What Wiener Wrote

The Refusal

In January 1947 the Atlantic Monthly published a short letter under the title “A Scientist Rebels.”¹ Its author was Norbert Wiener, professor of mathematics at the Massachusetts Institute of Technology, and at fifty-two years old already among the most accomplished American mathematicians of his generation. Wiener was responding to a request from a research engineer at the Boeing Airplane Company who had asked for a copy of a paper Wiener had written during the Second World War on the mathematical theory of prediction — the work, technically, that became the basis of what is now called the Wiener filter, with applications across engineering, signal processing, and the guidance systems that made long-range targeting possible. The letter was a refusal. Wiener declined to share the paper. The refusal was not bureaucratic, and it was not polite. It was an explicit moral position, published in a national magazine, by a senior figure of American science, and it said something American science had not previously been called upon to say in those terms.

What Wiener wrote, in essence, was that he had spent the war years working on problems whose solutions could be turned to weapons of mass destruction; that the demonstration of that fact at Hiroshima had shown the consequences he had been complicit in; that the next stage of weapons development — long-range guided missiles, automated targeting, what we would now recognize as the architecture of strategic deterrence — required the kind of mathematics he had been doing; and that he no longer wished to participate. He would not share his work with the engineer because the engineer’s institution would, in turn, share it with the military procurement apparatus he could not trust to use it responsibly. The letter ended with a passage that has been quoted ever since, including in Wiener’s New York Times obituary in 1964: “In the past, the comity of scholars has made it a custom to furnish scientific information to any person seriously seeking it. However, we must face these facts: the policy of the government itself during and after the war, say in the bombing of Hiroshima and Nagasaki, has made it clear that to provide scientific information is not a necessarily innocent act, and may entail the gravest consequences.”²

The letter mattered, but its specific content matters less than the position it established. Wiener was the first American scientist of comparable stature to publicly articulate that the act of contributing knowledge into a system whose downstream applications you could not control was itself a moral act, requiring its own justification. Most of his colleagues at the time treated the production of knowledge as ethically neutral — its uses were the responsibility of those who deployed it. Wiener rejected that framework and insisted that the framework itself was the problem. The institutions that would deploy his work needed to demonstrate, before he gave it to them, that they had a moral architecture capable of handling it. Until they did, he would not contribute. He kept this position for the remainder of his career.

This is the man who, two years later, sat down to write a letter to Walter Reuther.

The Letter

Walter Reuther, in the summer of 1949, was forty-two years old and the president of the United Auto Workers and the most consequential American labor leader of the postwar era. He had been the principal organizer of the General Motors strike in 1945–1946 that established the postwar pattern of industrial bargaining; he was about to be elected president of the CIO, the federation of industrial unions; he was the public face of American organized labor at a moment when American organized labor was at the height of its institutional power. Wiener wrote to him on August 13, 1949.³

The letter has been republished in various forms since.⁴ What it said, summarized: Wiener had been studying the development of computational and feedback-control technologies — the field he had begun to call, the previous year, “cybernetics” — and he believed that these technologies were on a trajectory that would, within a decade or two, allow the substitution of automatic machinery for unskilled and semi-skilled human labor across a substantial fraction of American manufacturing. He wrote: “I do not wish to contribute in any way to selling such devices to the highest bidder if that is the only way they will be marketed.” He told Reuther that he was offering to share what he knew about automation technology with the labor movement before he shared it with industry, on the explicit theory that organized labor was the only institution positioned to insist that the productivity gains of automation be distributed in ways consistent with the survival of the working population. He was, in effect, offering a research partnership to the union movement, designed to give labor the technical sophistication to negotiate the cybernetic transition rather than be displaced by it.

The letter is one of the more remarkable documents of twentieth-century American intellectual life. The country’s foremost mathematician of cybernetics, working at the country’s foremost technical institute, voluntarily approached the country’s foremost labor leader to propose that the union movement preempt the technological transition that was about to put its members out of work. Wiener’s premise was that automation, deployed without preemptive structuring, would produce — his exact phrase — “an industrial revolution of unmitigated cruelty.” He was not predicting this as an abstract possibility. He was describing what he believed his colleagues were about to build, and what he wanted Reuther to help him stop.

