job displacement - SKEPSIX.

TL;DR: Half of Americans are more worried than excited about AI, and that’s not ignorance. Energy drain, job displacement, and systems that confidently hallucinate on the things that matter most are legitimate grievances. But the real danger both sides miss is cognitive offloading: handing your thinking to a tool that still fabricates facts and calls it analysis.

Why half of America sees AI as a threat, not a tool

According to a June 2025 Pew Research Center survey, 50% of U.S. adults feel more concerned than excited about AI in daily life — up from 37% in 2021. Only 10% say they are more excited than concerned. That gap is not a blip. It has hardened over four years of AI becoming more capable and more present, and it reflects something more structural than media hysteria.

Tech insiders routinely frame this wariness as ignorance. The data doesn’t support that. Frontier models hallucinate on 1.0 to 2.5 percent of summarization tasks in 2026, a real improvement from 3 to 8 percent in 2023. But on long-tail factual queries — the kind that involve obscure medical details, local regulations, or niche technical specs — hallucination rates remain at 15 to 40 percent even on the strongest models. A system that confidently invents facts on niche questions is not something most people will trust with their taxes, their medical records, or their children’s education. That distrust is calibrated, not irrational.

The trust deficit deepens when AI’s apparent competence is revealed as a mirage. The MIRAGE paper from early 2026 showed that leading multimodal models could generate detailed, plausible medical diagnoses for X-rays they had never seen — sometimes beating benchmarks designed to test visual understanding — without any image input at all. The models weren’t reasoning about pixels. They were exploiting linguistic shortcuts in the test questions. To the non-technical public, this looks not like a tool that makes mistakes, but like a system that is fundamentally deceptive.

Energy, jobs, and opacity: the grievances that make AI feel like a hostile takeover

The discomfort goes deeper than accuracy. People see AI arriving not as an opt-in assistant but as an ambient condition of modern life. Microsoft embeds Copilot into Office. Google replaces search results with AI Overviews. Adobe pushes generative fill into Photoshop. The default is on, and opting out is a chore.

At the same time, the infrastructure powering these features carries a visible environmental cost. Hyperscalers are spending tens of billions of dollars on GPUs and building data centers that consume water and electricity at scales that strain local grids. The public sees billionaires siphoning resources to build opaque computing complexes that feel less like public infrastructure and more like private extraction. The framing is not “this will save lives”; it’s “we can’t afford to miss the next platform shift.”

Creative professionals, translators, voice actors, and entry-level knowledge workers see their economic niches being hollowed out — if not by AI directly, then by the promise of it, used as justification for headcount reductions and contract renegotiations. The fear is not speculative. It’s visible in the gig economy, in publishing, in game development. When the same technology that threatens your livelihood also gives you wrong answers and demands you accept it because resistance is futile, “evil” is an emotionally accurate descriptor, even if it’s technically imprecise.

Mocking the skeptics as NPCs only guarantees a deeper chasm

A recurring response from technology communities is to pathologize the critics. The habit of treating anti-AI views as the domain of uninformed Luddites who simply don’t understand the math has real consequences. When an entire demographic is told their legitimate grievances — energy consumption, algorithmic wage suppression, forced adoption — are just noise, they don’t become more rational. They dig in.

The anti-AI identity becomes fused with a broader distrust of institutions, making it nearly impossible to have granular conversations about which uses of AI are genuinely beneficial and which are exploitative. The result is a self-reinforcing cycle: industry dismisses the public, the public demands sweeping regulatory crackdowns, industry cries overregulation, and the middle ground evaporates. A techno-optimism that refuses to acknowledge harm is just as blinding as a doomerism that refuses to see any benefit. Both positions are ideological, not analytical.

The more honest framing is the one that most people land on eventually, often after using AI themselves: it is a tool that multiplies whatever you bring to it. For someone who already knows how to do their job well, it is a force multiplier. For someone whose job consists of predictable, repetitive tasks that AI can approximate, the threat is real and immediate. Both things are true at the same time, and neither audience is stupid for their reaction.

The underreported danger: cognitive offloading, not job replacement

For all the attention given to labor disruption, the more insidious hazard of mass AI adoption is cognitive offloading: the gradual transfer of thinking, reasoning, and judgment to a system that is not reliable enough to bear the weight. This is not a future scenario. It’s happening now when students use language models to write essays without evaluating the arguments, when doctors accept AI-generated differential diagnoses without verifying clinical reasoning, when programmers paste code they don’t fully understand because the tests pass.

