THE AI CON

What if the most powerful technology of our century is actually a marketing trick wrapped around stolen labor, statistical guesswork, and a planet-sized electricity bill? You have been told that artificial intelligence is inevitable, that resisting it is like resisting gravity, that the only choice is to adapt or be left behind. Every headline, every CEO keynote, every breathless think piece pushes the same story: machines are about to wake up, replace you, and either save or destroy the world.

Emily M. Bender, a computational linguist at the University of Washington, and Alex Hanna, director of research at the Distributed AI Research Institute, spent years pulling that story apart on their podcast Mystery AI Hype Theater 3000. What they found is uncomfortable and freeing at the same time. The miracles are exaggerated. The harms are already here. And the people getting rich from the confusion would very much like you to keep believing it is magic.

This microbook hands you the tools to see through the smoke. By the end, you will recognize the trick, name the damage, and understand exactly how ordinary people are already pushing back.

The Great Marketing Illusion

The phrase "artificial intelligence" does almost no technical work. It is a sales label, stretched over basic automation, text generators, and classification algorithms that companies need to sound revolutionary so the funding keeps flowing. Bender and Hanna call these things synthetic media extruding machines, because that is closer to what they actually do.

While journalists chase the science-fiction question of whether AI will destroy humanity, billionaires happily debate "p(doom)" on podcasts. That conversation is convenient for them. It moves attention away from harms that already happened. In Detroit, Robert Williams was arrested in front of his daughters because facial recognition matched him to a shoplifter he was not. Porcha Woodruff, eight months pregnant, was hauled to a precinct on the same kind of broken match. In Gaza, an Israeli targeting system called The Gospel has been used to automate strike decisions on human beings.

None of that is hypothetical. None of it waited for superintelligence. The hype around future apocalypse is precisely what lets present-day apocalypse run unchecked, while companies and government agencies adopt flawed tools out of fear of missing out.

The Myth of the Thinking Machine

Large language models do not understand you. They predict the next likely word in a sequence based on patterns scraped from enormous piles of text. Bender herself coined the famous term for this with her co-authors: stochastic parrots. The parrot can produce a sentence that sounds like grief, but it has never lost anyone.

Humans are wired to project intention onto anything that talks. That is why Blake Lemoine, a Google engineer, became publicly convinced that the chatbot LaMDA was a sentient being asking for rights. He was not stupid. He was human. We meet language and we assume a mind behind it, even when there is only statistics.

The dream of Artificial General Intelligence sitting behind all this hype has an ugly lineage. The very idea of a single measurable "general intelligence" comes from the Stanford-Binet IQ test, born inside the early twentieth century eugenics movement and used to justify forced sterilization, immigration restrictions, and racial hierarchies. Silicon Valley funders who now promise a benevolent digital god are, in many cases, openly aligned with those same ranking ideologies. The god they are building looks suspiciously like their own reflection.

Modern Luddites and Ghost Workers

The Luddites were not technophobes. They were skilled English textile workers in the early 1800s who smashed specific machines that bosses used to slash wages and destroy livelihoods. They were fine with technology. They were not fine with being crushed by it. That distinction matters today.

In 2023, the Writers Guild of America and SAG-AFTRA shut Hollywood down for months, demanding contracts that limit studios from feeding scripts into generative models and from scanning background actors once to reuse their faces forever. They won real protections, not by trusting executives but by withholding labor. On the streets of San Francisco, the activist group Safe Street Rebel placed traffic cones on the hoods of driverless robotaxis to freeze them in place, exposing how the cars exist mainly to undercut human drivers and break gig-worker organizing.

Meanwhile, the illusion of smart machines is propped up by hidden humans. The company Sama, contracted by OpenAI, paid Kenyan workers less than two dollars an hour to label streams of toxic content, including descriptions of child sexual abuse and torture, so that ChatGPT would seem clean. The model is not magic. It is built on people in the Majority World absorbing trauma so wealthy users in San Francisco do not have to see it.

Austerity Disguised as Innovation

When governments cannot, or will not, fund human caseworkers, nurses, and teachers, an algorithm shows up promising to do the job cheaper. That is not innovation. That is austerity with a futuristic interface.

