A technical essay on the pace of development. Three responses from executives. A proposed law aimed at an object with no definition. A presidential comment on a social network. A riposte from the state press of a rival country. By the end, a question about capability thresholds had become a marker of belonging. This shift is unremarkable, and it deserves a close look.
This morning we reported this public explosion of the debate on slowing down. Let us take the time to consider what it produces.
What going public brings
Let us start there, because it is real and often denied by those who mourn the era of the experts.
It makes commitments costly. A promise made in public commits you far more than an internal memo. This is precisely what makes it possible to break out of the deadlock we described: individual caution is punished as long as it is not collective, and the public sphere is what makes the collective observable.
It widens who decides. A technology that affects everyone should not be arbitrated by a few dozen people, whatever their expertise. Politicisation is the name we give to the fact that others are getting involved, and that is rather legitimate.
It forces translation. An argument that has to be understood by a broad public has to be formulated clearly, which sometimes reveals that it was not.
What it damages
Three mechanisms, and they are well documented.
Positions replace hypotheses. In an expert controversy, you can change your mind: that is the whole point. In a public debate, changing your mind reads as a climbdown. Arguments stop being propositions to be examined and become flags to be held.
Nuance disappears. The real disagreement was about thresholds, conditions, timelines. The public debate retains a binary opposition between those who want to slow down and those who want to speed up, when hardly anyone holds either of those pure positions.
Partisan alignment sets in. As soon as a question becomes a marker, the answer you give depends less on what you have understood than on the camp you belong to. Once that groove is worn, it does not come undone.
Climate followed exactly this path. A question of measurement and modelling, discussed among specialists, became within about twenty years a marker of political belonging in several countries. The result is not that people have bad arguments: it is that they no longer listen to arguments at all, because the answer is decided before the question. It would be surprising if AI escaped this mechanism, and this week suggests it is not escaping it.
The role of company executives
There is a complication specific to this file, and it is rarely named.
The main participants in the debate are neither independent researchers nor public officials: they are executives of directly competing companies.
This does not disqualify their arguments — someone with an interest can be right, and they know things no one else knows. But it means that every position expressed is simultaneously an analysis and a commercial positioning, and that no one, including them, can disentangle the two with certainty.
This is what makes the public debate here murkier than elsewhere: the experts are the sellers.
What can be protected
You do not depoliticise a subject, and it would be futile to try. You can, however, preserve the conditions for a useful disagreement.
Maintain places where you can be wrong. Spaces — publications, evaluations, technical bodies — where you can put forward a hypothesis and withdraw it without that becoming a confession.
Demand criteria rather than positions. This is what we do in every article on this subject: ask not whether someone is for or against, but what would change their mind. Someone who cannot answer that question is no longer debating.
Bring in third parties. A debate in which only the interested parties speak cannot be arbitrated. This is why we relentlessly call for independent evaluations with proper resources: they serve not only to measure, but to make the disagreement decidable.
What to take away
The answer to the question in the title is no, and that is not a tragedy.
No question that involves collective organisation stays technical, and it would be unhealthy if it did. What must be defended is not the purity of expert debate, but the survival of something the public debate can refer to — measurements, thresholds, findings produced by people whose careers do not depend on the outcome.
Without that, a politicised controversy is never settled. It takes hold, and you find it unchanged twenty years later.