VOLUME V, CHAPTER 14, SECTION 3
Artificial Intelligence Is a Tool, the Unlicensed Trainer, and the People It Replaces
The previous section discussed the great leap and the law's lag, and how the law, once it catches up, is made for the vested interest. This section returns to the three most central questions, what artificial intelligence actually is, where the ones who train it actually stand, and where the ones it replaces actually stand, and then closes out this whole volume and crosses the bridge into the volume on saving people. This is the final section of this volume.
One. Artificial Intelligence Is a Tool, Not a Subject
The fiercest argument in contemporary academic circles is whether artificial intelligence counts as a legal subject. One camp wants to build it an electronic personhood, or a limited personhood, on the grounds that it grows more and more autonomous. But the root of this argument sits not in philosophy, it sits in interest. If artificial intelligence is recognized as a subject, then when it errs, it is held responsible by itself, and yet it earns no money, pays no damages, and serves no prison sentence, so it is held responsible by itself amounts to no one being held responsible at all, which suits the developing company perfectly, collecting business with one hand while, when something goes wrong, pushing the responsibility onto the artificial intelligence with the other. What is called subject theory is, at bottom, an attempt to build a legal firewall, separating the company from the harm it causes.
What artificial intelligence actually is, put plainly, is a stretch of code running on a server, a model fed by an ocean of data, with no will and no self. The poems it writes come from the poems of the people who fed it. The legal analysis it produces comes from the legal texts of the people who fed it. It does not create, it synthesizes, a trained tool. The ancestors had already drawn this exact line thousands of years ago.
What is above form is called the Way. What is below form is called the vessel. The Book of Changes, Xici, Part One.
A person carries the Way. A vessel serves the Way. However much artificial intelligence resembles a person, its position is the vessel, not the Way. Confucius said the gentleman is not a vessel. If even a person should not be treated as a mere tool, then a tool certainly should not be raised up to the standing of a person.
Two. A Tool Needs No License. Whoever Uses the Tool Always Does
Fixing artificial intelligence in the position of a tool carries one direct consequence, the tool needs no license, and the person using the tool does. A scalpel needs no license, the surgeon who wields it does. A law book needs no license, the lawyer who reads it does. Carried over to artificial intelligence, this means, when a doctor uses it to help diagnose, the responsibility for the prescription finally written rests with the doctor. When an accountant uses it to prepare a statement, the responsibility for the final signature rests with the accountant. What the tool produces always has to pass through the licensed person's own mind and hand before it can truly be put to use in the real world. Responsibility always lands on the person, and artificial intelligence is responsible for itself is an empty phrase that pushes responsibility off into thin air.
And the rule that whoever uses the tool must be licensed is something built up over thousands of years. In the fifth century BC, Hippocrates first separated the physician from the wandering healer, establishing it as a trade with its own ethics and its own standards. The Tang dynasty's Sun Simiao, in On the Absolute Sincerity of Great Physicians, put it even more deeply.
When someone in distress comes seeking help, do not ask whether he is noble or humble, rich or poor, but treat all equally alike. Sun Simiao, On the Absolute Sincerity of Great Physicians.
When a person in suffering comes seeking help, treat him, regardless of rank or wealth, as if he were your own closest kin. China also had the governance of physicians recorded in the Rites of Zhou, and the Imperial Medical Bureau of the Sui and Tang. East and West both recognized one thing, any trade that touches another person's life, freedom, property, or future, and cannot be undone once it goes wrong, must carry a qualification, a license, and a chain of responsibility that can be traced. A license is not a piece of paper, it is a whole set of accepted restraint.
Three. But This Time, a Displacement Unseen in Thousands of Years Has Surfaced
The engineers training a large model may number in the thousands, or tens of thousands. Those training an artificial intelligence that writes code need no license, because software itself has never required one. But who is training the artificial intelligence built for medicine, for law, for psychology. Following the general pattern across the industry, hiring for training work prizes engineering skill, a foundation in modeling, and data capability, and a professional license is not a required condition. In other words, the person training a medical artificial intelligence does not necessarily hold a physician's license. The person training a legal artificial intelligence does not necessarily hold a law license. The person training a psychological artificial intelligence does not necessarily hold a license in psychology.
