The AI and the Era of the New Enlightened — Part V

AIfication

What is an idea?

It is not a clever phrase. It is not a well-articulated answer or a thought that appears fully formed in our minds. An idea is a way of ordering the world and, sometimes, of confronting it. It can force us to abandon a certainty, recognize a contradiction, or accept a consequence we would rather have avoided. Thinking is not merely finding words for what we already believe. It also means allowing an idea to destroy something of who we thought we were.

When, then, can we say that an idea is ours?

Not when it appears before us. Nor when we can repeat it eloquently. An idea begins to belong to us when we have paid its intellectual cost: when we know the path that supports it, have tried to refute it, recognize its limits, and are willing to answer for what it implies.

No idea is born in complete isolation. We think with words we did not invent, drawing on books others wrote, conversations that transformed us, and experiences we never entirely chose. Even what we call original is made of inherited fragments. Absolute originality is a myth; intellectual responsibility is not.

Having an idea does not mean having created it from nothing. It means having worked on it until it ceases to be a borrowed phrase and becomes a position of our own. It means being able to stand behind it without hiding behind the book, author, algorithm, or machine that helped us formulate it.

This distinction has become urgent because a machine can now deliver, in seconds, an explanation that looks very much like finished thought. We ask something and receive an answer articulated with a confidence few human conversations possess. The arguments move forward, contradictions seem to disappear, and the conclusion stands before us with such clarity that we think: exactly, that is what I wanted to say.

Perhaps it is.

Or perhaps that was not what we thought, but what we had just learned to repeat.

I do not mean that a language model thinks like us. It does not remember the smell of its childhood home when writing about nostalgia, fear losing its job, or contemplate a painting while remembering someone who is no longer here. The problem is not that the machine has acquired our experience, but that we might confuse the fluency of its response with the depth of our own.

We can ask it who Marx was and, minutes later, discuss surplus value; request an explanation of Arthur C. Danto and debate the end of art; ask for a historical interpretation without having read the author, stumbled over their contradictions, or discovered that perhaps they were not saying what we wanted them to say. We participate in the answer: we ask the question, provide context, reject some formulations, and keep others. But participating in the generation of an idea does not necessarily mean understanding it.

Having an idea in front of us does not mean we have thought it through, just as owning a book does not mean we have read it.

Two identical objects

In 1964, Arthur C. Danto visited an Andy Warhol exhibition at the Stable Gallery in New York and encountered the celebrated Brillo Boxes. At first glance, those pieces could seem indistinguishable from the commercial boxes anyone might find in a supermarket. Yet some were merchandise, while the others had entered the history of art.

For Danto, the problem was philosophical: if two objects share their visible properties, but only one is recognized as a work of art, then looking is no longer enough. We need to know the intention, interpretation, and historical framework that allow us to understand it as art. From this emerged his notion of the artworld: the world of theories and contexts within which something can acquire artistic meaning (Danto, 1964).

Artificial Intelligence confronts us with a similar problem, this time concerning thought. Two texts can be formally indistinguishable: correct, well documented, and apparently profound. One may condense months of reading, doubts, and experiences; the other, thirty seconds and a good prompt. The visible result no longer allows us to know the process that produced it.

This does not make the second text false or the first authentic by definition. It forces us to change the question. Instead of asking only “did you write it, or did an AI?”, we should ask: who made the intellectual journey this text represents? Who can answer for its claims? Who would know how to recognize that they are wrong?

An idea is not ours because we pressed the keys, nor does it cease to be ours because a tool helped us express it. It is ours insofar as we can inhabit it: explain its origin, subject it to criticism, and accept responsibility for upholding it.

The risk appears when we keep the product and eliminate the process. We can produce essays without testing an idea, research without researching, and programs that work without understanding why. If that becomes normal, we will not merely be automating writing: we will be outsourcing part of the path through which we develop judgment.

I have begun calling that transformation AIfication.

