The Challenge of AI is Social

A technological revolution is never only a technical event. It is an audit of a civilization. It asks whether schools still educate, whether institutions still work, whether citizens still trust one another enough to act together.

In 1871 the Meiji government sent officials and scholars abroad. The party visited the United States and Europe. Its task was practical: inspect institutions, factories, schools, ports, armies, and courts; record what could be copied; report what could be adapted; return with usable knowledge. Japan’s leaders had reached a hard conclusion. Industrial powers had acquired new productive capacity, new weapons, new transport, and new administrative competence. A preindustrial state could not protect its independence against countries that had mastered these tools.

The results were rapid. Japan reorganized education, industry, finance, transport, and state administration. By 1905 it defeated Russia. Between 1913 and 1938 its industrial production increased more than fivefold. Britain, once the early industrial leader, rose far more slowly in the same period. The lesson was not that machines alone make history. The lesson was that societies either reorganize around a new technology or become evidence in someone else’s archive.

Of course, that is a simplification, it is too tidy, and history is seldom tidy except in retrospect! Yet the evidence is stubborn. A technological revolution is never only a technical event. It is an audit of a civilization. It asks whether schools still educate, whether institutions still work, whether citizens still trust one another enough to act together, whether ambition has become vanity, and whether public language can still carry truth without collapsing under the weight of slogans, advertisements, and professional evasions.

Artificial intelligence has arrived at precisely the hour when many advanced societies have become expert in evasion. We have committees that know how to postpone, universities that know how to brand, bureaucracies that know how to protect their own procedures from their declared purposes, and public debates in which everyone is permanently informed and rarely wiser. Into this scene comes a technology capable of writing our memos, tutoring our children, designing our drugs, choosing our news, automating our offices, ranking our risks, advising our governments, flirting with our loneliness, and producing, with admirable fluency, the sort of prose that sounds intelligent until one notices that nobody has taken responsibility for it.

Here is the absurd truth, and it is funny only because it is so expensive: the country that buys the most chips may still lose to the country that teaches its citizens what to do with them.

The fashionable race is for data centers, frontier models, semiconductors, energy contracts, training runs, and the handsome gray buildings in which executives speak about ‘compute’ with the solemnity once reserved for grain reserves. I do not mock the hardware. Without it, there is no serious participation in this age. But hardware is not destiny. The decisive claim here is harder and more adult: ‘the competitive challenge of AI is primarily social, not technological.’

That is the argument worth taking seriously. It cuts against the current mood, because the current mood loves purchase orders. Purchase orders are comforting. They imply that renewal can be procured. A nation orders processors, cables, servers, and cooling systems; a minister gives a speech; a ribbon is cut; the future is photographed. Unfortunately, history has a vulgar habit of asking follow-up questions. Are the schools ready? Are the civil servants competent? Are the courts trusted? Are local institutions alive? We need to know if young people are forming judgment or outsourcing it, if workers are gaining power or merely being processed faster, and if citizens are using AI to extend their agency or being trained, politely and at scale, to become spectators of their own lives.

The central danger is not that machines become clever. The central danger is that human beings become passive. The device does not need to dominate us in any lurid theatrical way. It need only make judgment optional, effort unfashionable, attention rentable, and memory unnecessary. Then the damage will arrive with excellent user experience.

I think of this not as a problem of robots but as a problem of national character. The old components of competitive strength remain: ambition, coherent identity, shared opportunity, an active but not suffocating state, effective institutions, a learning society, and pluralism strong enough to survive disagreement. AI does not abolish these requirements. It tests them. It magnifies their virtues and their defects. A confident society can use AI to widen opportunity, accelerate discovery, repair public administration, and give ordinary people stronger tools. A brittle society can use the same systems to deepen inequality, automate arbitrary decisions, pollute public knowledge, and turn loneliness into a subscription product.

The moral issue and the strategic issue converge. Human agency is not a decorative liberal concern, suitable for speeches. It is a source of power. A society of capable, self-commanding, educated, argumentative, inventive citizens will adapt faster than a society of managed consumers. The first will use AI as an instrument. The second will be used by AI systems designed elsewhere, owned elsewhere, and optimized for purposes nobody voted on.

There is a grim comedy in the phrase ‘artificial intelligence strategy.’ It suggests that the problem is artificial intelligence. It is not. The problem is the condition of the country receiving it. AI enters the existing household. If the roof leaks, the accounts are chaotic, the children are neglected, the locks are broken, and everyone is shouting in separate rooms, the arrival of a brilliant new assistant will not produce order. It may simply help each person shout in better grammar.

The point is not anti-technological. Quite the reverse. The most serious pro-AI position is not boosterism. It is stewardship without sentimentality. We must build public-sector competence, train talent, use AI to widen opportunity, and protect autonomous agency. We need to fund discovery, repair the information environment, and streamline public bureaucracy while building foresight functions able to think ahead before events become emergencies with stationery. These are not ornamental reforms. They are the minimum duties of a state that wishes to remain serious.

The bureaucracy question deserves special attention, because it is one of the great tragedies of modern government. We created institutions to serve human purposes; then we trained generations of citizens to serve the institutions. Forms begat forms. Procedures acquired bodyguards. The citizen, arriving with a need, became a suspicious object moving through a system designed by people who had never met him and were faintly irritated by his existence. AI could reduce this cruelty. It could make public services faster, clearer, less humiliating, and more humane. Or it could do the opposite: produce perfect denials, instant opacity, and algorithmic authority with no face, no memory, and no shame.