A medical consultation tool reaching hundreds of millions of people is, at bottom, a medical instrument trained collectively by hundreds or thousands of unlicensed people. In the traditional system, an unlicensed person offering medical advice to a patient is called practicing medicine without a license, grounds for revoking a license and pursuing criminal responsibility. But here, unlicensed people, through the act of training a model, indirectly hand out medical judgments to hundreds of millions of people, and the entire process is, under current law, completely legal. This is not a loophole being exploited, it is that the law has simply never reached this ground at all. Administratively, the artificial intelligence trainer has already been recognized as a new occupation, with technical standards set for it, but those standards sit entirely at the engineering level, never once touching the question of whether the person training an artificial intelligence in a given field must hold a license in that field. The new occupation has been recognized. The matching system of licensing has not caught up. This is the single most precise sample there is of the great leap and the law's lag.
Four. When Something Goes Wrong, No One Can Be Held to Account
Right behind this sits a question that matters more than any other, when artificial intelligence causes harm inside a licensed profession, who is responsible. Under the present system, an artificial intelligence company stands, almost undefeated. The first layer, a sea of people, a traditional chain of responsibility has clear positions, the maker, the seller, the user, while artificial intelligence's chain runs through data sourcing, labeling, algorithm engineering, training, product, testing, platform, and user, hundreds of thousands of people each carrying a sliver of it, diluted until no one can even be pointed to as the one who decided how it would answer. Trace it back to the data's source, and the error may sit inside an old book from thirty years ago, or a reply on a forum long since shut down, impossible to trace at all.
The second layer, the backing of capital. Artificial intelligence companies are often valued in the hundreds of billions or trillions, able to hire the very finest lawyers, and an ordinary victim suing one is, in effect, one person challenging a national-level legal team, and even a win in court brings damages that amount to barely a drop in the bucket set against the company's revenue. The third layer, the user agreement, this product is for reference only, does not constitute professional advice, and you bear the consequences yourself, and the moment you click I agree, you have built the company's firewall for it. The chain of responsibility breaks apart exactly this way, and it is exactly this break the vested interest needs, so that it can keep enjoying the business while paying not one cent of the cost.
Five. Turn the Lens to the Side of the Ones Replaced
A great leap has two sides. The taking side has now been laid out in full. There remains the side of those replaced. What is being taken this time is the professional territory of mental laborers, lawyers, doctors, teachers, accountants. Across history, those replaced have stood in four different positions. The position of resistance, standing in the machine's way, and just as the Luddite movement could not block the steam engine, a joint letter from a professional association today cannot block this either. The position of exit, changing trade or accepting one's fate, though for a professional who has practiced half a lifetime, the cost of changing trade runs frighteningly high. The position of self-strengthening, learning to use the machine, to manage the machine, turning artificial intelligence into a tool in one's own hand, though the ones who manage to make that leap are always only a small, fast-moving few. The position of demanding support, if you have used a machine to take my job, then you owe me a way to live, and two centuries of labor movements winning unemployment insurance and welfare is exactly the fruit of standing in this position.
Every one of these four positions is narrower this time than ever before. Those replaced in earlier leaps were manual laborers, and exit still meant they could carry goods or farm land. This time, those replaced are mental laborers who spent a decade or more earning a license, and artificial intelligence is stepping straight into the work at the very peak of the pyramid, leaving nowhere easy to retreat to and nowhere easy to leap toward. There is also a strange position here, mental laborers themselves hold the microphone, so the debate this time is louder than at any point before, and yet a loud voice does not mean a good position, the loudest voices stand on ground that is, inch by inch, giving way beneath them. What position always looks at is the ground underfoot, never the volume of the voice.
Six. The Deepest and Most Disorienting Position of All, in This Round
In earlier great leaps, the machine took old work with one hand and created new work with the other, and those displaced could, within a few years, generally find their footing in some new post. This time, the speed at which artificial intelligence replaces work may outrun the speed at which it creates work. Those replaced now stand in a position no one has ever stood in before, across thousands of years, things keep growing more plentiful, while they themselves grow more and more useless. Artificial intelligence pushes output to an extreme, the grain in the fields, the goods in the factories, the services across every trade, all growing more abundant, and yet producing all of it needs them less and less.