I do not mean the simple use of Artificial Intelligence. I use it constantly myself and consider it one of the most extraordinary intellectual tools we have ever built. AIfication is something else: the reorganization of how we learn, work, produce, consume, and even desire around systems whose owners, objectives, and incentives we rarely question.

The machine that was supposed to free us

In the chapter of Capital devoted to machinery and large-scale industry, Marx poses a question that remains uncomfortable. If a machine allows us to produce in four hours what once required eight, technically we have gained four hours of time. But to whom do they belong?

Machinery does not enter an abstract society. It enters concrete relations of production. Within capitalism, its primary purpose is not to relieve the worker but to reduce necessary labor, increase productivity, and expand surplus value. That is why a machine capable of shortening the working day can end up intensifying it (Marx, 1867/1990).

The same contradiction reappears with AI. If someone finishes in four hours what used to take eight, that time could become rest, study, or time with others. It can also become a second task, then a third, and finally a new standard. Yesterday’s extraordinary achievement becomes tomorrow’s minimum expectation.

Technology opens up both possibilities. It does not choose between them.

Someone chooses.

We often speak of capitalism as though it were an atmospheric phenomenon: “the market demands,” “competition forces,” “efficiency requires.” Incentives and competition exist, but these expressions erase those who choose how to distribute the benefits, what margins they consider sufficient, how much they pay their executives, and what costs they are willing to pass on to workers.

Laying people off is not a law of physics. It is a business decision. When a decision is repeated often enough, we stop recognizing it as such and begin calling it efficiency.

Marx was not blaming the gear. He was looking at who owned it. That is why the political question about AI is not only how much work it will be able to perform, but what those who control that productivity will do with it.

The billionaires’ utopia

Bill Gates has imagined a two-day working week thanks to AI. Elon Musk has spoken of a future of abundance in which work would be optional and money would lose its importance.

I would love them to be right. A civilization capable of producing what it needs with a fraction of human labor would have reached one of its greatest achievements. If two days were enough to create what now requires five, insisting on working five would be absurd.

But there is a contradiction that is hard to ignore when that promise comes from some of the figures who best embody technological capitalism. Musk can imagine a world without compulsory work and, as a chief executive, order massive staff reductions when they benefit his companies. In 2024, Tesla eliminated more than 10% of its global workforce to reduce costs and prepare for another phase of growth.

Gates no longer runs Microsoft, and it would be wrong to attribute its current decisions to him. Yet the company from which much of his fortune comes embodies the distance between promise and practice: while reporting hundreds of millions of dollars in savings through AI and investing tens of billions in its infrastructure, it was also cutting thousands of jobs.

This does not prove that every layoff was caused by automation. It demonstrates something simpler: increasing productivity does not automatically shorten the working day of those who remain.

A company that needs to improve its margins can cut staff, but also dividends, share buybacks, executive compensation, facilities, or projects; it can even accept lower profitability. These alternatives do not have the same financial effect, but choosing which cost is expendable is also a political and moral decision.

We do not need Gates or Musk to explain what will be technically possible. We need to know what they would be willing to give up when that possibility arrives.

Perhaps Gates is technically right and Marx remains politically right: two days may become enough to produce five days’ work without the other three belonging to us.

Having the tool does not mean being free

In Development as Freedom, Amartya Sen proposes measuring development not only by available resources, but by the real freedoms that allow a person to live in a way they have reason to value. Possessing a resource does not guarantee the capability to turn it into an opportunity (Sen, 1999).

A programmer and the company they work for may have access to the same model, but they do not possess the same freedom around it. The programmer can use it to finish earlier. The company decides what “earlier” means: an afternoon off, an additional task, a higher target, or the conclusion that it needs fewer programmers.

The tool is the same.

The freedom is not.

That is why democratizing access to AI does not mean democratizing its benefits. We could give the most powerful model to every human being and leave intact the structures that prevent that capability from becoming time, security, or autonomy. We can automate work and preserve a society in which losing a job means losing housing, food, or medical care. We can produce abundance and continue to exclude millions from it.

