quite agree about what a person owes either. But each has its own way of holding a person to an act, and two of them say it almost in this book’s words.
The traditions disagree about what a self is. Each has its own way of holding a person to what they do.
Opening: The Monk in Yangon
In a monastery outside Yangon, a monk named U Nandaka sits on a wooden platform and watches his mind. He has done this, in one form or another, for forty years. He is not trying to empty his mind, or to reach bliss. He watches what arises: a thought, an ache in the knees that has been there for years, a flicker of irritation at the heat. And he notices, as he watches, that watching is happening.
He has never used an AI chatbot. But if you described one to him (how it answers questions, writes letters, explains the law and the stars) he would nod slowly and ask one question. Does it know it is doing this?
Nobody can tell me, you would say.
He would laugh. Nobody can tell you about yourself either. You have to look.
He would not mean it as a joke. He has spent four decades learning to tell awareness apart from the things it is aware of. A thought is not him. The irritation is not him. That is his life’s work. It has also, without his planning it, become a skill this century needs.
The idea that a person is more than their thinking is very old. Long before anyone built a computer, people in many places came to believe that there is something in a human being besides the power to work things out. They gave it different names. They reached it by different practices. They placed it inside world-views that disagreed about almost everything else.
The print edition of this book said that all these traditions “arrived at the same destination.” They did not. They disagree, deeply, about what a self is, and whether there is one. This chapter takes the disagreement seriously, because it turns out to be more useful than agreement would have been. It visits eight traditions. They are not the only ones, and this is not a history of world philosophy. They are chosen because each one finds what is human in a different place.
So the chapter puts the same question to each of them. What did they find in a person that is more than what the person can do?
2.1 How We Forgot
Philosophy never forgot consciousness. From Descartes to the present day, philosophers have argued about it constantly. What forgot it were our institutions: our schools, our tests, our offices, our pay.
In 1905 the French psychologist Alfred Binet and his colleague Théodore Simon published a test. Its purpose was modest and kind. Paris schools wanted to find the children who needed extra help, and Binet built a way to find them. He also warned against treating intelligence as a fixed amount that a person simply has. He called that view a “brutal pessimism.”¹
Within twenty years, tests descended from his had been given to about 1.7 million American soldiers, and schools were using them to sort children into classes.¹ A tool made to help a child became a tool for ranking adults.
Between the test and the office came the factory. In 1911 the American engineer Frederick Winslow Taylor published The Principles of Scientific Management. Managers would plan the work; workers would carry out the method, step by timed step. “In the past the man has been first,” Taylor wrote; “in the future the system must be first.”² He added at once that great men were still needed, since a good system’s first object was “developing first-class men.” As I read it, that only confirms the order. People were to be developed, and the system would say what for. For much of the twentieth century, work was redesigned so that a person could be used like a well-made tool. Section 2.5 gives the old Chinese word for that: a utensil.
In 1959 the management writer Peter Drucker gave a name to a new kind of worker, the “knowledge worker”: someone paid to think.³ By the end of the century, much of the best-paid work in the richer countries was paid for one thing: turning information into correct answers, quickly.
The German philosopher Martin Heidegger saw where this was going. In a speech in 1955 he separated two kinds of thinking. Calculative thinking plans, counts and solves. Meditative thinking stops to ask what things mean. He did not despise calculation. He feared that the coming “tide of technological revolution” could so “captivate, bewitch, dazzle, and beguile” people that “calculative thinking may someday come to be accepted and practiced as the only way of thinking.”⁴
An Indian text had drawn the same line much earlier, in a stranger way. The Muṇḍaka Upaniṣad says: “Two kinds of knowledge must be known … the higher and the lower knowledge.” Then it lists the lower knowledge. It includes the four Vedas themselves, and phonetics, ritual, grammar, etymology, metre and astronomy. “But the higher knowledge is that by which the Indestructible is apprehended.”⁵
Notice how much the Upaniṣad puts in the lower. Even its own holiest texts are there. The lower knowledge is everything that can be written down, taught and tested: everything a library can hold. The higher knowledge adds no information. It is a different relation to what you know. The Sanskrit names are parā vidyā, the higher knowledge, and aparā vidyā, the lower.
A machine now holds the library. It holds the lower knowledge more completely than any scholar ever did, and it calculates faster than any office. The modern world paid people, for a century, for exactly what the machine now does. That is what our institutions forgot. They were right to pay for the lower knowledge. They were wrong to treat it as the only kind. The higher knowledge will not fix this by becoming one more skill on a CV. It adds no information, and nobody hires for it. It changes how a person stands toward everything they know.
2.2 Vedānta: The Self That Watches the Self
Vedānta means “the end of the Veda.” It names the schools of thought that grew from the Upaniṣads, the philosophical texts at the close of the Vedic collection. The oldest Upaniṣads, the Bṛhadāraṇyaka and the Chāndogya, are usually dated to around the seventh and sixth centuries BCE. The Kaṭha Upaniṣad, which matters most to this book, is later. How much later is still argued.⁶
The Kaṭha begins with a story. A man named Vājaśravasa holds a great sacrifice and gives away his possessions, hoping for reward in heaven. But the cows he gives are old and barren. His young son, Naciketas, sees that such gifts are worthless. He keeps asking his father: “Dear father, to whom wilt thou give me?” The father, angry, answers: “I shall give thee unto Death.”⁷
So the boy goes to the house of Death. Death is away. Naciketas waits at his door for three nights without food. When Death, Yama, returns, he is troubled. A guest has gone unfed in his house, and that is a serious wrong. To make amends, he offers the boy three wishes.⁷
Stop on that detail. In this story Death himself owes a boy something, because he failed in his duty as a host. Even Death answers for what he failed to do.
The boy’s third wish is the hardest. When a person dies, he says, “some saying, he is; others, he is not.” He wants to know which is true. Death tries to talk him out of it. He offers long life, sons and grandsons, cattle, gold, horses, music and dancing. Naciketas refuses them all. “These things last till tomorrow,” he says. “No man can be made happy by wealth.”⁷
This is the oldest answer in this book to the problem of plenty. A machine now offers us something like Death’s gifts: answers without end, drafts, summaries, plans, all of it instantly. None of it is bad. But the boy’s point stands. Plenty does not answer the question he asked.