That is why reform cannot be left to vendors. A vendor will sell efficiency. A republic needs legitimacy. Efficiency can process a person. Legitimacy must answer him. The distinction is important. It is the difference between a state that earns obedience and one that merely extracts compliance.

The same is true of education. AI can become the greatest tutor ever placed in the hands of a poor child, or the most sophisticated excuse ever given to a lazy one. It can help a student learn languages, mathematics, history, coding, medicine, music, and debate. It can also complete the assignment, write the reflection, summarize the book, invent the citation, and leave behind a pupil who has received the outward signs of education while remaining inwardly unfed. We will then have achieved a rare administrative triumph: universal access to advanced learning without the inconvenience of learning.

What do we want citizens to become? If the answer is employable, governable, and entertained, AI will serve that thin ambition with terrifying competence. If the answer is capable of judgment, then schools must change not by banning AI or worshipping it, but by making thought visible again through oral defense, handwritten notes, live problem solving, sustained reading, apprenticeship, criticism, revision, and the patient formation of taste. AI should raise the standard, not lower the human.

The public information sphere is even more fragile. A society can survive disagreement. It cannot survive the permanent collapse of shared reality. AI can help detect fraud, compare claims, translate technical knowledge, and give citizens better access to evidence. It can also produce falsehood at industrial scale and with the emotional timing of a professional manipulator. The old lie had to find a printer, a broadcaster, or a demagogue. The new lie needs a prompt, a network, and a target list.

This is where the issue becomes political in the deepest sense. Freedom requires persons capable of judgment. Judgment requires some common ground of fact. Common ground requires institutions that can be criticized without being destroyed, trusted without being worshipped, and corrected without humiliation. AI will not create that culture for us. It will reward the culture we have.

I do not believe national renewal will come from a single central plan. The lesson is almost the opposite. The state must be active, but not possessive. It must fund, coordinate, convene, and protect; it must build competence and insist on public purposes. But the energy must also come from firms, schools, universities, cities, laboratories, libraries, civic associations, families, and individuals. The most successful societies will be those in which thousands of local experiments proceed at once, many fail, some succeed, and the successes travel. The future will not be licensed only in capital cities.

This is also why the contest between nations will not be settled by model rankings alone. Model rankings are useful, but they can seduce the mind into a childish view of power. A country can have impressive systems and still fail to integrate them. It can have clever engineers and foolish institutions. It can have dazzling firms and miserable schools. It can have the best chatbot in the world and citizens who no longer know how to ask a serious question.

We don’t need an AI strategy. We need a strategy for national renewal and competitive advantage that uses the capabilities and opportunities of the AI Era to achieve its goals. Countries must stop treating AI as a sector and start treating it as a test of social capacity.

The historical comparison with Japan is not comforting, because Japan’s success required urgency, humility, discipline, and a willingness to learn from abroad without ceasing to be itself. That combination is rare. Nations prefer flattery. They enjoy hearing that their values, their entrepreneurs, their traditions, their universities, their soldiers, their regulators, or their founders have already guaranteed success. Britain once had every reason to assume primacy. Russia once had size. The Ottoman and Austro-Hungarian empires had ceremony, memory, and inertia. None of these proved sufficient.

The American danger is not lack of talent. It is the substitution of talent for seriousness. There is astonishing brilliance in the labs, the firms, the universities, and the young. Yet brilliance dispersed through failing institutions can become a kind of national theater: impressive scenes, poor plot. The country can produce miracles in one room and paralysis in the next. It can train a model on a continent of data and fail to issue a permit, treat a veteran, teach algebra, or explain a tax form in English.

We may build machines that can reason across the frontier of science while preserving administrative systems that cannot answer the telephone.

Still, I am not pessimistic. Foreboding is not despair. The best reading of this age is that AI gives societies a rare chance to correct defects they had already learned to tolerate. It can help rebuild public administration around citizens rather than files. It can return expert help to places without experts. It can give small firms analytical capacities once reserved for giants. It can assist doctors, teachers, engineers, researchers, and judges without replacing the moral burden of their judgment. It can help a state see earlier, act faster, and explain better. It can make opportunity less dependent on birth, geography, and inherited access.

But only if we choose that path deliberately. The default setting of powerful technology is not human dignity. The default setting is extraction by whoever controls the system. Dignity has to be designed, defended, audited, taught, and renewed. Agency has to be practiced. A citizen who never has to remember, judge, search, revise, or decide is not liberated. He is being retired early from the human enterprise.

So the task before us is plain enough to be frightening. Build the machines. Build the data centers. Secure the chips. But then do the harder work: rebuild the society that will use them. Teach the young to think with AI without surrendering thought to it. Train officials who understand both code and consequence. Make institutions faster without making them colder. Use automation to return time to human beings, not merely profit to systems. Protect the public sphere from synthetic poison. Make opportunity broader, not more hereditary. Give citizens tools that enlarge their powers rather than comforts that shrink them.

In 1871 Japan sent people abroad with notebooks. They went to inspect the world that had overtaken them. The notebooks were less important than the disposition behind them: seriousness without self-pity, ambition without fantasy, learning without servility. Our instruments are different now. The old notebooks have become laptops, datasets, models, and terminals glowing late in offices, classrooms, libraries, kitchens, and laboratories. Somewhere, late at night, a student asks a machine a question, rejects the easy answer, asks a better one, and begins again.

Stay curious

Colin

Image: Europeana on Unsplash