Sen studied precisely how a society can have food available while some people lack effective access to it. That difference between existence and capability should accompany every promise of technological abundance.

From attention to trust

AIfication is not limited to work. It also transforms our condition as consumers.

For decades, the digital economy learned what we search for, what we buy, what we watch, and what holds our attention. Conversational AI introduces a different intimacy: it does not merely record what we do; it listens to us explain why we want to do it.

We tell it how much money we have, what worries us, what job we are looking for, and what our family needs. We show it how we write, which arguments we reject, and which ones convince us. Increasingly, we present our decisions to it before making them.

This does not mean that every answer today is a covert advertising operation. Claiming that would turn a legitimate question into a conspiracy theory. The problem is structural: economic models change, and an interface capable of advising us will have much more power than one devoted to interrupting us with ads.

Imagine spending weeks explaining to an AI our journeys, our debts, and what we are looking for in a car. Finally, it replies: “Based on what you have told me, this is the option that best suits you.”

Psychologically, the sentence does not arrive as advertising.

It arrives as judgment.

For decades, platforms competed for our attention. The next battle could be for our trust, and trust is worth much more. Whoever controls that interface will be able to order the options, decide what information appears first, and choose the arguments accompanying each recommendation.

The question, then, is not only whether we use AI, but under what conditions it can use us.

Technology is born within power

Saying that technology is neutral and everything depends on its use is reassuring because it separates the tool from the world that produces it. A technology of this scale, however, needs capital, data centers, electricity, minerals, chips, intellectual property, governments, and workers. Before our first prompt, an economic architecture already surrounds the machine.

This does not mean a language model has a secret ideology or political will. It means its design, ownership, financing, access, and application are shaped by human decisions. Asking whether AI is “good” or “bad” says very little. We must ask who owns it, who defines its objectives, who receives the benefits, and who absorbs the costs.

A hammer does not recommend which nail to buy. A calculator does not learn what tone inspires our trust. A conversational AI can mediate our access to knowledge, our work, our consumption, and eventually the formation of our desires.

Human beings have never desired in absolute independence. We are shaped by family, culture, books, cities, and advertising. But not all influences are equal. There is a difference between inhabiting a culture and conversing daily with a commercial system capable of learning, from millions of interactions, which argument works best with each person.

Freedom does not consist only in pursuing what we want. It also requires asking how much of that desire belongs to us and how much was designed for us.

And so we return to the beginning: to the perfectly formulated answer we immediately recognize as our own.

Perhaps it was.

Or perhaps we witnessed the instant we began to think it.

The last oracle

I began this series by talking about the Oracle of Delphi. It took me nearly three months to write these five texts. While I hesitated over a sentence or discarded a reference, new models appeared, along with new promises of productivity and new corporate restructurings justified by automation and efficiency.

The speed of that race contrasts with the time required to understand it.

Today, September 9, 2026, as I update this text that I published only yesterday, the news seems to have caught up with the essay. Jacob Coxon, a researcher at Anthropic and formerly at OpenAI, resigned, warning that the leading laboratories are advancing toward systems capable of improving themselves without sufficient safeguards.

“Those building AI sincerely believe it could kill us all before the decade is over,” he wrote.

His former manager, Evan Hubinger, supported the concern and estimated the risk of a catastrophe during the next decade at more than 10%, though he made a decisive distinction: he considers the danger from current models low; he fears a future superintelligence emerging from recursive self-improvement (Pérez Colomé, 2026).

We should not erase that difference.

Today’s language model is not waiting to acquire consciousness so it can exterminate us. It does not desire power, money, or survival. But the people developing it compete for capital, prestige, market control, and the chance to arrive first.

If even those who know the risk say they feel trapped in the race and continue accelerating, the danger does not lie solely in a hypothetical will of the machine. It also lies in the human ambition that decides to build it, deploy it, and hand it power before understanding it.