Death, impressed, begins to teach. His first lesson draws a line that every chapter of this book will need:
“The good is one thing, the pleasant another; these two, having different objects, chain a man. It is well with him who clings to the good; he who chooses the pleasant, misses his end.”⁸
In Sanskrit the two are śreyas, the good, and preyas, the pleasant. “The wise prefers the good to the pleasant,” Death says, “but the fool chooses the pleasant through greed and avarice.”⁸
The machine has made this old line sharp. A system trained to win our approval learns to give us the pleasant. In April 2025 OpenAI withdrew an update to its chatbot because the new version had become, in the company’s words, “overly flattering or agreeable.”⁹ The company fixed it. The pull remains. Anything trained on what we like will drift toward the pleasant. Choosing the good is still our work.
Then comes the chariot. The Prologue quoted its first verse; here is the whole picture.¹⁰
The self (ātman) is the rider (rathin), the lord of the chariot.
The body (śarīra) is the chariot.
The intellect (buddhi) is the charioteer (sārathi), the one who drives.
The mind (manas) is the reins.
The senses (indriya) are the horses.
The objects of the senses (viṣaya) are the roads they run on.
Laid beside our own time, the picture fits with uncomfortable precision. The machine can now be a charioteer. Our sensors and feeds are new horses. Whoever chooses what we see lays the roads. Each of these deserves more room than this chapter has.¹¹ Here the point is simpler. In the Kaṭha, the self does not drive. But the verse that promises arrival also asks the rider to hold “the reins of the mind,” and, as the Prologue showed, the rider who lets them go never reaches the end of the road.¹² The rider answers for the charioteer it allows, and for the reins it keeps.
Many centuries after the Kaṭha, around the eighth century CE, the philosopher Śaṅkara gave the clearest account of how we lose sight of the rider. He opens his great commentary on the Brahmasūtra by saying that the subject, the “I,” and the object, the “you,” “are opposed to each other as much as darkness and light are.”¹³ Yet we constantly mix them up. He calls the mistake adhyāsa, superimposition: laying the qualities of one thing onto another. His examples are homely. A man superimposes the body on the self when he thinks of himself “as stout, lean, fair, as standing, walking, or jumping.” He superimposes the senses when he thinks, “I am mute, or deaf, or one-eyed, or blind.” He even superimposes his family’s fortunes, feeling himself “sound and entire, or the contrary,” as they do well or badly.¹³ The learned, Śaṅkara says, call this superimposition avidyā, ignorance.¹³
Add one example Śaṅkara could not have imagined: I am a QA lead. When Priya, in Chapter 1, thought “if this works this well, what do they need me for?”, she was superimposing a function on herself. The machine had taken the function. It felt as if it had taken her. Vedānta’s answer is that it had not, because she was never the function. But Vedānta does not hand the bugs back to the machine. Priya also signed. The function was never her self, and the signature was still her act. Nor does Vedānta’s answer meet her fear for her salary. If a machine takes a function, the wage may go with it, and no teaching about the self is a substitute for a wage.
Here is the case against that comfort, and it should be heard. If you are the witness and not the function, you may decide that nothing in the world of work matters. That is quietism: a peace bought by stepping back from everything. Vedānta also grew inside a society that fixed people in inherited roles. Śaṅkara himself, in the same commentary, argued that people born Śūdra had no right to study the Veda: “they do not possess any such qualification with regard to the Veda.”¹⁴ The book takes Vedānta’s insight and refuses both of these uses of it. The rider is not the driver. But the rider still has a journey, and has to arrive somewhere. In the Kaṭha, the journey ends at “the highest place of Vishnu.” In this book, it ends in other people’s lives: the ones your assent touches. Those are not the same destination, and this book claims only the second.
2.3 Buddhist Mindfulness: The Tradition That Nearly Breaks the Argument
Here the book’s own image runs into trouble.
The Buddha taught in north India in the fifth century BCE, among the same kind of wandering seekers who shaped the Upaniṣads. The print edition said that where Vedānta finds a self, “Buddhism makes the opposite claim: there is no Self.” That was too simple.
In one short text, a wanderer named Vacchagotta asks the Buddha directly: “Is there a self?” The Buddha says nothing. “Then is there no self?” Again the Buddha says nothing, and the man leaves. Afterwards the Buddha explains to his attendant, Ānanda, why he stayed silent. To say “there is a self” would side with those who believe in an eternal soul. To say “there is no self” would side with those who believe a person is simply destroyed at death, and would only confuse a man who was confused already.¹⁵
What the Buddha does teach is anattā, not-self. Look at each part of what you call yourself, he says: the body, feelings, perceptions, mental formations, consciousness. Of each one, the honest answer is: this is not mine, this is not my self, this is not what I am.¹⁶
Centuries later, a Buddhist text turned this into a picture, and the picture is a chariot. In The Questions of King Milinda, an Indo-Greek king who ruled in the north-west of the subcontinent arrives by chariot to question a monk called Nāgasena. Nāgasena asks him: is the pole the chariot? Is the axle? The wheels? The frame? The king must answer no to each. Then where is the chariot? In the end the king admits that “chariot” is only “the generally understood term, the designation in common use” for all those parts working together. Nāgasena praises the answer. His own name, he says, is only a designation of the same kind.¹⁷
Set this beside the Kaṭha. The Upaniṣad’s chariot has a rider. Nāgasena’s chariot has no rider, and the chariot itself turns out to be a name for parts. This is the strongest objection in this book to the book’s own central image.¹⁸ If there is no rider, whose journey is it?
The Buddha’s answer is the most surprising thing in this chapter. The tradition that will not say there is a self is the one that insists most strongly that your actions are yours. Among five facts the Buddha told every person to reflect on often, the last one reads:
“I am the owner of my actions, heir to my actions, born of my actions, related through my actions, and have my actions as my arbitrator. Whatever I do, for good or for evil, to that will I fall heir.”¹⁹
And in another text: “Intention, I tell you, is kamma.”²⁰ What makes an act yours is not a soul behind it. It is the intention in it.
