AI Is the Culmination of Industrialization

AI Is the Culmination of Industrialization

ai blog
July 9, 2026

Joel Kowalewski, PhD

The curse of the wires

On the sixth of February, 1861, with the United States weeks away from tearing itself in two, a Philadelphia newspaper called the Morning Pennsylvanian ran a short editorial under a plain and confident heading: “The Telegraph—Its Abuses.” The telegraph was then the most advanced information technology on earth, a nervous system of copper strung across a continent, and the paper had made up its mind that it was a menace. “No doubt,” the editorialist conceded, “under the control of honest and conscientious men, and confined in its operations to the transmission of facts as they really exist, and events as they really and truthfully transpire, it would be productive of much good.” But this, he went on, was not the case. Driven by the pursuit of profit, the wires multiplied unreliable reports, manufactured excitement to boost their own revenues, and produced what he called a “diseased condition of the public mind.” The magnetic telegraph, he concluded, “instead of being a blessing, is a curse to the country.”

Strike out the word telegraph, write in the word AI, and almost nothing in that paragraph needs to change. The charge is identical: a powerful new machine for moving information, run for profit, flooding us with plausible falsehood, corroding our capacity to tell what is real. I begin here not because the resemblance is a charming rhyme of history, but because it is the tell. Every information technology of the industrial age has arrived wearing the same two costumes at once—apocalypse and salvation—and the reliability of that pattern is itself an argument about what kind of thing is actually happening. We are not, I will argue, living through a rupture. We are living through a continuation, and mistaking it for a rupture is the most expensive error we can make.

Artificial intelligence is not a break in human affairs. It is the culmination of industrialization—the same two-century project of breaking the world into manageable, repeatable, optimizable parts, now finally turned on information and judgment themselves. It is nonsensical to speak of a “failed Industrial Revolution”; the phrase does not parse, because industrialization was never a product that could succeed or fail but a direction an entire civilization took. For exactly the same reason it is nonsensical to speak of a failed AI. AI is not a gadget you can evaluate and decline. It is an automating, regulating technology aimed at managing the very complexity that industrialization produced—and it has already been made essential before we have finished deciding whether we want it.

This inverts the anxious question of the moment. Everyone wants to know whether AI is a bubble that will burst, and my answer is that the real danger runs the other way: the danger is that it will not. A bubble bursts and clears—the speculative money washes out, the derivative products die, and the world resets. Industrial infrastructure does not do this. The assembly line did not burst. The electrical grid did not burst. And, to name the cautionary case I will return to, the miracle chemicals that industry bonded into nearly every product of modern life did not burst either; they became forever. The serious question about AI is therefore not whether it works well enough to justify its valuations. It is whether, once it has been woven through everything, we will ever again be in a position to remove it.

There is no rupture, only the same machine, larger

Our first move has to be against the word revolution itself, because it does most of the misleading. Lewis Mumford, in Technics and Civilization (1934), insisted that the machine is not a device but a civilizational project stretching back long before the smokestacks—that the clock, not the steam engine, was the key machine of the modern age, and that what we compress into the phrase “Industrial Revolution” was in fact a slow, layered accumulation of techniques and habits of mind. The people living through it did not experience a single rupture; they experienced a thousand small continuities, each of which felt ordinary. “Revolution” is the name later historians gave the sum. We are making the same category mistake in real time, staring at one bright interval—call it 2019 to now—and calling it a new epoch, when it is a late chapter of a book that opened with the loom and the ledger.

Consider what actually happens when an automating technology lands. When the automated teller machine arrived in the 1970s, it was sold, and feared, as the end of the bank teller. The teller did not vanish. The job was redefined—fewer hands counting cash, more moved into sales and what the banks learned to call “relationship management”—and the number of tellers actually rose for years, because cheaper branches meant more branches. This is the pattern industrialization has always followed: a technology encroaches on the familiar and does not destroy it so much as obscure it, redistribute it, and rename it. Which is why the genuinely new thing about AI is not what it can do. AI is unprecedented not because it can replace human labor—at present it largely cannot—but because of the thought that it can. In an industrial economy the belief runs ahead of the capability, and the belief alone is enough to restructure a workforce, because the executive who reorganizes around a tool that half-works is rewarded long before anyone audits whether it worked.

