Global Governance & Public Policy
efficiently learn on your own. That asymmetry is collapsing, not because AI is superior to human expertise, but because it makes the products of expertise accessible without the institution that housed it. A farmer in Nairobi can now get a crop disease diagnosis that would have required a university extension agent ten years ago. The agent still exists. The farmer does not need her anymore. The second was coordination cost. Organizing collective action among strangers was expensive and slow, which is why humans built legislatures to aggregate preferences and banks to pool capital. Those costs are collapsing. AI makes the friction of institutional mediation feel optional, feel like overhead, feel like the thing standing between a question and its answer. The third was trust. Not trust in any individual judge or reporter, but in the institution itself, the accumulated weight of its history and its procedures. Gallup has tracked confidence in American institutions for decades. In the nineteen-seventies, roughly seventy percent of Americans expressed confidence in major institutions. By 2025, only three retained majority confidence. Congress stood at ten percent. Newspapers at sixteen. The criminal justice system at fourteen. The trust decline predates AI. It began decades ago. But AI is pouring solvent into every crack. When anyone can generate a convincing legal brief, a fabricated news article, or a synthetic expert voice, the signals that once distinguished institutional authority from noise become unreliable.
The floor is giving way faster than anyone can repair it. The obvious objection is that we have heard this before. The printing press broke the Church’s monopoly on scripture. The telegraph collapsed coordination advantages. Television eroded trust in print. Each time, institutions adapted. The answer is structural. Those technolo- gies attacked one institutional dependency at a time. The printing press disrupted in- formation asymmetry but did nothing to coordination costs. Television eroded trust in one medium while building it in another. In each case, institutions could absorb the blow because the other two legs of the stool held firm. AI is kicking all three legs at once. It collapses information asymmetry, slashes coordination costs, and degrades the trust signals institutions depend on, not sequentially but simultaneously, and not for specialists but for everyone with a phone. The instinct to regulate AI is correct. But regulation is an institutional act. It requires legislatures to draft laws, courts to enforce them, and citizens to trust that both are ac- ting in good faith. If AI is eroding that trust, then regulation faces a timing problem that has nothing to do with the quality of the re- gulation itself. Regulation is medicine. But medicine requires a patient who trusts the doctor. If the illness is the erosion of that very trust, the medicine must work before the patient stops believing it can. And every actor in the conversation has reasons not to see the full picture. Academics write from universities. Journalists write from media companies. Regulators write
from agencies whose existence presupposes that regulation works. For all of them, the Industrial Revolution metaphor is more than analytical convenience. If AI is the new steam engine, everyone keeps their position. If AI is the Agricultural Revolution in reverse, the university conducting the analysis and the agency drafting the regulation may not survive the phenomenon they are attempting to describe. Thomas Kuhn documented this identical dynamic in scientific paradigm shifts: the individuals most invested in the existing paradigm are the last to perceive that it is changing, not because they are dishonest, but because acknowledging the shift means acknowledging that their own position may not survive it. The Agricultural Revolution gave humanity roughly eight thousand years to build the institutions it needed. Eight thousand years between the first granaries and the earliest known legal codes. Between the first surplus and the first written law. Those eight thousand years were not gent- le. The gap between creating conditions that demanded institutional life and actua- lly constructing institutions capable of sus- taining it was long, violent, and survivable only because the pace of change permitted thousands of years of iterative failure. We do not have thousands of years. The conditions that produced every existing institution are dissolving in years, not millennia. The institutional trust that took centuries to accumulate is being withdrawn at the pace of a bank run that no deposit insurance can cover. Çatalhöyük persisted for roughly a thousand years. Granaries packed with surplus. Walls painted with scenes of the lives its residents were assembling. No streets between the houses. No public space. No way to move among strangers without passing through someone else’s home. Humanity built streets eventually. The question that matters now has nothing to do with whether new coordination mechanisms can be constructed. Of course they can. The question is whether they can be constructed fast enough. Agriculture gave eight thousand years. AI is giving us maybe ten.
“AI is not changing how we work. It is changing how we organize.”
DAVOS CLUB 29
Powered by FlippingBook