When the monster becomes the wall socket

On 11 October 1889, a Western Union lineman named John Feeks climbed into the web of cables above Centre and Chambers Streets in Manhattan. The telegraph wires he was working on were supposed to carry low voltage. Somewhere in the tangle, they met a high-voltage lighting circuit.

Feeks died above the street while a crowd gathered below. Two weeks later, the satirical magazine Judge put an electrical demon on its cover. A skull shaped like a light bulb glowed at the top of a pole. Wires spread from it like the legs of a spider, catching horses and people in a city that appeared to have lost control of its own invention. The title was “The Unrestrained Demon.”

Seen from a room lit by safe lamps, the image can look like a period joke at the expense of frightened Victorians. We know how the story ends. Electricity becomes an ordinary utility, then a condition of modern life. We build hospitals, homes, communication networks, and climate policy around it. The demon becomes the wall socket.

That ending invites a reassuring lesson about technological change. People panic. Cartoons exaggerate. Progress continues. The feared machine settles into daily life, and the next generation wonders what the fuss was about.

The lesson is attractive because it turns history into a forecast. It is also too easy. The wires over Manhattan really were killing people. The machines attacked by the Luddites really did threaten wages, skill, and control over work. Other alarms, such as the Victorian diagnosis of “bicycle face,” tell a different story: an old social order dressing its discomfort as medical advice.

Past fears do not tell us whether today’s fears about AI are right. They teach us how to listen more carefully.

What the demon knew

Electricity did not become safe because the public learned to stop worrying. It became safer through engineering, standards, inspection, legal conflict, public investment, and rules about where dangerous equipment could go.

New York’s overhead network mixed telegraph, telephone, and lighting lines on crowded poles. Insulation failed. Companies resisted the cost and disruption of moving wires underground. City and state authorities spent years forcing a change that the industry had reasons to delay. The spectacle above Centre Street made a hidden contest visible: who would carry the risk of electrification while its owners pursued the benefit?

The Judge cover distorted the danger into a supernatural creature. It also described the street truthfully enough. A system that no single company appeared able or willing to control had entered public space. People beneath it could not inspect the wiring, choose a safer route for the current, or negotiate the level of danger they would accept.

Society kept electricity and changed the material and legal conditions under which it could become familiar. Public acceptance followed work on the danger; familiarity was never a substitute for that work.

We often forget this middle part. Once an infrastructure works, its protections merge with the technology in our memory. We point to the light and call electrification a success. The conduits, circuit breakers, building codes, trained workers, public inquiries, and political fights recede from view. Safety comes to look like a natural property of the invention rather than a settlement people demanded around it.

This matters when electricity is offered as an analogy for AI. If AI becomes as common as electric power, that would say something about dependence and scale. It would not prove that early critics had misunderstood the risks. Electricity became indispensable after its dangers changed streets, laws, professions, and institutions. Its ordinariness is partly the record of those disputes.

The men who broke the frames

The word “Luddite” has survived as a mild insult for someone who rejects technology through ignorance. The historical Luddites were more exacting.

Textile workers in Nottinghamshire, Yorkshire, and Lancashire attacked particular machines and employers during the disturbances that began in 1811. They worked in different trades and defended different customs, but their protests repeatedly concerned the terms under which machinery was used: lower wages, diluted skill, deceptive manufacture, and the loss of rules that had given workers some claim over quality and livelihood.

An 1812 declaration by framework knitters invoked an old charter against frames used to make goods in a “fraudulent and deceitful manner.” This was not a manifesto against every machine. Many of the protesters were skilled machine workers. The frame was objectionable when an owner used it to evade standards, replace trained labour with cheaper labour, or produce inferior goods while keeping the gains.

The state answered with troops, punishment, and laws that made machine-breaking a capital crime. Industrial production continued. Mechanised weaving eventually made cloth far cheaper and helped turn durable clothing from a scarce possession into an ordinary good. Later machines, including the Jacquard loom with its punched cards, became part of the ancestry through which we tell the history of computing.

Those outcomes did not make the Luddites mistaken about their own lives. A technology can enlarge total wealth while ruining particular workers, weakening their bargaining power, and changing the meaning of a craft. Cheap cloth was a real gain. So were the wages, time, autonomy, and knowledge that workers lost. They belonged to different people and arrived on different schedules.

Calling the whole period progress compresses that distribution into one word. It lets the eventual consumer price answer a question the workers were asking about power.

The Luddite example is therefore less comforting than the familiar analogy suggests. It warns against treating technical capability as an independent force to which society can only adapt. Factory owners chose how to introduce machines. Parliament chose whose property to defend and which forms of protest to criminalise. Workers had few lawful ways to bargain collectively. Machinery entered a political arrangement and strengthened some parts of it.

AI also arrives through choices about ownership, job design, pace, surveillance, credit, and the division of gains. A worker who objects to an automated system may be defending a task they value. They may also be defending the wage, discretion, apprenticeship, or leverage attached to that task. “Resistance to technology” is too blunt a name for either concern.

