AI Is a Human Product, Not a Magic Force. A New Book Wants You to Know the Difference.

Key points

• Eleanor Drage's book argues that AI is built on human labour, from mining raw materials to labelling training data, not on any mysterious intelligence.
• Drage says big tech companies can't deliver utopian promises while prioritising profit over public interest.
• The cultural obsession with AI-caused extinction, she argues, distracts from practical fixes that could make AI safer right now.
• Demystifying marketing language, such as "the cloud" and "hallucination", is a concrete way citizens can push back against tech company narratives.

The word "artificial intelligence" carries a lot of weight. It conjures something vast, almost supernatural. Eleanor Drage, an AI ethicist at the University of Cambridge, wants to dissolve that impression entirely.

Her new book, reviewed by The Guardian, argues that AI is not a mystical force. It's the product of workers in silicon mines, microchip factories, and data labelling centres, where people are paid to tag images and text so that machine learning systems, the software that learns patterns from enormous datasets, can be trained. Understanding that human chain, Drage says, is how citizens begin to reclaim power from the companies building these systems.

What does the jargon actually mean?

Drage's plainest move is attacking the language itself. "The cloud" sounds like a natural phenomenon. It's just someone else's computer. An AI "hallucination", the term used when a chatbot confidently states something false, is really a system error or a data labelling mistake. Calling it a hallucination implies the machine has a mind that occasionally wanders. It doesn't.

"Intelligence" is the biggest sleight of hand. The word is broad, loaded, and, Drage argues, potentially misleading when applied to software that predicts the next word in a sentence.

Should people worry about AI wiping out humanity?

Drage is sceptical that apocalyptic framing helps ordinary people at all. Silicon Valley's dominant story, that AI will either deliver a golden age or destroy civilisation, is compelling but convenient. It pulls attention toward existential drama and away from practical questions: who controls the data, how should AI be regulated, what rights do workers and users have.

Apocalypse talk also creates a false hierarchy of problems. Climate change and economic inequality already cause enormous harm to real people today. Fixing those problems would reduce many AI-linked risks too.

The book's sharpest argument is simple. Big tech can't honestly promise utopia while it's chasing profit and serving a narrow group of shareholders.

The pattern is one we've tracked before. Our story on Anthropic's push for stricter AI laws from 16 July 2026 showed how companies can dress up commercial interests as public safety, the exact tension Drage is writing about.

What can an ordinary person actually do?

Drage's practical advice centres on informed scepticism. Question the marketing terms. Ask who benefits when a company calls its product "intelligent" or its mistakes "hallucinations". Demand clearer rules about how personal data is collected and used to train AI models.

None of that requires a computer science degree. Critical reading applies here just as it does to any industry making large promises.

Common questions

What is a data labelling centre?

A workplace where people are paid to review and tag content, such as marking which images contain a cat or flagging harmful text, so that AI systems have organised information to learn from.

Why does the language tech companies use matter?

Words like "hallucination" or "intelligence" shape how the public thinks about accountability. When a chatbot "hallucinates", it sounds like a glitch beyond anyone's control. When it makes a labelling error, someone is responsible for fixing it.