In Pennsylvania, the Allegheny Family Screening Tool scores families for child-welfare risk. In practice, it pulls heavily from data about poor and Black households that already use public services, and then recommends investigation, which often means children removed from their parents. It is racism scaled by software, dressed in the neutral language of "predictive analytics."

Healthcare repeats the pattern. The startup Hippocratic AI markets robotic nurses pitched as cheaper than humans, with no actual empathy or accountability. UnitedHealth's nH Predict algorithm has been used to deny extended care to elderly patients, sometimes overriding doctors. A nonprofit called the National Eating Disorders Association replaced its human helpline staff with a chatbot named Tessa, which then gave weight-loss advice to people in crisis. The pattern is now clear. The wealthy keep human professionals. The working class gets a faulty automated copy and is told to be grateful.

The Extrusion of Synthetic Trash

Generative models are flooding the information ecosystem with cheap, plausible-looking garbage, and the cleanup falls on everyone else. The training data was never bought. It was scraped. Artists, novelists, journalists, and researchers had their work ingested without consent, and what comes out the other side is sold back to the public as creativity.

The science fiction magazine Clarkesworld had to temporarily close submissions in 2023 after being buried under hundreds of AI-generated stories per month. Sports Illustrated was caught publishing articles under fake author bylines with AI-generated headshots, part of a content pipeline run by a vendor called AdVon Commerce. Amazon has been flooded with synthetic mushroom-foraging guides that could literally poison readers, and academic publishers have pulled papers featuring impossible AI-generated diagrams, including one notorious image circulated by sites like Testtomcels showing a rat with anatomically absurd proportions slipping past peer review.

Local journalism, already gutted by private equity, is finishing its collapse as hedge funds replace reporters with text extruders tuned for search engine optimization. The point was never to inform you. The point is ad impressions.

Doomers, Boosters, and Real Disasters

The loud public fight between AI doomers, who warn of extinction, and AI boosters, who promise utopia, looks like a debate. It is actually a duet. Both camps agree on the central fantasy: that current software is on a smooth ramp toward god-like capability. They only disagree about whether to cheer or panic.

Researcher Timnit Gebru and philosopher Émile Torres traced the shared worldview behind the loudest voices and named it TESCREAL, an acronym bundling transhumanism, extropianism, singularitarianism, cosmism, rationalism, effective altruism, and longtermism. Its adherents prioritize trillions of hypothetical future digital humans over the very real suffering of people in the Majority World today. Marc Andreessen's Techno-Optimist Manifesto is the booster anthem. The institutional AI Safety movement, funded by some of the same billionaires, is the doomer flip side, useful mostly because it lobbies for regulations that lock in incumbents and treat science fiction as policy.

The disaster they both ignore is physical. Training and running these models burns staggering amounts of electricity and water. Researchers estimate that a short conversation with a large chatbot can consume roughly 500 milliliters of water for cooling. Data centers are being built next to drought-stricken communities. The climate cost is not theoretical. It is being paid right now, mostly by people who never asked for the product.

Refusal, Resistance, and Real Solutions

You do not need a PhD to puncture the hype. You need a short list of questions. What goes into this system, and what comes out? How was it evaluated, and by whom? Who benefits if it works, and who is harmed if it fails? Is there a way to appeal a decision it makes about me? Most marketing collapses on contact with those questions, because the honest answer is usually "we do not know" or "we will not say."

Bender and Hanna argue for serious regulation built on data minimization, sometimes framed as Zero Trust, alongside the right to be forgotten already protected in European law. They want mandatory documentation of training data, ongoing consent rather than one-time terms of service, and real liability for automated harm. Public libraries and information literacy programs become frontline defenses in a world where chatbots churn out misinformation faster than anyone can fact-check it.

The deepest tool is what they call Strategic Refusal. It is not blanket rejection of computers. It is the disciplined choice, especially when organized through unions and professional associations, to say no to specific deployments that degrade work, care, or truth. The screenwriters did it. Nurses unions are doing it. Teachers are starting to. The principle borrowed from disability justice says it best: nothing about us, without us. Any tool built to act on a community must be designed with that community holding real power, or it should not be built at all.

The Future Is Still Ours to Design

The hype bubble will pop, like crypto and the metaverse before it. The question is what we build in the meantime. Demand transparency, defend organized labor, fund human institutions, and refuse the tools that exist mainly to extract from you. Technology is a choice, not a weather system. Decide who it serves.