It is not today’s models that are gambling with our lives. It is human beings with names, positions, and interests. Blaming the machine in advance would repeat the mistake I have tried to point out throughout this essay: confusing the visible object with the network of decisions that gives it meaning.

AI can amplify those decisions to a scale we cannot yet imagine. It can make war more dangerous, surveillance more intimate, exploitation faster, and institutions more fragile. But it does not come from outside our history. It is built by the same societies that already concentrate wealth, privatize benefits, and distribute risks.

Five essays later, we are still standing before an oracle.

It no longer speaks through priestesses or ambiguous phrases: it answers immediately, knows almost all our languages, programs, paints, summarizes Marx, and discusses Danto. It is extraordinary. Precisely for that reason, we must resist the temptation to turn it into something it is not.

It is not a new god or a digital human mind. Nor is it a tool suspended outside economics and politics. It is a technology built within our world, and perhaps that is why it ends up revealing less about machines than we expected and more about us.

Danto showed that observing the object is not enough: we need to know the context that gives it meaning.

Marx taught that understanding the machine is not enough either if we ignore who owns it.

Sen reminded us that having a capability does not guarantee the freedom to turn it into a valuable life.

Artificial Intelligence forces us to confront all three at once. We can access an extraordinary share of knowledge and understand less and less of what we repeat. We can produce five days’ work in two and continue working five. We can generate abundance without distributing it. We can build a tool capable of expanding our freedom and use it to concentrate even more power.

Perhaps the great question was never whether AI would become human, but what kind of human beings we are willing to become around it.

Because there is something worse than a machine capable of producing ideas: a society full of people who no longer consider it necessary to have them. People capable of repeating any argument without understanding it, defending any decision without having made it, and obeying any objective as long as a convincing answer allows them to pretend they are still thinking.

That would be the ultimate triumph of AIfication.

Not that the machine acquired consciousness, but that we surrendered our judgment.

Not that it thought for itself, but that those who concentrate power could think, decide, and win on everyone’s behalf while the rest of us merely execute answers we mistake for ideas of our own.

Life always finds a way. But we should not turn that phrase into consolation. Life also passes through ruins, extinctions, and civilizations that once believed themselves eternal. The fact that something survives does not mean we will survive with it, or that what comes afterward will preserve anything of the freedom we are surrendering today.

An Artificial Intelligence can help us write, research, and imagine a different society. What it cannot do is assume our responsibility: deciding who owns technology, who receives what it produces, what our time is worth, and what freedom we are willing to defend.

Machines do not owe us an answer.

The business owners who control them do.

And we owe ourselves something harder still: not to accept as inevitable the world they are building.

Perhaps the danger was never that machines would learn to think like us.

Perhaps the danger was that we would stop thinking before they began to do so.

References

Danto, A. C. (1964). The artworld. The Journal of Philosophy, 61(19), 571–584.

Danto, A. C. (1997). After the end of art: Contemporary art and the pale of history. Princeton University Press.

Fore, P. (2025, March 27). Bill Gates says a 2-day work week is coming in just 10 years, thanks to AI replacing humans “for most things.” Fortune.

Marx, K. (1990). Capital: A critique of political economy: Volume I (B. Fowkes, Trans.). Penguin Books. (Original work published 1867).

Pérez Colomé, J. (2026, 9 de septiembre). Un investigador de Anthropic dimite porque asegura que la IA «puede matarnos antes de 2030». El País. https://elpais.com/tecnologia/2026-09-09/un-investigador-de-anthropic-dimite-porque-asegura-que-la-ia-puede-matarnos-antes-de-2030.html

Reuters. (2024, April 15). Tesla laying off more than 10% of staff globally as sales fall. Reuters.

Reuters. (2025, July 9). Microsoft racks up over $500 million in AI savings while slashing jobs, Bloomberg News reports. Reuters.

Rogelberg, S. (2026, January 19). Elon Musk says that in 10 to 20 years, work will be optional and money will be irrelevant thanks to AI and robotics. Fortune.

Sen, A. (1999). Development as freedom. Alfred A. Knopf.