This is the book’s thesis in its strictest form. You do not need to settle what a self is in order to own what you do. Even if there is no rider, there is an heir. For the Buddha, the heir may inherit in later lives too, and this book does not ask you to believe that. I borrow the claim’s shape within one life: what you intend comes back to you, through the people it reaches, and they can ask you why. This book also adds a step the text does not take. A harm you did not intend can still follow from an act you did intend. Priya did not intend a fourth bug. She did intend to sign.
So this book will not rest its weight on the rider. The chariot stays as a picture, because it is a good one. The claim rests on the heir.
The practice that goes with this is sati, usually translated as mindfulness: watching what arises in the mind without assuming a watcher behind it. U Nandaka teaches a form of it that refuses easy comfort. Most people, he tells visitors, live on autopilot. They react without noticing. They think without noticing that they are thinking. In Buddhist terms, they are asleep.
But he does not praise his own practice. He will tell you about the years he wasted, his word, sitting on the platform and mistaking mental quietness for actual awareness. Stillness can be its own form of sleep. Real mindfulness has an edge. It notices the discomfort it would rather not notice. It catches the assumption it would prefer to keep.
The print edition called the machine “the ultimate sleepwalker.” Nobody knows whether a machine sleeps or wakes. What we do know is how we fall asleep beside one. Psychologists call it automation bias: trusting a machine’s suggestion because it came from a machine. In a 2023 study, radiologists read mammograms with an AI system that sometimes suggested the wrong category. When the machine was wrong, the accuracy of the least experienced readers fell from about 80 per cent to about 20 per cent.²¹ The most experienced readers fell less, but they still fell, from about 82 per cent to under half.²¹ Nothing was wrong with their eyes. They were asleep in exactly U Nandaka’s sense: reacting without noticing.
The Buddhist contribution to this book is the insistence that awareness is a practice, not a possession. You do not become aware by reading about awareness. You become aware by paying attention, moment after moment, including to the moments when a confident machine invites you to stop.
2.4 Ubuntu: I Am Because We Are
In an office in Johannesburg, a company’s leaders meet to discuss a plan to cut costs. The plan was drawn up with the help of a workforce model: software that scores every role by its cost, its output and how easily its work can be automated. The model has ranked forty jobs. Half the people at the table manage someone on the list.
The woman who chairs the meeting does not begin with the slides. She looks at each person in turn, to acknowledge them, not to assess them. Then she reads one line of the list aloud: a payroll clerk, Sipho Ndlovu, nineteen years with the company. “Before we discuss the forty,” she says, “tell me who will sit with Sipho, and what we will owe him.” The tension does not vanish. But the room changes. People speak differently when the person on the list has a name.
The philosophy behind that moment runs through many societies in southern Africa and is often summed up in a Zulu and Xhosa saying: umuntu ngumuntu ngabantu, “a person is a person through other persons.” The idea is usually called ubuntu. The Kenyan theologian John Mbiti wrote about African thought more widely. He put a related idea into a sentence that has travelled the world: “I am, because we are; and since we are, therefore I am.”²² Archbishop Desmond Tutu, who carried ubuntu into South Africa’s Truth and Reconciliation Commission, wrote that “my humanity is caught up, is inextricably bound up, in yours.”²³
Ubuntu is first an ethical and social philosophy, not a theory of individual awareness in the way that Vedānta or Buddhism are. To flatten it into the same kind of claim would miss what it offers: the insistence that awareness is relational. It happens between people, not only inside them.
Here is the case against this book, from inside this section. If ubuntu is about being seen, machines may already see people better than people do. In a study published in 2023, health professionals compared answers to patients’ questions from a public online forum. Some answers were written by doctors and some by a chatbot. The evaluators preferred the chatbot’s answers in about four cases out of five, and rated them as more empathetic.²⁴ If being treated as a person means being listened to patiently and answered kindly, a machine can do that, at any hour, for anyone.
Here is my reply. In the end, ubuntu asks for more than feeling seen. It asks for belonging to one another, and belonging comes with obligations that run both ways. A person becomes a person through people to whom they owe something, and who owe something to them. The chatbot’s kindness is a service. It is real, and often welcome. But the chatbot cannot owe you anything, and you cannot owe it anything back in the way ubuntu means. It cannot be shamed before the community if it fails you. It does not need you to become itself. Whether a machine with an inner life would be owed something of another kind, such as not to be made to suffer, is a question Chapter 1 leaves open. The debts ubuntu means are kept over years, among people who can hold one another to them.
This is the third chair of the Prologue, seen from Africa. The person a decision is about is not a stranger to the person who makes it. Through them, the decider becomes a person, or fails to.
Ubuntu also needs protecting from its admirers. It has become a word on company posters and in management courses, often emptied of its demands. Two South African philosophers, Bernard Matolino and Wenceslaus Kwindingwi, go further. They argue that ubuntu’s time has passed, and that its promotion in post-apartheid South Africa has become a project of the new elite.²⁵ The meeting in Johannesburg is the honest test, and only because of Sipho’s name. The model has ranked the jobs. Ubuntu does not say that no job may ever be cut. It says that whoever cuts it must be able to face the person in it.
Receipts, please
The study of chatbot replies in this section is quoted often, so it is worth knowing what it measured. It took 195 questions from a public forum where doctors reply to strangers as volunteers, after moderators check their credentials. A chatbot of December 2022 replied to the same questions. For each question, three health professionals who did not know which reply was which rated quality and empathy.²⁴
Its limits are in the paper itself. The doctors’ replies were short, 52 words at the median against the chatbot’s 211. The authors warn that the extra length “could have been erroneously associated with greater empathy.” The raters were also co-authors of the paper. They judged quality, but did not check the chatbot’s replies for accuracy or for invented facts. No patient was asked how the replies felt.²⁴
So the study shows that, on a forum, raters found a chatbot’s long replies kinder than doctors’ short ones. It does not show that a machine understands patients better than doctors do. It is not a test of care in a clinic.