A radiologist who offloads image interpretation to a system that fabricates clinical findings — not as a second opinion but as a primary reader — is not being replaced. They are being deskilled in a way that compounds error silently. The same dynamic plays out in legal analysis, engineering design, and journalism. The focus on “evil AI” masks this quieter erosion.

The non-technical public fears losing agency to machines; the tech crowd insists those machines are just tools, like calculators. Both miss the point. Calculators don’t write paragraphs of plausible-sounding nonsense when you ask a question outside their training distribution. An AI that hallucinates on long-tail facts while sounding authoritative is not a calculator. It’s a confidence machine with a broken calibration. Using it as a cognitive crutch doesn’t make you a cyborg. It makes you an unwitting amplifier of its errors.

A practical rule for using AI tomorrow without giving up your thinking

The solution isn’t to reject AI or to embrace it uncritically. It’s to reimpose the cognitive friction that the technology is designed to remove. For any task where the answer matters, adopt a single rule: verify at the source, not from the model’s own output. If you ask for a fact, follow the citation. If the model summarizes a document, read the original. If it generates code, understand the logic before merging.

For fact-heavy queries, combine retrieval-augmented generation with high-quality sources, which benchmarks show can reduce hallucination by 50 to 80 percent. When the stakes are high, use function calling to query authoritative APIs directly, avoiding the model’s internal knowledge entirely. And when a model starts sounding too smooth, pause and ask: would I be able to explain this answer without the AI? If the answer is no, you’ve already offloaded more than you should.

None of this will make the AI evil perception disappear. But it shifts the question from “is AI evil?” to “am I using it in a way that preserves my own judgment?” That is a question anyone, tech-savvy or not, can act on tomorrow morning. The fear isn’t wrong. The response to it can still be right.

TL;DR: Meta reassigned 7,000 employees to train the AI replacing their laid-off colleagues. The result: a parody video, calls to poison the data, and a model that will hallucinate exactly where humans were most valuable.

When morale collapses, the training data collapses with it.

When Meta laid off 8,000 workers and reassigned another 7,000 to train AI models, a departing engineer posted a parody video set to “American Pie” on the company’s internal message board. The video captured a sentiment that had been simmering for weeks: you are being asked to build the thing that will make you redundant. The clip’s popularity was not just a morale collapse. It was the visible tip of a practical problem companies rarely acknowledge. Forced AI training mandates, combined with large-scale layoffs, create the conditions for unreliable training data and models that fail in ways nobody is measuring.

The real risk is not that employees will quietly comply. It’s that resentment turns into deliberate data poisoning, or equally corrosive, that demoralised workers stop surfacing the errors the model inevitably makes. Once the models are deployed, the business inherits a system that looks competent on a dashboard but hallucinates on the exact tasks that used to require human judgment. The fix is not better model cards or more alignment research. It’s a set of defensive moves that employees can take right now without resorting to sabotage.

The numbers behind the video: 8,000 layoffs, 7,000 reassignments, one pipeline.

Meta’s spring 2026 restructuring was extreme but not unique. The company explicitly tied the headcount reduction to a strategic pivot toward AI, and hundreds of the retained employees were moved into roles where their daily output would directly train the models intended to absorb their former colleagues’ work. The internal message board began filling with calls to “confuse the AI,” feed it false information, and trigger infinite recursion loops. A software engineer named David Frenk posted a farewell video set to the chords of “American Pie” and it spread instantly, becoming a rallying point.

The practical problem hits the machine learning pipeline immediately. Training data from workers who resent the project carries a higher noise floor. But even when employees stay honest, the mechanism of forced participation erodes a subtle safety net: the informal feedback loops that catch edge-case failures during normal operations. When people stop caring because they believe the model is there to eliminate them, they stop correcting it. The business is then flying blind, substituting the false confidence of a leaderboard metric for the messy reality of work.

The deeper reason forced AI training backfires: language models are built to guess, not to admit ignorance

Most commentary treats this as a labour relations story. The more consequential layer is technical. Large language models hallucinate because their training and evaluation pipeline treats a guess that sounds plausible as better than a refusal to answer. In pretraining, any incorrect statement that matches the statistical shape of a fact is rewarded if it minimises cross-entropy loss. In post-training, the dominant benchmarks use 0-1 scoring: a wrong answer and a hesitant “I don’t know” are equally penalised, so the model learns to produce confident-sounding outputs even when uncertainty would be the accurate response. This is the same guessing mechanism that drives AI moderation false positives — the model never learned to say it doesn’t know.