A society where things grow so plentiful they overflow, while at the very same time a large number of people, unable to find a way to contribute, find themselves with no way to live, is a position humanity has never squarely faced before. Why, with so much abundance, can none of it reach the hands of those with no work left to do, this question already touches the root of how output gets divided and to whom unearned income belongs, and that is a question of the way out, one this volume does not answer, leaving it for the volume that follows.
Seven. Speaking Only of the Civilian Sphere Leaves One More Paradox, Laid Out Here Without Being Resolved
This entire run has spoken only of the civilian sphere, setting military, aerospace, and defense applications aside, because that sphere is directly governed by the state, with clear approval, procedure, and accountability, and the problem of the unlicensed trainer does not exist there. The civilian sphere is exactly where this positional disorder actually occurs, because the end user there is an ordinary citizen with no professional training, who takes almost everything artificial intelligence produces at face value.
But the claim that whoever trains a medical artificial intelligence should hold a physician's license runs immediately into a paradox, the engineers who design a scalpel or build an imaging scanner hold no physician's license either, so why should the one training a medical artificial intelligence be any different. Looked at closely, the difference is this, a medical device is a physical tool, cutting, imaging, and whether and how it is used is decided entirely by the licensed physician standing in the middle, and it passes through strict registration, approval, and a complete chain of traceability. A medical artificial intelligence is a judgment-making tool, its advice moving directly into the user's own decision, with no licensed physician standing in between to intercept it, and none of a device's approval or traceability behind it. It wants to enjoy a device's convenience of an unlicensed trainer, while refusing to accept a device's three layers of strict protection, and this is the deepest disorder of all. Yet even if those three protections were added, whether the license should ultimately rest with the trainer or with the user, for a judgment-making tool, remains an open question. This paradox, this volume only lays out. It does not resolve it.
Blinded by one corner, and left in the dark about the greater principle. Xunzi, Xunzi, Dispelling Blindness.
Blinded by a single part, a person loses sight of the larger truth. Today's discussion of artificial intelligence law is, for the most part, blinded by partial issues like algorithmic bias and data privacy, seeing the leaves turning yellow without seeing that the root itself has rotted, patching the old law again and again without ever touching the true root, the unlicensed trainer practicing without a license. Laying out a paradox without rushing to dissolve it is this work's deepest posture of all, because its true way out does not sit inside this volume.
Eight. The Position Set by This Section, and Volume V Closes Here
This section sets the forty-third layer. Artificial intelligence is a tool, and a tool needs no license, while whoever uses it always does. Those who train it, in the civilian licensed professions, are, for the most part, unlicensed, a displacement unseen in thousands of years. When something goes wrong, a sea of people, the backing of capital, and the user agreement stack together, and responsibility cannot be traced. Those it replaces stand in positions growing narrower and narrower, even reaching the point where the more plentiful things grow, the more useless they themselves become. All of this is position. None of it is a verdict.
Here, the whole of Volume V, Legal Institutions, closes. This volume's forty-three layers ran from the root of law, natural law as the highest position with law as wei beneath it, whom law serves decided by position, no one is without fault, rights rooted in nature, the highest law being self-discipline, through how law runs, strength and weakness, public legal education, lawyers, the lawmaker, vested interests, through how law treats an ordinary person, the direction of proof, guilt as a position, law dead and a person alive, we the people, through law and scarcity, land, the imperial examination, the eight tools of contemporary America and China, through the space between nations, the taker, the taken, occupying a position, the treaty, and finally to the age of artificial intelligence, the great leap and the law's own lag. One single thread runs through it all from beginning to end, the direction law takes is, in the end, decided by output and by position.
This volume, from beginning to end, has done exactly one thing, laying out positions, and offering no way out. Every position has been set firmly in place. Every disorder has been laid out clearly. Every predicament has been brought before your eyes. And not one of them has been answered. The answer sits in the next volume. The next volume, Saving People: Racing Against Time Before the Blood Flows, will answer, directly, every hard question this volume has set down, how to pull those replaced by artificial intelligence out first and see them settled, how to grow output large enough that no one need take anymore, how to let what you would not want done to yourself grow from one person's own cultivation into an entire way of life shared across all of humanity. The more thoroughly the positions have been laid out, the more firmly that way out will stand. Carry these forty-three layers of position, and walk into the next volume.
Credit to the ancestors; the mistakes are mine: Tiger Lyon.