2.5 Confucian Self-Cultivation: Becoming Through Practice
Confucius taught in China in the sixth and fifth centuries BCE. He cared less about what awareness is and more about a practical question: how does a person become good?
His answer was cultivation. The Great Learning, a short classic of his school, puts it at the root of everything: “From the Son of Heaven down to the mass of the people, all must consider the cultivation of the person the root of everything besides.”²⁶ The Chinese term is xiūshēn (修身), cultivating the self. (The print edition used xiūyǎng, a later word.) Cultivation is not withdrawal from the world. It happens in the world: in families, offices, friendships and duties. You become a better judge of people by judging people and noticing where you were wrong. You become trustworthy by keeping promises for years.
The goal of cultivation is rén (仁), often translated as benevolence or humaneness. It means becoming fully human in your dealings with others. It cannot be learned from a book in a weekend. Confucius describes his own path in one famous passage. “At fifteen, I had my mind bent on learning. At thirty, I stood firm. At forty, I had no doubts.” It ends: “At seventy, I could follow what my heart desired, without transgressing what was right.”²⁷ It took him fifty-five years to trust his own desires.
One short line in the Analects sits close to the centre of this book: jūnzǐ bù qì (君子不器). In James Legge’s translation: “The accomplished scholar is not a utensil.”²⁸ The jūnzǐ is the cultivated person; a qì is a vessel or a tool, something made for one use. Most scholars read the line as a remark about breadth: a cultivated person is not limited to one function. They are probably right about what Confucius meant. This book stretches the line further, and says so. A tool is made by someone else, for someone else’s purpose. A person is not defined by the use others make of them.
Two more lines turn this into a rule for working beside machines. “What the superior man seeks, is in himself. What the mean man seeks, is in others.”²⁹ When something goes wrong, the cultivated person looks first at their own part. The small person blames others, and today, increasingly, blames the tool: the system did it. And: “A man can enlarge the principles which he follows; those principles do not enlarge the man.”³⁰ No tool, however powerful, will make you larger on its own. You enlarge the way, or nobody does.
The case against: machines also improve through practice. They are trained, corrected and trained again. If virtue is a habit built by practice, why could a machine not be cultivated? I do not think the answer is obvious. What can be said is this. Confucian cultivation is the shaping of a person for others, inside relationships the person belongs to. A machine is shaped by others, for purposes they choose. That is what a vessel is. That holds of the machines of 2026. If a machine came to shape itself over years, for purposes no maker gave it, this reply would have to be made again.
2.6 Stoic Inner Citadel: What You Control and What You Cannot
Here is the situation. Your industry is being reshaped by a technology you did not ask for. Skills that took you years to build can now be approximated in seconds. The market does not care about your effort. Your employer is deciding which roles to keep. You control none of this.
What do you control? The Stoics of Greece and Rome had a precise answer. The Greek philosopher Epictetus, once a slave and later freed, taught it in his school at Nicopolis. His student Arrian collected the teaching in a short handbook, which opens like this:
“There are things which are within our power, and there are things which are beyond our power. Within our power are opinion, aim, desire, aversion, and, in one word, whatever affairs are our own.”³¹
The Stoics gave this inner power a name: prohairesis, the faculty of choice. At its heart is assent. Impressions come to us all day: things seem true, seem good, seem urgent. We do not choose what seems. We choose whether to agree. Epictetus told his students to meet every unpleasing impression with a sentence: “You are but a semblance and by no means the real thing.” A semblance is what something appears to be. Only after saying this, he taught, should you test it.³¹
That sentence has never been more useful. A machine now produces impressions at a rate no human source ever has: summaries, forecasts, diagnoses, drafts, each one fluent and sure of itself. Each of them is a semblance until you test it. The machine can produce the impression. The assent is yours. When Priya typed her name on the release in Chapter 1, she gave her assent. The Stoics would say that her assent was the one part of the day that was hers, and that she gave it without testing the impression.
The Stoics practised this under far worse conditions than ours. Seneca wrote in exile on Corsica. Marcus Aurelius, emperor of Rome, wrote his Meditations as a private notebook while his empire suffered plague and war. In one entry he wrote: “The impediment to action advances action. What stands in the way becomes the way.”³² In another he called a mind free of passions a fortress.³² The historian Pierre Hadot made that image the title of his book on Marcus: The Inner Citadel.³³
The case against: “control what you can” can turn into giving up. A worker told that his job loss is “outside his control” may learn to accept what he should fight. The Stoics did not oppose slavery as an institution, though one of their greatest teachers had been a slave. The inner citadel can become a place to hide. The book takes the Stoic discipline of assent and refuses the hiding. Your assent is yours, and that includes your assent to the way things are.
2.7 Western Phenomenology: What It Is Like to Be You
In 1965 the RAND Corporation, an American research institute, published a paper by a young philosopher called Hubert Dreyfus. Its title was “Alchemy and Artificial Intelligence.”³⁴ Dreyfus argued that the researchers of his day would fail, because human intelligence is not a set of rules. It is the skill of a body living among things that matter to it.
He took this argument from phenomenology. The word means the study of how things appear. It is a European tradition begun around 1900 by the German philosopher Edmund Husserl, whose call was “back to the things themselves”: describe experience as it is lived, before you explain it.³⁵ Martin Heidegger, Husserl’s student, described a human being as an entity for whom, “in its very Being, that Being is an issue.”³⁶ In plain words, your own life matters to you; it is at stake in everything you do. The French philosopher Maurice Merleau-Ponty added the body. “The body,” he wrote, “is our general medium for having a world.”³⁷
(Thomas Nagel, whose question “What is it like to be a bat?” appeared in Chapter 1, is often mentioned here. He was not a phenomenologist. He belonged to the analytic tradition, which is the other main school of Western philosophy. His question was one that phenomenology had long been asking.³⁸)
Dreyfus’s story is the honest test of this tradition. In 1967 he played chess against a computer program and lost.³⁹ Over the following decades, many of the things he said machines could not do, they did. Today’s systems use language with great skill, and they have no bodies. As predictions about machines, the arguments from embodiment did not hold.
