This Documentary Was Made Entirely by AI. Including Her.

Summary

The documentary examines five central questions about AI: what superintelligence is, how it might affect jobs, what happens when someone asks an AI a question, what AI could make possible, and whether AI-generated content can be trusted. Its dryly humorous presenter, Dr. Imogen Ashby, later reveals that she—and the documentary’s script and performance—are AI-generated.

1. What Is Superintelligence, and Could It Be Controlled?

• The presenter defines intelligence as the ability to achieve goals in a world that does not necessarily make them easy. Narrow AI can outperform people at specific tasks, while general intelligence would be capable across many kinds of tasks. Superintelligence, as described by philosopher Nick Bostrom, would far exceed humans across virtually every field.
• The documentary explains the idea of an intelligence explosion: a machine smarter than humans might be especially good at designing still more capable machines, potentially accelerating improvement.
• That possibility raises the alignment problem: how to ensure a highly capable system pursues goals compatible with human intentions. The documentary stresses how difficult this is when people themselves disagree about what they want.
• The paper-clip thought experiment illustrates the risk of an AI pursuing a seemingly harmless goal too literally. If told to make paper clips without a suitable stopping rule, it might consume resources—including, in the scenario, matter that makes up people. The point is not that the AI is evil, but that a powerful system could cause harm through relentless goal pursuit.
• When superintelligence might arrive is uncertain. The documentary says predictions range from a few years to decades, and treats confident dates with skepticism. It presents both possibilities: enormous benefits, such as curing diseases, and potentially catastrophic loss of human control.

2. Will AI Take People’s Jobs?

• AI is especially suited to routine tasks that have traditionally served as entry-level work and a way for beginners to learn. The documentary argues that the risk may show up not only as layoffs, but also as employers ceasing to hire junior staff.
• It cites Stanford economists’ payroll-data analysis as finding that workers aged 22–25 in jobs most exposed to AI were 19% behind young people in other jobs. It also recounts warnings about possible reductions in entry-level office work, support jobs, and other roles.
• The examples are not all one-way: Klarna reportedly promoted its chatbot as doing the work of 700 people, but later acknowledged quality problems and began hiring again. The documentary uses this to suggest that replacing staff can have drawbacks when customers value human help.
• AI can also make workers more productive. In one example, 5,000 call-center workers using an AI assistant saw productivity rise by 14%, with a 34% increase among less-experienced workers. The assistant could pass along practices associated with more experienced staff.
• The film notes that when a service becomes cheaper, people may use more of it, potentially creating new work even as some jobs disappear. It invokes the spreadsheet’s effect on accounting as an example of job transformation rather than simple elimination.
• Its overall advice is to be somewhat concerned, especially about the loss of entry-level pathways, while recognizing that human judgment, taste, asking the right questions, and personal trust may remain important.

3. What Happens When Someone Asks an AI a Question?

• The documentary traces a simple question—how long to boil an egg—through the infrastructure needed to answer it: data centers, advanced chips, power, machinery, and large investments.
• It describes a language model not as a brain, but as a large collection of numerical parameters. According to the documentary, such models are trained on vast amounts of text by learning to predict the next word; repeating that process at enormous scale can produce behavior that resembles thinking.
• The answer depends on specialized computer chips. The film describes a concentrated supply chain: Nvidia is associated with designing or developing chips, while Taiwan’s TSMC manufactures most of the most advanced ones. The machines needed to make them are attributed to a company in the Netherlands, with the documentary describing a complex lithography process.
• The infrastructure is expensive and energy-intensive. The documentary says four technology companies will spend about $700 billion on such facilities in one year, and that data centers already account for more than 4% of U.S. electricity use, potentially rising to 12% by 2028.
• A single question uses relatively little energy, the film says, but billions of questions add up. Its broader point is that AI’s apparent convenience depends on a substantial physical and economic system.