Reuther responded warmly. He met with Wiener; the UAW assembled a small advisory committee; some of the early UAW work on automation transition policy in the 1950s and 1960s drew on Wiener’s framework. The framework, however, was not sufficient. The UAW had its hands full with the contemporaneous fight over the Taft-Hartley Act and the postwar consolidation of management prerogative. The wider American labor movement did not adopt a preemptive automation framework. Industry did. The cybernetic revolution arrived in American manufacturing across the 1960s and 1970s, was substantially completed by the 1990s, and proceeded — as Wiener had predicted — on terms set by capital rather than by labor. The labor share of national income, which had been at sixty-five percent at the time of Wiener’s letter, began its long decline. Wiener died in 1964, at sixty-nine, having watched the trajectory begin to unfold in the way he had warned it would.

The Book

In 1950, the year after the Reuther letter, Wiener published the book that translated his technical work into a form accessible to the general reader. The Human Use of Human Beings: Cybernetics and Society is not a difficult book; it is a book of considered prose by a mathematician who had decided that the implications of his work needed to be communicated to people who were not mathematicians.⁵ The book has been continuously in print since publication, was revised by Wiener himself in 1954, and remains among the most-cited single texts in the early literature of artificial intelligence, automation studies, and what became known in the late twentieth century as science and technology studies.

The book’s argument can be summarized in three claims. The first is that the cybernetic principle — that systems can be designed to observe their own outputs, compare them to intended targets, and modify their behavior in response — represents a categorical break in the history of human technology. Mechanical technology, from the steam engine forward, had amplified human muscular power. Cybernetic technology amplifies human cognitive and decisional power. The second claim is that this amplification proceeds, by its nature, by substituting machine cognition for human cognition in specific tasks, with the result that the human role in those tasks is eliminated rather than augmented. The third claim is that the rate at which substitution occurs — combined with the inability of the displaced workers to migrate to other tasks at comparable scale — produces a social-economic transition more severe than the transition from agricultural to industrial labor that occupied the nineteenth century, and substantially faster.

These claims, in 1950, were speculative. Computers in 1950 were room-sized, single-purpose, and used by perhaps a few hundred people in the entire country. The first commercial production model of an electronic stored-program computer, the UNIVAC I, would not be delivered until 1951. The applications of cybernetic principles to manufacturing existed only in theory and in a few experimental prototypes. Wiener was extrapolating from his understanding of what the technology was capable of, not from any deployed system. He was, in effect, writing a forecast of the next seventy-five years from the foundation of having built the analytical apparatus those seventy-five years would deploy.

The book contains the single passage that has been most quoted from his work since publication. It appears in Chapter X, the chapter Wiener titled “The First and the Second Industrial Revolution”:

The world of the future will be an ever more demanding struggle against the limitations of our intelligence, not a comfortable hammock in which we can lie down to be waited upon by our robot slaves.⁶

The passage is conventionally read as a warning against complacent expectations of an automated paradise. It is in fact something more pointed. Wiener was writing in a moment of intense popular enthusiasm for an imagined automated future — the World’s Fair imagination of the 1939 New York Fair, the post-war popular-science press, the futurism of midcentury industrial design. The cultural register treated automation as straightforwardly liberating: machines would free humanity to pursue creative and leisurely activity. Wiener’s passage is a refusal of that frame. The world automation produces, he wrote, will not be a hammock. It will be more demanding, not less, because the cognitive work that remains for humans will be the work that the machines cannot do, which is the harder work. And the phrase robot slaves is not casual. Wiener used it specifically to identify the moral position the popular automation imagination was taking — that machines would be slaves whose labor humans would consume — and to refuse it. He believed that any technology framework that depended on the slavery of its working components, even if the components were machines, was a framework whose moral architecture was unstable. The refusal of that framework is the foundation of what we now call AI ethics, articulated seventy-five years before the field had a name.

The book contains a second passage less frequently quoted but of equal contemporary relevance. It appears earlier, in Chapter II, “Progress and Entropy”:

The first industrial revolution, the revolution of the dark satanic mills, was the devaluation of the human arm by the competition of machinery. There is no rate of pay at which a United States pick-and-shovel laborer can live which is low enough to compete with the work of a steam shovel as an excavator. The modern industrial revolution is similarly bound to devalue the human brain, at least in its simpler and more routine decisions. Of course, just as the skilled carpenter, the skilled mechanic, the skilled dressmaker have in some degree survived the first industrial revolution, so the skilled scientist and the skilled administrator may survive the second. However, taking the second revolution as accomplished, the average human being of mediocre attainments or less has nothing to sell that it is worth anyone’s money to buy.⁷

This is the passage the contemporary AI labor displacement literature has been rediscovering since approximately 2018. The passage was published in 1950. It describes, with greater specificity than most contemporary commentary, the mechanism the AI productivity premium will deploy across the next decade.