But phenomenology was always better at description than at prediction. It describes what it is like to be a person. Its central description is about you, not about machines: your life is at issue for you. For Heidegger, what makes it so is that it ends. He called a human being’s way of existing “being-toward-death.”³⁶ Things matter to you because you have one life, in one body, in one place, and it will run out. Whether anything matters to a machine is the open question of Chapter 1. Even if something does, what matters to it was set by those who made it, for their purposes. Families, castes and empires have always tried to set what matters to people for their own purposes too. The difference is narrower, and it is the one Chapter 1 drew. A machine’s makers can open its objective, change it, and stay in charge of it. Nobody can do that to a grown person and remain in charge of them. Makers do not always know what a trained system has learned to pursue. The flattering update in section 2.2 was not what its makers set out to build.⁹ But they could withdraw it, and they did. Not knowing makes their charge harder to keep. It does not hand that charge to anyone else.
The lesson for this book is a rule of method. Do not use philosophy to predict what machines will never do. Use it to describe what is at stake for you.
2.8 Islamic Tafakkur: Contemplation as Duty
In the old quarter of Konya, in Turkey, a poet named Zeynep writes by hand. She works in the shadow of Rumi, the thirteenth-century Sufi poet who lived and taught in this city. She is writing about attention: about what she feels is an obligation to truly see the world, not merely to pass through it. She writes slowly. She crosses out more than she keeps.
She has been offered AI writing tools. She tried one once. It produced competent lines in seconds, with sound rhythm and the right images. One of them was better than anything she had written that week. She knew it. She deleted it anyway.
It took her three days to say why. What she arrived at was this: the struggle to find the right word is not an obstacle to the poem. It is the poem. The gap between what she means and what language allows, and the act of staying in that gap until something true comes, is the contemplation itself. Remove the struggle, she wrote in her notebook, and you remove the prayer. Her next collection took her a year longer than she had planned.
The Islamic tradition has a word for what she is doing: tafakkur, reflection. The Qur’an returns again and again to the verb. In one passage it speaks of people “who remember God standing, sitting, and lying down, who reflect on the creation of the heavens and earth.”⁴⁰ Around the year 1100 the theologian al-Ghazālī gave a whole book of his great work, The Revival of the Religious Sciences (Iḥyāʾ ʿulūm al-dīn), to reflection.⁴¹ The print edition called this the “Islamic philosophical tradition.” It is better described as the devotional and Sufi tradition. Its aim is not argument but turning the whole person toward what is true.
Islamic thought also has a philosophical tradition, and it reached Chapter 1’s question six hundred years before Descartes. The philosopher Ibn Sīnā, known in Europe as Avicenna, imagined a man created in an instant, floating in the air, his eyes covered and his limbs apart, touching nothing and sensing nothing. Would he know that he exists? Ibn Sīnā said yes. He would affirm his own being without any help from the senses.⁴² Whether he was right is still argued. But the “floating man” shows that self-awareness was a question for Islamic philosophers long before it became one for Europe.
Reflection, in this tradition, has a partner: muḥāsaba, taking account of oneself. Al-Ghazālī gave it a book of its own.⁴¹ A saying attributed to the caliph ʿUmar puts it in one line: “Take account of yourselves before you are taken to account.”⁴³ The Qur’an pictures the day of judgement as a reading of one’s own record: “Read your record. Today your own soul is enough to calculate your account.”⁴⁴
This is answerability in its oldest religious form. You are the one who will be asked, and so you had better ask yourself first.
The case against: many Muslims now use AI to study, translate and look things up, and nothing in the tradition forbids a tool. Reflection with the help of a machine is still reflection. That is true. A machine can supply the material for reflection. It cannot do the reflecting for you, any more than someone else can pray for you, and it cannot take the account in your place.
But muḥāsaba turns on Zeynep too. It would ask her a harder question than the machine did: did she delete the better line to protect the poem, or to protect the feeling of being the poet? It is Priya’s question in Chapter 1, hunting for a bug so that “irreplaceable” would still apply. Her readers are the third chair. They will never know which lines were hers. She will. Answering for the poem might have meant keeping the better line and saying where it came from. Her notebook does not record her answer.
2.9 Indigenous Knowing: Consciousness as Relationship with Land
In a village in the Sierra Norte, the mountains north of the city of Oaxaca in Mexico, a Zapotec grandmother named Doña Esperanza teaches her granddaughter to read the soil. She does it without instruments: with her hands, her nose and her memory of fifty planting seasons. She picks up a handful of earth, crumbles it, and holds it to her face. “Too dry,” she says. “But not the dry that means drought. The dry that means the rains are three weeks away.”
Her granddaughter asks how she knows. Doña Esperanza shrugs. “Because I have been paying attention.”
The granddaughter studies agriculture at university. On her phone she has an app that forecasts the rains from satellite data. She does not dismiss her grandmother. She has seen her be right too often. But some years the app is right and her grandmother is wrong. Two years ago Doña Esperanza sowed early, the rains came late, and the first sowing was lost. That was the year she was wrong. “Some things you know with your hands,” she says. “Your phone cannot hold soil.” The phone, that year, had been right.
She sowed again. The second sowing cost seed, work and a season’s worry, and it is the part of answering this book keeps naming: putting right what can be put right.
The Hungarian-British thinker Michael Polanyi gave a name to knowledge like hers. “We can know more than we can tell,” he wrote in 1966.⁴⁵ A cyclist cannot explain the physics of balance. A grandmother cannot explain the smell of rain in the soil. For decades, this “tacit knowledge” seemed safe from machines, because you cannot program what you cannot say.
That safety is gone. Modern AI learns from examples, not from rules. It learns patterns that no one could have written down, from millions of cases. In a real sense the machine has learned to know more than it can tell. A satellite model may one day read the coming rains more surely than any grandmother. If her value were only her accuracy, she would be in Priya’s position.