Fears that guarded an old boundary

Some historical anxieties deserve less rehabilitation.

During the bicycle boom of the 1890s, newspapers and medical commentators circulated warnings about “bicycle face”: a fixed, exhausted, or unfeminine expression supposedly produced by the strain of riding. The concern fell especially heavily on women, for whom the bicycle offered unusual freedom of movement. Other warnings touched fertility, clothing, propriety, and the danger of being beyond supervision.

Cheap novels had attracted similar suspicion long before. Critics worried that immersive fiction would inflame emotion, corrupt young readers, waste time, and give women access to desires and worlds that escaped household authority. The supposed injury moved easily between mind, body, and morals.

Here the feared object was entangled with a person who was becoming harder to govern. A bicycle carried a woman away from the route chosen for her. A circulating book gave a reader private company. Medical and moral language made the threatened boundary sound natural.

These episodes are useful because the critics look so obviously wrong to us. They are also dangerous as ready-made parables. Laughing at “bicycle face” can encourage us to place every concern about a new technology in the same cabinet of obsolete fears. That move protects the new system from scrutiny by associating its critics with the silliest people in the archive.

The archive itself needs scrutiny. The often repeated claim that early railway passengers were expected to suffocate or suffer bodily collapse above twenty or thirty miles per hour is difficult to trace to reliable contemporary evidence in the form usually quoted. A story can become familiar because it flatters the present. We enjoy meeting ancestors who feared what we now find harmless.

Technological history contains genuine panic, satire, commercial propaganda, practical danger, labour conflict, and later folklore. They should not be blended into a single human tendency called fear of change.

How a technology disappears

A mature technology often becomes invisible. We stop seeing it as an invention and encounter it as the way things are.

The light comes on. The jumper has a low price. The book waits on a shelf. The bicycle belongs to the street. Each object carries an earlier conflict inside an arrangement that now feels settled.

Invisibility can mean that a technology has earned trust. Engineers understand its failure modes. Rules have been written. People know how to use it. Repair is available. The benefits are broad enough that withdrawal would feel like deprivation.

It can also mean that costs have moved out of sight. Electricity is clean at the socket while generation, mining, transmission, and waste happen elsewhere. Cheap clothing depends on workers and landscapes the buyer may never encounter. A platform feels automatic after the labour required to train, moderate, maintain, and correct it has been hidden behind the interface.

Normalisation settles habits before it settles every moral account. People build homes, jobs, institutions, and expectations around the system. Refusal becomes expensive. The technology then appears inevitable partly because society has removed the practical routes around it.

Indispensability is a poor verdict on an earlier dispute. A system can become essential because it is beneficial, because institutions were rebuilt around it, or because alternatives were allowed to disappear. Usually all three processes are present.

The better historical question is not whether people eventually accepted the technology. It is what had to change, who had to yield, which protections were won, and which losses became ordinary enough to escape notice.

Before AI becomes ordinary

AI has already begun its movement from spectacle to infrastructure. Generated text enters office documents, search, customer service, education, software, medicine, and public administration. The blank box on a screen is gradually becoming a layer inside other systems, encountered without ceremony and sometimes without a clear announcement.

Anxiety gathers around it. Some people fear mass unemployment. Others fear a slower degradation of work: fewer entry-level tasks, lower fees, tighter measurement, and less control over how expertise is used. Teachers worry about writing without learning. Artists object to the use of their work and to markets flooded with plausible substitutes. Citizens worry about surveillance, discrimination, fabricated evidence, concentrated ownership, energy use, and decisions that no reachable person can explain. Some people fear emotional dependence on systems designed to simulate attention.

These claims will not all prove equally important. Some will fade as people learn what the systems can and cannot do. Some will be answered by better design. Some will survive because they name a conflict of interest that technical improvement cannot remove. The ability to generate a better answer does not decide who owns the training material, who loses income, who may appeal, or whose purposes the system serves.

History gives no permission to dismiss the entire field of concern as another “bicycle face.” It also gives no reason to treat every disturbing prediction as an approaching fact.

We can ask what kind of claim each fear contains. Is there a demonstrated hazard, like live current in a crowded street? Is a worker describing how an employer will use a machine against them? Is an established profession protecting a privilege that wider access may rightly weaken? Is a moral warning trying to preserve someone’s dependence? Is the story itself documented, or repeated because it makes the present feel wise?

The answers lead to different responses: safety standards, bargaining rights, public alternatives, disclosure, education, liability, limits on use, or patient observation. “Embrace” and “ban” are rarely enough to govern the conditions under which a technology enters a life.

One day, much of AI may be as unremarkable as a motor in a lift or current behind a socket. If that day comes, people will be tempted to say that today’s anxiety was the unavoidable price of meeting the future.

They should look again at the demon above Manhattan. The public was frightened because the wires were dangerous. Electricity became ordinary after people fought over the streets, forced hazards into view, and built protections into the network. The monster disappeared partly because the world around it changed.

Our task is to build institutions capable of hearing enough inside that fear to make the ordinary future worth inhabiting.