A.I. Revolution | Full Documentary | NOVA | PBS
Can we harness the power of artificial intelligence to solve the world’s most challenging problems without creating an uncontrollable force that ultimately destroys us? That is the question at the heart of this NOVA documentary, which premiered on PBS on March 27, 2024. ChatGPT and other new AI tools can now answer complex questions, write essays and generate realistic-looking images in a matter of seconds. They can even pass a lawyer’s bar exam. Should we celebrate? Or worry? Or both?
The film begins with how the technology works. Modern AI is built on artificial neural networks, systems that learn by training on vast amounts of data rather than by following rules written by a programmer. Large language models like ChatGPT generate answers by recognising patterns in enormous quantities of text and guessing the next word, one word at a time.
The most moving part of the film is in medicine. Computer scientist Regina Barzilay at MIT began working on AI in the 1990s, when rule-based systems like Deep Blue were giving way to neural networks. She used the techniques to decipher dead languages. Then, in 2014, she was diagnosed with breast cancer, and she began asking a different question: why could her cancer not have been diagnosed earlier? Working with radiologist Constance Lehman at Massachusetts General Hospital, she trained a neural network on more than 128,000 mammograms collected at seven sites in four countries. The resulting software, called Mirai, is between 75 and 84 percent accurate at predicting whether a patient will develop breast cancer within five years. When a friend developed lung cancer, the same approach was applied to CT scans, producing a model called Sybil that recognises the patterns of developing lung cancer.
Then the film turns to the other side. The same technology that can spot a tumour years early can also create deepfakes: convincing fake videos and voices of real people, made with tools that are now available to almost anyone. The film shows a deepfake of Jordan Peele and a fake image of a terrorist attack that spread online and moved the stock market. Researcher Hany Farid explains that the technology lowers the barrier to creating fakes while increasing the potential for misuse, and the film follows the argument to its conclusion: if anyone can make a convincing fake of anyone, then trust itself becomes a casualty. The film also notes the quieter uses of AI for good that rarely make headlines, including systems deployed in California to detect wildfires early, before they rage out of control.
What the Film Explores
The film explores the promise of AI through the people building it. Barzilay’s story is the film’s centre, because it shows what the technology is for: a researcher who survived cancer and used the tools of her trade to make sure the next patient is diagnosed earlier. The prosthetics work makes the same point in a different register: a machine that learns to read a person’s muscles is not replacing the person, it is giving them back a hand.
It also explores the peril, and it does so without hype. The danger the film describes is not a robot uprising. It is something more ordinary and more immediate: the erosion of trust. A deepfake does not need to fool everyone. It needs to fool enough people, for long enough, to do damage — to a reputation, to a market, to an election.
Finally, the film explores the question of control. The researchers it interviews — including Mustafa Suleyman and Yoshua Bengio — disagree about how dangerous advanced AI could become, but they agree on one thing: the technology is moving faster than the rules that govern it. The film ends not with an answer but with a choice, the same one it opened with. AI can help solve the world’s hardest problems. Whether it does, the film suggests, depends less on the technology than on the people who decide how to use it.
Key Themes
- The question: can AI solve the world’s hardest problems without becoming an uncontrollable force? The film never stops asking it.
- How it works: neural networks learn from data, and large language models answer by predicting the next word — abilities that arrived years ahead of schedule.
- Cancer, caught early: Regina Barzilay’s Mirai model predicts breast cancer within five years with 75 to 84 percent accuracy; the Sybil model does the same for lung cancer from CT scans.
- Prosthetics that learn: Coapt’s pattern-recognition system reads muscle signals and gives an artificial arm more natural control.
- Deepfakes and trust: convincing fakes of real people are now easy to make, and the film shows how quickly they can do damage.
- AI for good, quietly: from wildfire detection in California to drug discovery, the film shows uses of AI that rarely make headlines.
Production and Background
The film is a NOVA production from GBH, the award-winning PBS science series, and it premiered on March 27, 2024. NOVA has been explaining science to a general audience for half a century, and this film follows its established method: a correspondent, Miles O’Brien, travelling to meet the researchers, with the science explained as it is encountered rather than lectured from a studio.
The film arrived at a moment when AI had suddenly become a public argument rather than a specialist subject. ChatGPT had been released a little over a year earlier, and the questions the film asks — about deepfakes, about jobs, about control — had become questions that ordinary people were asking too.
Recommended Companion Viewing
Explore authoritative investigative documentaries on related historical events and investigative subjects.
OpenAI is Suddenly in Trouble
YouTube Is Already 20% AI Slop
How The Internet's Favorite AI Employee Went Rogue
Yikes
Frequently Asked Questions
What is A.I. Revolution about? It is a NOVA documentary, presented by Miles O’Brien, that explores the promise and perils of new AI technologies: how they work, what they can do in medicine and prosthetics, and the dangers of deepfakes and misinformation.
Who is Regina Barzilay, and what did she build? Barzilay is a computer scientist at MIT who was diagnosed with breast cancer in 2014. With radiologist Constance Lehman, she built Mirai, a model trained on more than 128,000 mammograms that predicts breast cancer within five years with 75 to 84 percent accuracy. A related model, Sybil, detects developing lung cancer from CT scans.
Can AI really pass a lawyer’s bar exam? Yes. As the film notes, tools like ChatGPT can now answer complex questions, write essays, generate realistic images in seconds, and pass a lawyer’s bar exam.
What is the danger of deepfakes? Deepfakes are convincing fake videos and voices of real people, made with tools that are now widely available. The film shows how a fake image of a terrorist attack spread online and moved the stock market, and argues that the real casualty is trust itself.
More to watch
- AlphaFold – The Most Useful Thing AI Has Ever Done
- AI Supremacy: The Artificial Intelligence Battle Between China, USA and Europe
- Will Artificial Intelligence Save Us Or Kill Us?
- How AI causes serious environmental problems (but might also provide solutions)
- AI and the Future of Health
The film’s most hopeful example is AI in medicine, and AlphaFold – The Most Useful Thing AI Has Ever Done shows another AI system transforming biology.