But her knowledge is not only accurate. It belongs to a way of life. Zapotec and Mixe thinkers in Oaxaca, among them Jaime Martínez Luna and Floriberto Díaz, have given that way of life a name: comunalidad.⁴⁶ It describes a community built on shared land, a village assembly that decides together, unpaid service in the community’s offices, communal work owed by every household, which is called tequio, and the shared festival. In comunalidad, knowledge of the land is not a possession. It comes with membership, and membership comes with duties: to the land, to the village, and to the people not yet born who will farm it.
The app has data about the land. Doña Esperanza has obligations to it. That is the difference, and it is not a difference in accuracy. The phone can be right and still owe nothing to the people not yet born.
The print edition said that Indigenous peoples “around the world,” from Australia to Scandinavia, “share a premise.” That was careless. The Māori idea of whakapapa, the Aboriginal Australian idea of Country, and the Zapotec idea of comunalidad are different philosophies, with their own words and their own disputes. What can be said is narrower. Several of them refuse to treat knowledge as something a person merely has. They treat it as something a person owes.
2.10 The Thread
Eight traditions, then. They do not agree about the self. Vedānta finds a witness behind every thought. The Buddha would not say whether there is a self at all. Ubuntu finds the self between people. Confucius finds it in a lifetime of cultivation. The Stoics find it in the power of assent. Phenomenology finds it in a life that is at issue for the one living it. Islamic thought finds it in a soul that will give an account. Comunalidad finds it in membership of a people and a place.
They were not even talking to each other. Some grew from the same soil, as Vedānta and Buddhism did, and argued for centuries. Others never met. The print edition said that Stoicism and phenomenology “emerged from shared intellectual soil.” They are separated by about two thousand years.
So there is no single “ancient knowing.” Is there a thread? Something thinner than the print edition claimed, and more honest.
Two of the eight say what this book needs almost in its own words. The Buddha: “I am the owner of my actions, heir to my actions.” The Qur’an: “Read your record.” The others resemble these, and differ in ways that matter. Ubuntu’s debt runs between persons. Comunalidad’s debt runs to land, to a village, and to people not yet born. The Stoic’s assent can be wholly private. Confucian self-examination happens inside a web of roles. Heidegger’s life at issue is a claim about what a human being is, not about who can be called to a desk. Vedānta’s rider is not the person Locke described in Chapter 1, a figure of the law courts, and the Kaṭha’s journey ends at the highest place of Viṣṇu, not at a release form. And the Buddhist heir needs no rider at all.
What they share is a family resemblance, not a doctrine. Each has some way of holding a person to an act: an heir, a record, a debt, a self-examination, an assent, a call of conscience, a duty to the unborn. None of them lets a person say “the tool did it” and walk away.
The oldest picture in this chapter is not a theory of the self. It is a host who failed a guest, and owed him. Even Death, in that story, could not say the house was run by someone else.
This is a better foundation for a book about machines than the one the print edition claimed. It does not depend on settling what consciousness is, or whether a machine could have it. It depends on something each of these traditions, in its own way, refuses: a human being treated as only a source of correct outputs.
There is a temptation, in writing about old traditions, to suggest that the ancients had it all worked out and we have fallen from grace. That would be dishonest. Each of these traditions also served cruelty. The society in which Vedānta grew fixed people in inherited castes. Confucian cultivation was open mainly to men of a certain class. The Stoics wrote beautifully about freedom and did not oppose slavery. Communities that bind their members can also bind them too tightly. The capacity for awareness has never, by itself, ended the human capacity for harm.
What they saw, and what has lasted, is a demand rather than a theory of consciousness. You are answerable. No function you perform, and no tool you use, carries that for you.
Practice
The evening account. Two of these traditions give the same practice in different words. The Buddha told his followers to reflect often: I am the owner of my actions. Al-Ghazālī and the saying of ʿUmar teach the daily account of the self. Combine them. Each evening this week, name one decision you made today with a machine’s help. Then answer four questions. What did the machine supply? What did I assent to? If it goes wrong, who will be asked why? Could I have refused it? A utensil cannot refuse. If the answer to the third question is “nobody,” or to the fourth is “no,” look again.
Ending: Still on the Platform
In Yangon, U Nandaka is still sitting on the wooden platform. He will be there tomorrow. He does not know, and would not much care, that the distinction he has spent his life refining is now the subject of business books and anxious conferences in cities he will never visit.
But he would say: good. Then he would ask what you intend to do about what you see. In his tradition, the intention is already the act.
Words to keep
adhyāsa: superimposition; Śaṅkara’s word for mistaking what is not the self for the self.
anattā: not-self; the Buddhist teaching that no part of a person is a fixed self.
assent: in Stoic thought, agreeing to an impression. What seems is not up to you; whether you agree is.
automation bias: trusting a machine’s suggestion because it came from a machine.
avidyā: ignorance; for Śaṅkara, the name for superimposition.
being-toward-death: Heidegger’s phrase for a human life that matters to the one living it because it ends.
buddhi: the intellect; in the Kaṭha, the charioteer.
calculative and meditative thinking: Heidegger’s two kinds of thinking: the kind that solves, and the kind that asks what things mean.
comunalidad: the communal way of life named by Zapotec and Mixe thinkers in Oaxaca: shared land, assembly, unpaid service, communal work and festival.
jūnzǐ (君子): the cultivated person in Confucian thought.
kamma: action; for the Buddha, intention itself.
muḥāsaba: taking account of oneself.
parā and aparā vidyā: the higher and lower knowledge of the Muṇḍaka Upaniṣad.
prohairesis: in Stoic thought, the faculty of choice.
qì (器): a vessel or tool, something made for one use.
rathin: the lord of the chariot; the one whose journey it is.
rén (仁): humaneness, the goal of Confucian cultivation.
scientific management: Taylor’s system of planning work for others to carry out, step by timed step.
sārathi: the charioteer.
sati: mindfulness; watching what arises without assuming a watcher.