When a company asks the very people who understand the domain to train the replacement model, the hope is that those workers will inject precision. But the incentive structure of the model itself pushes toward overconfidence. Employees know the edge cases. The model, rewarded for guessing, will paper over them with fluency.

Data on hallucination rates backs this up. On summarisation benchmarks, the best frontier models in 2026 still hallucinate on about 1.0 to 2.5 percent of outputs. On tasks that require integrating retrieved context with internal knowledge, the rate jumps to 4 to 9 percent. And for long-tail facts, exactly the kind of niche, company-specific details that an internal model would need to nail, hallucination rates sit at 15 to 40 percent even on the most capable systems. A model trained under duress on the tacit knowledge of staff who are being shown the door will inherit those error rates, and the errors will land precisely where the human was most valuable.

What workers can do when asked to train their own AI replacement

Sabotage is not a solution. Deliberately poisoning data is detectable, often illegal, and ultimately makes the model worse in ways that hurt the people still using it. Three practical alternatives give employees leverage without ethical or legal exposure.

Document the model’s failure modes instead of sabotaging it. Every time the internal AI generates an output that is factually wrong, misattributes a source, or draws a false connection, record it. Create a simple log with the prompt, the model’s response, and what the correct answer should be. This is not obstruction. It is the quality assurance step the business claims to want. A growing catalogue of concrete errors shifts the conversation from “the model is generally capable” to “here are the 47 tickets it got wrong last week.” A well-maintained failure log makes the case for human-in-the-loop oversight in terms the organisation understands: cost of error.

Shift your work toward decisions the model isn’t rewarded to make. The same evaluation structures that reward guessing also leave a gap: whenever the correct answer requires surfacing doubt, checking a citation, or abstaining, the model underperforms. If your role involves deciding when to trust a source, verifying a chain of reasoning, or determining that a question is unanswerable with current information, you are operating in the space where models are most brittle. Make that portion of your work explicit. Write down the judgment calls the AI cannot make and tie them to concrete business outcomes.

Treat the AI as an unreliable junior, not a replacement. Forced-ranking a human against a language model rarely makes sense when the model hallucinates on internal data. The most defensible posture is to treat the AI as a first-draft tool that must be checked. Propose a process where the model produces an output, a worker validates it against known ground truth, and the corrections feed back into a dataset used only to improve retrieval, not to replace the checker. This turns a threat into a collaboration that mirrors what the best-performing RAG systems already do: retrieval with high-quality sources cuts hallucination by 50 to 80 percent. The person who designs that feedback loop becomes the one the business cannot remove.

Workers in rote documentation roles feel the pressure now. Judgment-heavy roles have breathing room.

The employees most immediately exposed are those whose output is close to the training pipeline: handling routine support tickets, generating status reports, or organising inboxes. These are tasks the model can handle passably until it silently mangles a critical date or client name. In those roles, AI deployment is already underway and the need to build a failure log is urgent.

Workers in roles that depend on weighing conflicting evidence, making decisions with asymmetric downside, or interpreting ambiguous internal policy are in a slower-burning situation. The models still confidently hallucinate under those conditions because they are not penalised for guessing, and the non-deterministic nature of prompts makes the error pattern unpredictable. The transformation is real, but it is a long grind of S-curves, not a sudden singularity. The breathing room exists because the technology hasn’t smoothed out the rough edges that matter most for autonomous high-stakes work.

If your employer is deploying AI without fallback plans, your leverage lies in knowing exactly where it breaks

The Meta internal video was a flare, not the fire. The fire is the assumption that you can replace the people who built the intelligence with a system that still guesses on the hard parts. The employee who can walk into a review and say, “Here are the 30 failure cases from this month, here is what they cost, and here is the validation layer that catches them,” stops being a cost to be cut and becomes a hedge against reputational and operational damage. That is not a morale argument. It is a hard business case that the hallucination benchmarks already make. The models are not going away, but neither is the need for someone who knows exactly where they will be wrong.

The Real Roots of AI Evil Perception — and What the Tech Crowd Overlooks