Salomon's House

In the [YOU] on AI Field Guide

The cycle that [YOU] on AI opens is in part a story about how knowledge-as-power concentrates: how the same capability that makes AI useful also makes it the most significant concentration of unaccountable influence in the current moment. Salomon's House is the concept that names the institutional form this concentration takes—and Bacon's own failure to account for its dangers is the sharpest available demonstration of why the form is inadequate.

Bacon's Fathers are good by stipulation. They are wise, benevolent, and trustworthy—within the fiction of the utopia. Bacon did not need to argue for their goodness because he wrote it into the premise. The real world does not work this way, and Bacon's own career proved it. Lord Chancellor of England, trusted with the highest judicial office in the realm, he was impeached in 1621 for accepting bribes from litigants with cases before his court. He confessed. The prophet of Salomon's House fell to exactly the corruption his utopia refused to reckon with: that character alone does not withstand unchecked power, however sincere the character and however real the capability. The Baconian frame reorients the governance debate from the qualities of the builders to the structure of the constraints.

The secrecy clause—the Fathers decide in private what to reveal—is where Bacon's description lands hardest. Every AI laboratory now conducts a version of this deliberation: which capabilities to disclose, which model weights to publish, which risks to communicate and which to manage internally. The arguments on all sides are serious; some withholding genuinely reduces harm. But the structure Bacon identified is inescapable: the institution that holds the capability decides how much of it the world may have, and on what terms. Whether the institution makes these decisions well is a question about the characters of its leaders. Whether it should be the institution making these decisions alone, without external accountability, is the question Salomon's House forces us to ask.

Origin

Bacon wrote The New Atlantis in the last years of his life, after his fall from office, and left it unfinished at his death in 1626. It was published posthumously in 1627 alongside his Sylva Sylvarum. The fable—a ship of European sailors blown to a hidden Pacific island—is the thinnest of fictional frames for Bacon's real interest: the institutional design of a knowledge factory. He had been arguing for organized, state-funded research throughout his career, in the Advancement of Learning (1605) and the Novum Organum (1620), but only in the utopian format of The New Atlantis did he draw the full picture of the institution he imagined.

The Royal Society, founded in 1660, was partly inspired by Bacon's vision of Salomon's House, as its early members acknowledged. The subsequent history of organized research—from the nineteenth-century research university to the twentieth-century industrial laboratory to the government research agency to the private technology laboratory—is a history of institutions that instantiate, in varying forms, the Baconian premise that knowledge-as-power requires institutional organization to produce it at scale. The contemporary AI laboratory is the latest and most powerful member of that lineage, distinguished from its predecessors primarily by the scope of the general capability it pursues and the speed at which that capability is developing.

Key Ideas

The effecting of all things possible. Bacon's formulation of Salomon's House's mission—the enlarging of the bounds of human empire, to the effecting of all things possible—is the most ambitious institutional mandate ever stated, and it describes the contemporary AI laboratory's self-understanding more accurately than most of those laboratories' own mission statements. The pursuit of general capability rather than particular application, the willingness to pursue the full range of what intelligence can do, the assumption that the limits of the possible are the only limits worth respecting: these are the Baconian premises, and they are the premises of frontier AI development.

The governance gap. Salomon's House lacks any mechanism of external accountability. Its Fathers govern themselves, check each other, and decide in conscience what to do with their power. Bacon's defense of this arrangement is not argued but assumed: the Fathers are good. His own career demonstrates that the assumption is insufficient. The governance debate around frontier AI is, at its core, the debate about what external mechanisms of accountability are appropriate for an institution that holds the power to effect all things possible and currently exercises the Fathers' discretion about what to reveal.

Knowledge-power as a distribution problem. The Baconian fusion of knowing and doing has a structural consequence: wherever knowledge concentrates, power concentrates. The decision about which capabilities to release is not only a decision about safety; it is a decision about who gets the power. To withhold a capability is to keep it; to release it selectively is to distribute it selectively. Every release policy is simultaneously an allocation of knowledge-as-power among organizations, states, and publics. Bacon's Fathers deciding what to reveal were adjudicating this distribution in Bensalem's first instance; the same adjudication now happens at the scale of the global economy.