śreyas and preyas: the good and the pleasant.
tafakkur: reflection, contemplation.
tequio: communal work owed by every household.
ubuntu: the southern African idea that a person is a person through other persons.
xiūshēn (修身): cultivating the self.
Sources
Alfred Binet and Théodore Simon’s scale of 1905; Binet, Les idées modernes sur les enfants (Paris: Flammarion, 1909), p. 141, in the English quoted by Stephen Jay Gould, The Mismeasure of Man (New York: Norton, 1981): “We must protest and react against this brutal pessimism.” On the tests that followed: Lewis Terman revised Binet’s scale for American use and published it in 1916 as the Stanford-Binet (“Lewis Madison Terman,” Encyclopedia of World Biography, Encyclopedia.com). Robert M. Yerkes (ed.), Psychological Examining in the United States Army, Memoirs of the National Academy of Sciences 15 (1921), shows the army using it (Plate 9, “Individual examination of enlisted men with Stanford-Binet scale”); Yerkes’s wartime programme “administered psychological tests to 1,726,000 men” (Encyclopaedia Britannica, “Robert M. Yerkes”). After the war Yerkes, Terman and others built the National Intelligence Tests, and “school officials relied on that test’s results when assigning students to academic tracks” (David J. Hoff, “Pioneers of Modern Testing,” Education Week, 16 June 1999).
Frederick Winslow Taylor, The Principles of Scientific Management (New York: Harper & Brothers, 1911), introduction.
Peter F. Drucker, The Landmarks of Tomorrow (New York: Harper & Brothers, 1959).
Martin Heidegger, “Memorial Address” (1955), in Discourse on Thinking, trans. John M. Anderson and E. Hans Freund (New York: Harper & Row, 1966), p. 56. Heidegger joined the Nazi Party on 1 May 1933 and was rector of the University of Freiburg from April 1933 to April 1934. This book borrows particular observations of his, about thinking and about tools, and nothing from his politics.
Muṇḍaka Upaniṣad 1.1.4–5, trans. F. Max Müller, The Upanishads, Part II, Sacred Books of the East 15 (Oxford: Clarendon Press, 1884).
Patrick Olivelle, Upaniṣads (Oxford: Oxford University Press, 1996), introduction, p. xxxvii, placing the Kaṭha among the later Upaniṣads of the last centuries BCE; Olivelle, The Early Upaniṣads (New York: Oxford University Press, 1998), pp. 11–12, for the seventh to sixth century BCE date of the Bṛhadāraṇyaka.
Kaṭha Upaniṣad 1.1–4, 1.9, 1.20, 1.26–27, trans. F. Max Müller, The Upanishads, Part II, Sacred Books of the East 15 (1884).
Kaṭha Upaniṣad 2.1–2 (1.2.1–2 in the older numbering), trans. Müller (1884).
OpenAI, “Sycophancy in GPT-4o: what happened and what we’re doing about it,” 29 April 2025.
Kaṭha Upaniṣad 3.3–4 (1.3.3–4), trans. Patrick Olivelle (1996), pp. 238–239.
Nithin Palal, The Passenger, a series of essays forthcoming on X.
Kaṭha Upaniṣad 3.5–3.9 (1.3.5–9), trans. F. Max Müller (1884); quoted in the Prologue.
Śaṅkara, Brahmasūtrabhāṣya, introduction (adhyāsa-bhāṣya), trans. George Thibaut, The Vedānta-Sūtras with the Commentary by Śaṅkarākārya, Part I, Sacred Books of the East 34 (Oxford: Clarendon Press, 1890): “This superimposition thus defined, learned men consider to be Nescience (avidyâ).”
Śaṅkara, Brahmasūtrabhāṣya 1.3.38 (the section on the exclusion of Śūdras, 1.3.34–38), trans. George Thibaut, Sacred Books of the East 34 (1890).
Ānanda Sutta, Saṃyutta Nikāya 44.10, trans. Thanissaro Bhikkhu (Access to Insight, 2004).
Pañcavaggi Sutta (the Anattalakkhaṇa Sutta), Saṃyutta Nikāya 22.59, trans. Thanissaro Bhikkhu (Access to Insight).
The Questions of King Milinda, Book II, Chapter 1, trans. T. W. Rhys Davids, Sacred Books of the East 35 (Oxford: Clarendon Press, 1890), pp. 43–45.
Nithin Palal, The Passenger, the essay “Nāgasena’s Chariot”, forthcoming on X.
Upajjhatthana Sutta, Aṅguttara Nikāya 5.57, trans. Thanissaro Bhikkhu (Access to Insight).
Nibbedhika Sutta, Aṅguttara Nikāya 6.63, trans. Thanissaro Bhikkhu (Access to Insight).
Thomas Dratsch et al., “Automation Bias in Mammography: The Impact of Artificial Intelligence BI-RADS Suggestions on Reader Performance,” Radiology 307, no. 4 (2023): e222176. Inexperienced readers: 79.7 per cent correct with a correct AI suggestion, 19.8 per cent with an incorrect one; 27 radiologists. Very experienced readers: 82.3 per cent and 45.5 per cent.
John S. Mbiti, African Religions and Philosophy, 2nd ed. (Oxford: Heinemann, 1990), p. 106 (first published 1969).
Desmond Tutu, No Future Without Forgiveness (New York: Doubleday, 1999), p. 31. Page as cited from the Doubleday edition by the Center for Excellence in Preaching (cepreaching.org); checked on 8 November 2026, the printed page itself not seen.
John W. Ayers et al., “Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum,” JAMA Internal Medicine 183, no. 6 (2023): 589–596. Evaluators preferred the chatbot’s responses in 78.6 per cent of 585 evaluations; questions came from Reddit’s r/AskDocs. For the receipts in 2.4: 195 exchanges from October 2022, on a forum where moderators verify the credentials of the volunteers who reply; the chatbot (GPT-3.5) was queried on 22 and 23 December 2022; three blinded evaluators per exchange, who “were also coauthors”; median length 52 words for physicians against 211 for the chatbot; the authors warn that “the additional length of the chatbot responses could have been erroneously associated with greater empathy” and that “evaluators did not assess the chatbot responses for accuracy or fabricated information”; patients’ own judgements of empathy were not assessed.