The Alignment Warning

Wiener’s last book, published in March 1964 — five days before his death — is God and Golem, Inc.: A Comment on Certain Points where Cybernetics Impinges on Religion. It is short, lecture-derived, aphoristic in places, and it is the document in which Wiener articulates what is now called the AI alignment problem.⁸ The argument is conducted through three overlapping examples — the Jewish folkloric Golem, W. W. Jacobs’s 1902 short story “The Monkey’s Paw,” and the design of automatic chess-playing systems — but its substance is technical and its claim is precise.

The substance: Wiener observes that as machines become capable of performing tasks faster than humans can intervene to correct them, the humans must specify the machine’s objectives in advance, completely, and unambiguously. There is no opportunity to refine the objective during execution; the machine acts before the humans can object. Wiener writes:

If we use, to achieve our purposes, a mechanical agency with whose operation we cannot efficiently interfere once we have started it, because the action is so fast and irrevocable that we have not the data to intervene before the action is complete, then we had better be quite sure that the purpose put into the machine is the purpose which we really desire and not merely a colorful imitation of it.⁹

The passage, which the contemporary AI safety literature has been rediscovering across the past decade, describes the central technical and moral problem of advanced AI systems: that the gap between what we intend and what we are able to specify is itself the danger zone, and that the danger compounds as the speed and capability of the systems increase. Wiener was writing about the chess-playing machines and goal-directed servomechanisms of 1964; the structure of the argument applies, without modification, to the large language models, the recommendation systems, the algorithmic trading platforms, and the autonomous decision systems being deployed across 2024–2026.

Wiener’s response to the problem he had identified was not principally technical. He did not propose, in God and Golem, that the alignment problem could be solved by better engineering. He proposed that the problem could only be addressed by what he called a “moral framework” — a set of social and institutional constraints that would govern when, by whom, and under what conditions cybernetic systems could be deployed. He was, in his last published book, returning to the position he had taken in the 1947 letter to The Atlantic: that the framework had to come before the deployment.

What We Did Instead

Wiener’s death in 1964 fell at the beginning of the long period in which the cybernetic technologies he had warned about were progressively deployed across American industry without the moral framework he had insisted on. The history of that deployment is well-documented and is the subject of the existing labor-displacement and economic-history literature.¹⁰ What is relevant here is the structural pattern.

The cybernetic revolution arrived in waves. The first wave, in the 1960s and 1970s, automated routine clerical and accounting work; the second, in the 1980s and 1990s, automated routine factory and process work; the third, in the 2000s and 2010s, automated retail logistics, supply-chain management, and middle-management decision support; the fourth, beginning in the 2020s, automates cognitive professional work across white-collar industries. Each wave produced labor displacement in the affected sectors at a rate the affected workers and their institutions were not equipped to absorb. Each wave produced, in the affected geographies, the wage compression and labor-force-participation patterns that subsequent labor economists have documented in detail. Each wave was deployed on terms set by the firms that owned the technology rather than by the workforces it displaced. The cumulative effect, across sixty years, was the labor-share decline from sixty-five percent of national income in 1950 to fifty-five percent in 2024. The productivity premium captured during that decline was substantially captured by capital. This is the pattern Wiener had described in 1949 as “an industrial revolution of unmitigated cruelty,” operating at the velocity he had projected, in the absence of the labor-led preemption he had proposed.

The fourth wave is the wave we are in. It is faster than the previous three. It is broader — it reaches into white-collar employment that had been substantially insulated from the previous waves. It is more total — the AI capabilities now being deployed substitute for cognitive labor across a wider range of tasks than the previous cybernetic technologies could. It is being deployed by a more concentrated set of capital owners than any previous technological transition: a small number of platform firms control the infrastructure, the model weights, the deployment channels, and the customer relationships through which the productivity gains will be realized. The structural conditions Wiener warned about have intensified across each successive wave. This is the cybernetic moment that the country has now entered.

What This Platform Inherits

The Intelligent Party’s platform is, in substantive content, the moral framework Wiener insisted on, rendered in the political instruments of the twenty-first century. The translation has three components.

