BRUNO BERTEZ

July 7

When will critiques of American AI become common knowledge?

Public opinion, equity markets, capital flows, and popular sentiment are all products of perception. Reality as such produces its effects only after it has passed through the thickness of the human mind—after it has overcome resistances, denials, and the collective “I do not wish to know”—that is, after it has undergone the alchemy of perception.

When perceptions remain minority views, scattered and limited in scope, they do not necessarily produce consequences.

For them to become catalytic, they must coagulate. Something must occur—something mysterious.When social knowledge coagulates, when everyone knows that everyone knows, that knowledge becomes what is termed “common knowledge”: knowledge of which the world knows that everyone knows.

The moment common knowledge is established, a formidable mechanism is set in motion.

One enters a regime of all or nothing; a tipping point is crossed.

The central question that arises today, upon reviewing a considerable body of specialized publications and analyses, appears simple yet carries formidable implications: when will the critiques directed at American choices in artificial intelligence finally produce their effect?

When will these critiques cross the threshold that transforms them into common knowledge?

The concept of common knowledge is far more than mere shared information. Common knowledge (or common knowledge in the technical sense) is not limited to the fact that “many people know something.” It designates a precise epistemic structure, formalized in game theory and the philosophy of knowledge (notably by David Lewis and Robert Aumann).Information becomes common knowledge when:

• everyone knows it;
• everyone knows that everyone knows it;
• everyone knows that everyone knows that everyone knows it;
• and so on, to infinity.

This infinite regress is not a technical detail. It is precisely what alters collective behavior.

As long as information remains merely “widely disseminated” or even “mutually known” (everyone knows it, yet it is still unclear whether others know that one knows it), it can coexist with an apparent consensus to the contrary.

The most significant formation of common knowledge I have analyzed is the one that, in a remarkably short time, precipitated François Hollande into rejection and unpopularity to such a degree that he could not seek re-election. Suddenly, through this coagulation of shared knowledge, he became utterly « untouchable ».

When the instant knowledge becomes common knowledge, agents simultaneously adjust their expectations, strategies, and public discourse. Markets tip, majority opinions reverse, coalitions re-form.It is the moment when “the mayonnaise sets,” to borrow an expression familiar in both markets and politics.

I observe converging critiques that are already substantial, yet still confined largely to private or specialized circles regarding AI. Anyone who genuinely reads the full range of recent analytical output is struck by the rise—and above all by the convergence—of criticisms directed at the American model of AI.

These critiques address several structural points:

-the scale of expenditures, judged excessive and often poorly financed;

-accounting practices and financial structures regarded as questionable;

-a dangerous circularity among investments, valuations, and capital flows;

-the near-exclusive wager on scaling (the pure increase in model size and data volumes) as the sole path to progress;

-the deliberate choice of a closed, proprietary approach at the expense of openness and broad technological diffusion;

-the dynamic of frenzied competition aimed at “being first,” at the risk of sacrificing robustness, safety, and long-term sustainability.

These observations no longer reflect isolated positions. They appear recurrently among analysts, researchers, practitioners, and financial observers.

For anyone who follows the literature closely, they have become self-evident.Yet the essential point does not lie in the publication of these critiques.The essential point lies in the path they take—or have not yet taken—toward collective awareness.

As those familiar with my work know, I frequently draw on the intellectual tools of psychoanalysis. Here I discern a social resistance to the tipping point: a form of denial, a refusal to open one’s eyes. A determined refusal to awaken, coupled with a fierce will to continue dreaming. The mechanism I wish to convey is that of the dreamer who takes pleasure in the dream and is determined, at all costs, that it should continue. We are operating under the pleasure principle—the force that pushes away the negative and unwelcome news.

For the time being, we remain below the threshold of generalized awareness. The structural flaws of the American approach have not yet attained the status of common knowledge. A form of social and cognitive resistance continues to operate: overall sentiment remains predominantly positive; the narratives of technological leadership and inevitable destiny retain their attractive power.

Critiques circulate, accumulate, and reinforce one another, yet they have not yet triggered the self-referential mechanism that transforms dispersed knowledge into an inescapable social fact.

The identical phenomenon is observable in other domains. The negative aspects of certain military policies—failures, weaknesses, inconsistencies—may be patent, documented, and analyzed in detail. They do not thereby become common knowledge. The mayonnaise has not yet set.Yet in markets as in politics, it is precisely the waves of common knowledge that produce large-amplitude reversals.

The question therefore imposes itself: what would be required for the catalysis to occur? The open question remains that of the conditions for catalysis. What events, what revelations, what sequences would be capable of tipping the entire structure? What encounters with Truth, with the Symbolic, would be necessary for the spell to be broken? There is no social psychoanalyst.

I have no definitive answer, yet several mechanisms are conceivable:

-a technical or economic failure of such magnitude that maintenance of the dominant narrative becomes impossible;

-the disclosure of financial or operational information of Enron-like proportions that renders the circularity or fragility of the arrangements visible in an irrefutable and simultaneous manner;

-a shift in posture by actors previously silent or aligned, whose public reversal would immediately create the second-order knowledge (“everyone now knows that even they know it”);

-a succession of weak signals that, by accumulating under the gaze of a now-attentive public, cross the threshold of cognitive coordination.

Nothing indicates that this moment is imminent.Nothing indicates that it is impossible either.

The history of technologies and markets demonstrates that phases of common knowledge sometimes lag for a long period, only to be triggered with disconcerting rapidity.

In the meantime, be aware that the gap is already present: the critiques are solid, convergent, and substantial—for those willing to read them.They have not yet completed the journey that would transform them into a social fact shared at every order of knowledge. It is within this interval—between the publication of analyses and the constitution of genuine common knowledge—that the fate of collective perception of American AI is currently being decided.