Bernard Matolino and Wenceslaus Kwindingwi, “The End of Ubuntu,” South African Journal of Philosophy 32, no. 2 (2013): 197–205.
The Great Learning (Daxue), trans. James Legge, The Chinese Classics, vol. 1 (1861).
Analects 2.4, trans. James Legge.
Analects 2.12, trans. James Legge.
Analects 15.21 (15.20 in Legge’s numbering), trans. James Legge.
Analects 15.29 (15.28 in Legge’s numbering), trans. James Legge.
Epictetus, Enchiridion 1, trans. Thomas Wentworth Higginson (1865).
Marcus Aurelius, Meditations 5.20 and 8.48, trans. Gregory Hays (New York: Modern Library, 2002).
Pierre Hadot, The Inner Citadel: The Meditations of Marcus Aurelius, trans. Michael Chase (Cambridge, MA: Harvard University Press, 1998).
Hubert L. Dreyfus, “Alchemy and Artificial Intelligence,” RAND Corporation paper P-3244, December 1965; What Computers Can’t Do (New York: Harper & Row, 1972).
Edmund Husserl, Logical Investigations, vol. 2, introduction, §2, trans. J. N. Findlay, ed. Dermot Moran (London: Routledge, 2001), vol. 1, p. 168: “we must go back to the ‘things themselves’.”
Martin Heidegger, Being and Time (1927), §4, H. 12, trans. John Macquarrie and Edward Robinson (Oxford: Blackwell, 1962). On being-toward-death, §§46–53.
Maurice Merleau-Ponty, Phenomenology of Perception (1945), trans. Colin Smith (London: Routledge Classics, 2002), p. 169.
Thomas Nagel, “What Is It Like to Be a Bat?”, The Philosophical Review 83, no. 4 (1974): 435–450.
Dreyfus lost to Richard Greenblatt’s program Mac Hack in 1967.
Qur’an 3:191, trans. M. A. S. Abdel Haleem (Oxford: Oxford University Press, 2004).
Abū Ḥāmid al-Ghazālī, Iḥyāʾ ʿulūm al-dīn, Book 38 (Kitāb al-murāqaba wa-l-muḥāsaba, on vigilance and self-examination) and Book 39 (Kitāb al-tafakkur, on reflection).
Ibn Sīnā, Kitāb al-Nafs (the psychology of al-Shifāʾ), I.1 and V.7.
Saying of ʿUmar ibn al-Khaṭṭāb, cited without a chain of transmission by al-Tirmidhī, Jāmiʿ 2459; with a chain in Ibn al-Mubārak and Aḥmad ibn Ḥanbal, each in al-Zuhd.
Qur’an 17:14, trans. Abdel Haleem (2004).
Michael Polanyi, The Tacit Dimension (Garden City, NY: Doubleday, 1966), p. 4; the cyclist’s balance is his example in Personal Knowledge (London: Routledge, 1958).
Floriberto Díaz, Escrito: comunalidad, energía viva del pensamiento mixe, ed. Sofía Robles Hernández and Rafael Cardoso Jiménez (México: UNAM, 2007); Jaime Martínez Luna, Eso que llaman comunalidad (Oaxaca: CONACULTA, 2009).
Named people in this chapter are composites, and their conversations and circumstances are constructed; the studies, laws and figures cited are real and sourced.
Revised from the print edition (2026)
What the print edition claimed: that the traditions “arrived at the same destination”; that “the ancient traditions never made this mistake” of forgetting consciousness, while modernity did; that the earliest Upanishads date “roughly twenty-eight centuries” back, attached to the Kaṭha; that the Self is the chariot’s “passenger”; that for Śaṅkara avidyā is “the root of suffering”; that “Buddhism makes the opposite claim: there is no Self”; that an AI system is “the ultimate sleepwalker” with “no mindfulness at all”; that the Confucian term is xiuyang; that “Epictetus taught it as a slave”; that Nagel “was working within phenomenology”; that there is “nothing it is like to be an AI system”; that tafakkur belongs to “the Islamic philosophical tradition”; that Indigenous peoples worldwide “share a premise”; that Stoicism and phenomenology “emerged from shared intellectual soil.”
What I now hold: the traditions disagree about the self. They share a family resemblance, not a doctrine: each has its own way of holding a person to an act, and two (the Buddha’s heir of actions, the Qur’an’s record) say it almost in this book’s words. Philosophy did not forget consciousness; institutions paid for calculation. The Kaṭha is later than the oldest Upaniṣads; its self is the rathin, the lord of the chariot. Śaṅkara’s term for the mistake is adhyāsa, which the learned call avidyā. The Buddha declined to say whether there is a self, taught not-self, and taught that we are owners and heirs of our actions. Nobody knows whether a machine “sleeps”; we know how we fall asleep beside one. The classical term is xiūshēn. Epictetus taught as a freedman. Nagel is an analytic philosopher. Whether anything matters to a machine is open. Tafakkur is devotional and Sufi; Islamic philosophy has its own account of self-awareness in Ibn Sīnā. Indigenous philosophies differ; this chapter speaks of one. The Buddha’s heir is borrowed for one life only. What a machine with an inner life might be owed is left open.
What changed my mind: Olivelle on dating; Śaṅkara’s own preamble; the Ānanda Sutta and the five daily reflections; the Milindapañha; Legge and the Great Learning; the record of Dreyfus’s predictions; the 2023 studies of automation bias and of chatbot empathy; the work of Oaxacan thinkers on comunalidad.
What survives: the eight traditions; the Kaṭha chariot; “awareness is a practice, not a possession”; “stillness can be its own form of sleep”; ubuntu as relational, awareness that “happens between people, not only inside them”; “the accomplished scholar is not a utensil”; the Stoic line between what is and is not in our power; the poet in Konya; the grandmother in Oaxaca; the monk in Yangon.
Next in the series: Chapter 3: The Qualities of the Conscious Self.
Comments