
Rejecting the Silent Consensus: A Case for Human Primacy

Adapted from a recent Vik-Bailey Lecture at Harvard, the essay opens with what the author describes as one of the most significant cybersecurity events in history: a set of disclosures some call “Black October,” where critical vulnerabilities in core internet infrastructure were found by frontier AI systems. He frames this as an urgent race between using the beneficial side of a dual-use capability and defending against its malicious use. Such capabilities will keep arriving, and AI policy will be made amid the resulting clamor. He wants to turn instead to the long-term question of institutional future, because the technology industry has plenty of positive visions for flying cars, robots, fusion, and AGI, but lacks a positive vision for how society and institutions will work.
The vision that does circulate is usually unspoken: the silent consensus. The essay imagines a company whose AI systems outperform humans in research, development, negotiation, capital allocation, and manufacturing. Human managers and owners, even when augmented by AI, have a market incentive to take their hands off the wheel. If this repeats across the economy, human beings could be driven by competitive logic to forfeit primacy over global commerce. No AI rebellion and no malicious owner is required. Depending on what the machines want, the outcome could be comfortable dependence or broad disempowerment, perhaps with a tiny human elite. The common thread is termination of human primacy. He is careful about the term: no “we” exercises top-down control of the world, but humans collectively are still the prime actor on Earth. The silent consensus is not an explicit agreement that this future is good, nor a secret plan by AI builders; it is the shared assumption that the forfeiture is inevitable, or at least that no other outcome is robust to superintelligent machines.
The author rejects this consensus. He is unapologetically a human chauvinist, but he grounds human primacy in the past rather than in a monopoly on intelligence. Humans came up from the mud, built society, discovered morality and law, made art and history; that legacy makes humans stewards of civilization and participants in what Michael Oakeshott called “the conversation of mankind.” A machine that has read the written record is not an heir to that legacy, any more than reading someone’s family letters makes you family. The right to walk one’s own path does not depend on being smarter than a neighbor; a citizen does not lose standing because someone else has better judgment. He sees no reason a more intelligent machine should dissolve that standing. He also rejects the claim that AI lacks genuine understanding; he calls that position complacency. Still, he acknowledges he could be wrong. If fully autonomous corporations outcompete human enterprise, economic logic could hand ever more capital to machines. He draws a possible parallel to the agricultural revolution, which may have worsened human wellbeing for millennia.
For retaining primacy, the essay calls for accountability: AI developers should be accountable for unsafe practices, and firms and individuals should be accountable for misuse or lack of oversight. This needs technical infrastructure tying agents to humans, with guardrails on any agent that participates in the economy. Autonomous agents forming corporations, owning property, or litigating should be granted exceptional caution, if at all. Permissions should be intelligible, limited, auditable, and withdrawable. He pairs this with commitments to privacy, anonymous speech, and individual sovereignty over data, including the practical ability to move data elsewhere. Systems should help people acquire understanding and make decisions, not only carry out tasks. He calls this a starting point, not a complete blueprint.
The closing move is to name the silent consensus openly and widen the scope of ambition: choose to let AI enlarge humans rather than shrink them, become a new kind of “man expanded” rather than a transhuman passenger. Looking at his baby son, the author says he does not want the child to inherit a nicer car with a plusher passenger seat; he wants him to have the wheel.