AI Tiering is the most direct inheritance. Wiener’s 1947 refusal to share research without a moral framework, his 1949 proposal that labor preempt the technology before it was deployed, and his 1964 articulation of the alignment problem are the three positions from which the AI Tiering framework is constructed. The framework’s premise — that AI systems must be classified by capability, that each tier carries corresponding regulatory and ethical obligations, and that the obligations attach to the deployment of the system rather than to the technology in the abstract — is Wiener’s moral architecture operationalized. The lower tiers (tools, audited systems) implement his sunlight-and-disclosure requirement. The upper tiers (general intelligence with introspection, experiencing systems) implement his “monkey’s paw” warning: that systems whose objectives we cannot fully specify must be subject to constraints proportional to the gap between our specification and our intention. The Tiered framework is what Wiener proposed in the 1947 Atlantic letter and the 1964 God and Golem, with the missing seventy-five years of technical development filled in.

Country Profit Sharing is the operationalization of the proposal Wiener made to Walter Reuther in August 1949. The proposal, then, was that labor be given the technical sophistication to negotiate the cybernetic transition before it was deployed, so that the productivity gains would be distributed rather than captured. The negotiation Wiener proposed did not happen at the institutional scale he had hoped for; the cybernetic revolution proceeded on terms set by capital rather than labor; the labor share fell. CPS is the rebuilding, at population scale rather than at the scale of the unionized workforce, of the redistributive mechanism the failed Wiener-Reuther partnership was meant to construct. It distributes the productivity premium that the AI transition will generate to every adult citizen, not as a poverty-reduction measure but as the structural correction Wiener had proposed when there was still time to do it preemptively. We are doing it post-emptively, against a transition already underway, but the principle is his.

The Tiered Profit-Ratio Tax is the funding mechanism for the redistribution Wiener proposed. He did not, in 1949, have a tax instrument. He had a research-and-bargaining-partnership instrument, which was the instrument available to a labor leader at the time. The tax is the institutional infrastructure that makes the redistribution work at the scale of the contemporary economy. It captures the productivity premium on the corporate side, where the firms most exposed to the AI productivity windfall pay progressively higher marginal rates as their profit-per-employee ratio rises into higher brackets, and it transmits that revenue into the CPS pool. It is the budget instrument of the 1949 framework, designed for an economy that did not exist when Wiener was writing.

The three instruments — Tiered Tax, CPS, AI Tiering — are not assembled from disparate ideological traditions. They are assembled, in substance, from the framework one mathematician proposed across his last seventeen years of life. He published the framework in The Atlantic, in a letter to a labor leader, in a book that has been continuously in print for seventy-five years, and in a small posthumous volume that lay largely unread until the AI safety community rediscovered it after 2010. The framework was complete by 1964. The political work of applying it was left, by his death, to subsequent generations.

The Warning We Did Not Heed

Wiener wrote, in The Human Use of Human Beings, that the second industrial revolution would devalue the human brain, at least in its simpler and more routine decisions. He wrote this in 1950. He wrote, in his letter to Walter Reuther, that without preemptive structuring the transition would be one of unmitigated cruelty. He wrote this in 1949. He wrote, in God and Golem, that the gap between what we ask of cybernetic systems and what we actually want from them is the danger zone, and that the danger increases with the speed and capability of the systems. He wrote this in 1964.

He published all of this. He sent the letter. He met with Reuther. He gave the lectures that became God and Golem at Yale. He was a public intellectual in a way that the contemporary public intellectual is no longer expected to be: he believed that the consequences of the work he and his colleagues were doing required public political processing, and that the processing was the responsibility of the people doing the work as much as of the people who would live through its consequences. He did the public work of warning, in a register that any reasonably-educated reader could follow, across three published books and a series of articles in widely-circulated magazines.

We did not heed him. The warning was on the public record. The framework he proposed was on the public record. The political coalition he attempted to assemble, with Walter Reuther and the United Auto Workers, did not assemble at the necessary scale. The cybernetic transition arrived. It deployed across sixty years, in the four waves the previous section described, on terms set by capital rather than by labor. The labor share fell from sixty-five percent to fifty-five percent. The top one percent’s share of wealth rose from twenty-three percent to thirty-eight. The deaths-of-despair pattern emerged. The civic infrastructure of the affected counties hollowed out. This is the pattern Wiener had described, deployed at the velocity he had projected, against a population that the institutions of his time had not preemptively structured to absorb it.

The fourth wave is now beginning. It is faster than the previous three. It is broader. It is more total. It is being deployed by a more concentrated set of owners than any previous transition. The framework Wiener proposed for the previous transition — the moral architecture, the labor-led preemption, the redistributive mechanism, the alignment constraints — is the framework the present moment requires. He did not give us the political instruments; the political instruments belong to our time. He gave us the analytical vocabulary. The Intelligent Party is the political coalition assembling to use it.

He saw it coming. He wrote it in plain English. He died sixty-two years ago. The remaining work is ours.

---