Why we build machines to manage the machines

The archaeologist Joseph Tainter argued in The Collapse of Complex Societies (1988) that complexity is an investment, and like any investment it is subject to diminishing returns. A society solves a problem by adding a layer—a bureaucracy, a specialization, a coordinating institution—and the first layers pay handsomely. But each additional layer buys a little less than the one before, until a society is spending enormous energy simply to maintain the complexity it already has. Industrialization is precisely such a bet on complexity, sustained for two centuries: more connection, more coordination, more interdependence, more moving parts that must be kept in synchrony. When the returns on that complexity begin to thin—when the mergers stop creating value, when the supply chains become too tangled to see—a society has only two moves. It can simplify, which no industrial society has ever willingly done, or it can find a new subsidy that lets the complexity keep paying. AI is the subsidy. It is a control technology, built to manage a world that has grown past the point where any human, or any pre-AI institution, can hold it in view.

The word control here is not loose, and it is not mine. Norbert Wiener named the science of control and communication cybernetics in 1948, and the term “artificial intelligence” descends directly from that lineage of automatic regulation—from feedback loops and anti-aircraft predictors, from the problem of steering a system too fast for a human hand. Jacques Ellul, in The Technological Society (1954), gave the deeper pattern its name: la technique, the relentless drive to optimize every means for efficiency, migrating out of the factory and into administration, into science, into conversation itself, until it becomes the water we swim in and stops looking like a choice at all. Seen this way, AI is not the arrival of a mind. It is technique reaching a kind of self-reference—a technology whose entire function is to optimize the management of other technologies. That is why the unease it produces is so hard to place. Our survival has quietly come to depend on our faith in automating systems that no one of us can any longer comprehend, and we reach for older words—intelligence, understanding, even soul—to re-enchant a world we built to be fully explained and then made too complex to follow.

AI does not end industrialization. It is industrialization becoming able to manage itself.

The parable of the forever chemical

In 1938 a young DuPont chemist named Roy Plunkett opened a canister of tetrafluoroethylene he had left frozen overnight and found, instead of the gas he expected, a slick white powder—a polymer that shrugged off heat, chemistry, and friction as nothing before it had. It was Teflon. It went first into the Manhattan Project, sealing the corrosive uranium gases of the bomb; then, after the war, into a plant in Parkersburg, West Virginia, in 1951, and from there into cookware, packaging, textiles, cosmetics, and firefighting foam—and, eventually, into the blood of very nearly every human being alive. The processing chemical that made it, PFOA, or “C8,” turned out to be toxic, and the company’s own scientists understood as much for decades. But the compounds do not break down. We call them forever chemicals for this reason, and we cannot simply stop making them, because whole industries were built on their backs and their revenue now depends on manufacturing the next PFAS-like derivative to replace whichever one was just banned.

My point about this history is not the familiar one about corporate villainy. It is structural. The brittleness—the impossibility of ever getting free—did not originate in wickedness. It originated in the lack of competition in chemical manufacturing that went back to the 1930s: a few firms, deeply entangled, each too central to fail, so that a known poison could become foundational infrastructure without anyone ever quite deciding that it should. Now watch AI run the same play. A handful of firms—OpenAI, Anthropic, and the cloud giants whose compute they ride on—are becoming the C8 of the information economy. Companies large and small are bonding these systems into their core operations; and when the systems underperform, as they routinely do, the rational move is not to tear them out but to realign the operation around the tool—to reorganize the work so that the model’s shortcomings become invisible, and the dependency becomes permanent. That is how a technology stops being optional. Not by working with increasing effectiveness, but by becoming essential to the economic interests of individuals and businesses. Economic interests are increasingly impersonal mathematical objectives. Human interests are not easily quantifiable and removed by necessity. That is what inverts the logic, giving us forever chemicals. The end result of pure mathematical optimization detached from human interests is production for production’s sake. It is a self-sufficient system that produces, recycles existing products, and maintains itself and the existing power structure. AI is inevitable simply because it is essential to satisfying the numbers. Realistically, it cannot be rejected in this economic landscape. So we gain little in becoming technophobic. There is a path toward ideal alignment with AI technologies, avoiding a similar parable in which a disagreeable form or practice becomes essential, but the current trajectory of industrialization likely does not lead here. That is the lesson of forever chemicals.

