There is no longer any press devoted to electricity. No columnist comments on the progress of the electric motor, no public debate concerns the risks of alternating current, no international summit brings together heads of state to decide the pace of its rollout. Yet this technology remade the world from top to bottom. It has simply ceased to be a subject.
Every day we cover announcements, figures, controversies. It is useful, from time to time, to ask what becomes of all of it — and what happens when the attention goes away.
The window
Any technology that profoundly transforms a society goes through a period when it is the only thing anyone talks about. That period has an identifiable beginning, a growing intensity, and then it closes. Rarely more than two decades.
Electricity occupied newspapers, parliaments and drawing rooms for some forty years, between the first urban grids and its general adoption in the home. It was seen in turn as a promise of emancipation, a deadly danger, an instrument of industrial domination. People debated the ownership of the grids, the safety of installations, the night work it made possible.
The motor car had its own between 1900 and 1930: people discussed speed in the city, the rights of the pedestrian, liability in the event of an accident, noise, smell. Associations demanded it be banned outright. Some cities limited traffic to the speed of a man walking in front of the vehicle with a flag.
The consumer internet had its own between roughly 1995 and 2015. It was seen as a universal library, then as a danger to children, then as a democratic tool, then as its opposite.
None of these debates was settled by a conclusion. They simply stopped.
What happens when the window closes
Here is the point that interests us, and it is counter-intuitive.
One imagines that a technology ceases to be a subject because the questions have found their answers. It is the opposite. It ceases to be a subject because the answers have been taken, through concrete decisions, established habits, built infrastructure — and because it becomes too costly to reopen them.
The question of whether the car should have priority over the pedestrian in the city was not resolved by debate. It was resolved by pavements, traffic lights, road signs, a highway code and a generation of children taught to look both ways rather than play in the street. Once all of that was in place, the question no longer arose. Not because anything had been agreed: because the world had taken a shape that made it absurd.
This is how most of the great technical trade-offs are settled. Not by deliberation, but by sedimentation.
This process has already begun, and it is further along than we think. No one debates automatic spellcheck any more, even though it decides every day what strikes you as a mistake. No one comments on recommendation systems, which determine a considerable share of what billions of people watch and listen to. No one writes articles about machine translation, which has changed how entire continents relate to information. These technologies left the conversation without anything being settled: they simply became the backdrop.
The asymmetry of attention
There is a systematic lag, and it explains why the consequences almost always escape notice.
Attention focuses on capabilities, because they are spectacular, datable and easy to narrate. A model that solves an open problem, a system that crosses a threshold, an impressive demonstration: these are events.
Consequences, by contrast, are diffuse, slow and dateless. No one can write the article announcing that teenagers' relationship with boredom has changed, because that did not happen on any particular day.
The result is mechanical: attention peaks when the technology does new things, and runs out precisely when those things begin to produce their effects. We watch intently during the phase when there is almost nothing to see, and we look away when everything is being decided.
This is not a failing of journalism. It is the very structure of what makes an event.
What gets decided in silence
Four things, and they share the fact that they are very difficult to undo once in place.
Infrastructure. Data centres built for twenty years, power contracts signed through to 2048, supply chains that structure entire regions. These objects will largely outlive the debate that accompanied their construction, and they will impose their constraints on people who never discussed them.
Habits. A generation that will never have looked up information without assistance, never written a difficult text without help, never faced a question without getting an immediate answer. Not because it was forbidden, but because the other way will have ceased to be the ordinary way.
Implicit norms. What it becomes acceptable to delegate, what it becomes strange to do yourself. These boundaries shift without ever being discussed, and they end up seeming natural.
What stops being counted. This is the most discreet. Every technological transition produces losers who are visible during the attention phase, then cease to be. Coachmen, typesetters, switchboard operators were the subject of reports, then of statistics, then of nothing. Not because their lot improved, but because the subject had passed.
The particular case of this technology
One objection is unavoidable: perhaps this one is different, and will remain a subject for a long time.
It is possible, and two things suggest it. This technology touches on language, that is, on the very thing through which we conduct all our debates, including this one. A technology that settles into the medium of discussion is harder to forget than one that settles into a wall.
And it produces interlocutors rather than objects. A washing machine does not raise the question of what it is; a system that answers does so with every interaction.
But two things suggest the opposite, and they carry more weight. Familiarity does not depend on the depth of the change: it depends on familiarity. What is present everywhere ceases to be visible, whatever its importance. And this technology is spreading faster than all its predecessors, which suggests that its window of attention will be shorter, not longer.
There is a simple indicator of the moment when a window starts to close: when announcements stop surprising. A model more capable than the last no longer counts as an event; a broken record interests no one but specialists; the same debates return with the same arguments. It is not that nothing is happening — it is that what is happening has ceased to be information. That threshold is near, and it will be crossed long before the open questions are settled.
What can be done with it
This observation is not a call to alarm, and certainly not a call for more talk about AI. There is already too much noise, and noise is not attention.
It suggests three things, and they are modest.
What must be decided must be decided now. Not because the urgency would be technical, but because the window during which a society agrees to discuss a subject is limited, and it is open. Questions of measurement, of accountability, of who pays for what, of what gets delegated: this is the moment when they can still be asked out loud.
We need to build instruments that outlast attention. A debate fades; an obligation to publish figures remains. A controversy dies down; an authority that inspects carries on inspecting. That is why the boring mechanisms — mandatory reporting, independent evaluations, public statistics — matter more than declarations of intent. They are designed to work when no one is watching any more.
We must be wary of what becomes obvious. The moment when a practice ceases to be discussed is not the moment when it was approved. It is the moment when it became invisible. These are two very different things, and the second disguises itself remarkably well as the first.
What to take away
In fifteen or twenty years, there will probably be no media devoted to artificial intelligence any more, just as there are none devoted to electricity. Not because it will have failed or because it will have ceased to matter, but because it will be everywhere and what is everywhere is no longer noticed.
The questions we are asking today — who is accountable for mistakes, who pays for the infrastructure, what deserves to be delegated, what we owe to future generations — will not be resolved by then. They will have ceased to be asked, which looks a great deal like a resolution and is not one.
That is why this moment matters more than it appears to. Not because the technology is at a turning point: there will be others. But because we are in the brief period when a society still consents to ask itself what it is doing.
That period will not reopen. It did not reopen for any of the technologies that came before us. And what we make of it will be the backdrop in which others will live without ever knowing it was chosen.