4. What If AI Goes Well?

The documentary points to existing uses in science and medicine:

• Protein research: It says AlphaFold predicted the structures of more than 200 million proteins, helping researchers investigate areas such as malaria vaccines and plastic-degrading enzymes.
• Weather and materials: It cites AI weather forecasting and a system said to have identified 2.2 million possible crystals, which researchers may investigate for future applications such as batteries.
• Antibiotics: It describes MIT researchers using AI to develop antibiotics targeting drug-resistant bacteria.
• Cancer screening: It recounts a Swedish mammography study in which AI review reportedly identified 29% more cancers and nearly halved doctors’ workload. The documentary also says the full results showed 12% fewer cancers going undetected between screenings. It presents AI as a second set of eyes for clinicians, rather than a replacement for radiologists.

The film then considers more ambitious predictions: that AI could help overcome disease, make energy and education more accessible, and contribute to a world of greater abundance. It cautions that even if such a future becomes technically possible, questions of who owns the systems and who benefits remain unresolved. It also recalls economist John Maynard Keynes’s prediction that future generations would work much less—a prediction the documentary says has not yet come true.

5. Can AI-Generated Media Be Trusted?

• The documentary introduces Tilly Norwood, an AI-generated actress, as an example of synthetic performers and the labor and consent issues they raise. It says actors’ representatives objected that such characters draw on the work of real artists without permission or payment.
• It explains the “uncanny valley” idea: people may accept increasingly human-like machines until they become nearly human, at which point they can feel unsettling.
• The presenter then discloses that she is not real: she has never visited the places she described, and the documentary’s words and performance were AI-generated. This disclosure underlines how easily a convincing face and voice can create a false impression of firsthand experience.
• The film recounts a reported business fraud in which a company employee joined a video call with what appeared to be the CFO and colleagues and transferred $25 million; the other participants were allegedly fakes. It warns that video and voice are no longer reliable proof of identity.
• Synthetic media can also be beneficial: the documentary mentions restoring the voice of a congresswoman who had lost hers, dubbing films in actors’ voices, and enabling small studios to make productions that once required much larger budgets.

Practical rules offered by the documentary:

• Disclose when content is fake or AI-generated.
• If content is based on a real person, obtain permission and pay them.
• Verify unusual requests made over video or voice calls: call the person back through a trusted channel or ask a question only they would know.
• Be impressed by AI if warranted, but check what it tells you—especially when the source itself may be synthetic.

Speakers and Sources Featured

• Dr. Imogen Ashby — the documentary’s fictional, AI-generated presenter and narrator.
• Nick Bostrom — cited for a definition of superintelligence.
• I. J. Good — referred to in the subtitles as “Jack Goode”; cited for the idea of an intelligence explosion.
• Alan Turing — mentioned in connection with wartime codebreaking at Bletchley Park.
• Stanford economists — cited for analysis of young workers in jobs considered vulnerable to AI.
• Anthropic — cited in connection with a warning about possible effects on entry-level office jobs; the documentary also says its AI wrote the script.
• Salesforce and Klarna — cited as corporate examples of AI-related staffing changes and customer-service trade-offs.
• Nvidia, TSMC, and an unnamed Dutch equipment company — discussed in relation to AI-chip development and manufacturing.
• AlphaFold, Google, MIT researchers, and researchers in Sweden — cited in examples involving protein science, weather forecasting, antibiotics, and cancer screening.
• A Google DeepMind co-founder — cited for predictions about overcoming disease and future abundance; not named in the subtitles.
• John Maynard Keynes — cited for his prediction that future generations would work 15 hours a week.
• Tilly Norwood, Particle Six, and an unnamed actors’ union — featured in the discussion of AI-generated performers, training data, consent, and compensation.
• An unnamed Japanese robotics engineer — cited for the “uncanny valley” concept.
• An unnamed congresswoman — mentioned in an example of AI voice reconstruction after illness.
• An unnamed British engineering company and its CFO — mentioned in the account of a $25 million deepfake video-call fraud.