Old money, chasing new

Each generation of scholars is compelled to ask a single question: How did we arrive at the present moment? Alfred Chandler, in The Visible Hand (1977), provided a solid explanation that anticipates events of the late 20th and early 21st centuries. Writing about the American railroads, the first enterprises large enough to require a managerial hierarchy, Chandler describes the formation of a “visible hand” of salaried administrators coordinating what the market’s invisible hand no longer could, and the pattern of administrative control propagated into steel, oil, and everything after. Joseph Schumpeter called this expansion of corporate control creative destruction in Capitalism, Socialism and Democracy (1942), but he was clear-eyed about where the destruction tends: not toward a permanent churn of scrappy upstarts, but toward scale, toward the large firm that ultimately absorbs the disruption it cannot prevent. When Charles Wilson, the former head of General Motors, was asked at his 1953 confirmation as Secretary of Defense whether he could ever act against his old company’s interest, he said he could not conceive of the case, “because for years I thought what was good for the country was good for General Motors and vice versa.” The line is almost always misquoted as “what’s good for GM is good for the country,” and the misquotation is the more honest version, because it says the quiet part: governments are economic institutions that have learned to pretend they are something else. Accordingly, an AI bubble, on this reading, is unlikely to fully burst. It will centralize—into a few too-big-to-fail providers whose collapse the rest of the economy will have arranged itself to prevent. That’s the story of industrialization. The overall system, then, which includes government agencies, absorbs these disruptions accommodating business interests at the expense of the general public.

Chandler writing in 1977 did not anticipate the implications of the analog-to-digital transformation. He described a pattern and I have so far argued that industrialization repeats this pattern with seemingly mathematical precision. But my claim is also that industrialization culminates with AI (makes it inevitable). So this particular moment (the 21st century, 2026) must be distinct, and the main difference is the asset being consolidated is less tangible. It’s data. Whereas the conglomerates of old had a controlling interest in tangible resources like land, water, and mining, that had, for better or worse, a necessary human connection, digital data began to sever the connection with a physical human being. Taken to its logical extreme, the administrative or “visible hand” controlling data becomes AI. Not a person. Because this targets the upper class, the final game is for a controlling interest in AI distribution, application, and access to data that gives AI specialized insight, redefining innovation on these terms. I previously discussed the risk of production for production’s sake becoming the mathematical objective. We can now refine this as production for data’s sake. That is, production that leads to the accumulation of proprietary data for AI training, which says nothing about the questions or problems motivating the accumulation and whether they are even appropriate characterizations of a reality we would choose or desire.

Shoshana Zuboff, in The Age of Surveillance Capitalism (2019), traced the origins of the trend to the last two decades when digital platforms began systematically turning human experience into data to be captured, refined, and sold—a new commodity conjured out of something that had never before been anyone’s property. AI is what that vast accumulation is finally for: the engine that converts the stockpile into prediction and control. Thomas Piketty, in Capital in the Twenty-First Century (2013), supplied the arithmetic. When the return on capital outruns the growth of the economy as a whole—his famous r > g—wealth concentrates structurally, not because anyone cheats but because that is simply what the mathematics does when it is left to run. An automating technology that lets capital do the work formerly done by paid labor is, among its other functions, a machine for widening the gap between r and g. The industrial logic and the epistemic one are the same logic. Efficiency was always the point. And human labor is inefficient in this mathematical framework.

Hold two things at once

The temptation is to become one more voice in a long and distinguished tradition of technological lament—The Lonely Crowd (1950), Amusing Ourselves to Death (1985), Technopoly (1992), Bowling Alone (2000)—each a real and often prophetic diagnosis of a culture hollowed out by its own machines. But that literature is right about the symptoms and, I think, wrong about the cure, because it keeps treating technology as the disease when technology is the form our response to complexity takes. The crisis those books describe—the thinning of community, the commodification of attention, the retreat of the public person—predates AI by a century. AI did not cause it, and refusing AI will not cure it. You cannot walk out of industrialization by declining its latest instrument any more than the Morning Pennsylvanian could have restored the republic by cutting the telegraph wires.

We are left with the balancing act: to hold, at the same time and without collapsing them, the claim that AI is a genuine instrument of consolidation and dispossession, and that it may nonetheless be the only control technology yet built that is commensurate with a world we have already made too complex to run by hand. Complexity is not merely inscrutable networks of corporate partnerships. Technology has genuinely exposed the complexity of, for instance, human diseases, weather prediction, climate change, and the fabric of reality, space and time. We genuinely inhabit an increasingly foreign world and our continued existence as a species depends on managing and decoding complex relationships. The critic who sees only the negative has explained why we should be afraid and has left us no way to live; the enthusiast who sees only the second has sold us the necessity of control without specifying what ideal control looks like. Dropping either half is how the argument goes wrong.

Look up and you will find the wires over Philadelphia. It is the winter of 1861 all over again. The editorialist was not wrong that the telegraph was being used to disease the public mind—it was, and it did. He was wrong only to think that this made it a curse to be undone, rather than a part of ourselves, one we would have to live with, argue over, and—this is the entire work—keep in some kind of repair. This is our burden. AI is not the end of that story. It is the same machine, larger. The task was never to reject it or to worship it. It was to stay awake inside it. We can rewrite history, together.