Documentaries
Ten films that show the field rather than explain it
Documentary is the format AI is worst served by and best explained through. Worst, because a subject with no visible mechanism resists the camera — there is nothing to film but faces and screens. Best, because the films that work solve that problem by finding a person to follow, and a person is what most explanations of AI are missing. The list below is ordered by how much each film adds to a reader who already knows the outline, not by acclaim.
Number 1: The AI Doc: Or How I Became an Apocaloptimist
2026 · dir. Daniel Roher · Sundance
Wikipedia — The AI Doc: Or How I Became an Apocaloptimist
The only major documentary filmed after ChatGPT, with the people who built it.
Roher gets extended access to Sam Altman, Demis Hassabis and Dario Amodei — the three people whose companies define the frontier — and lets them make their own cases at length. That access is the film's value and its problem: critics have noted how much uncontested airtime the CEOs receive. Watch it as primary source material rather than as an assessment, and it becomes the most useful record of what the people in charge said they believed in 2026.
Number 2: AlphaGo
2017 · 90 min
A sports film that happens to be about a neural network.
DeepMind's match against Lee Sedol had everything a documentary needs and AI usually lacks: a champion, a challenger, a crowd, and a result nobody could argue with. The film's real subject is Lee Sedol's face across five games — and Move 37, the moment a machine played something no human would have and commentators had to decide whether it was a blunder or genius. The most human film on this list, and the one to show someone who thinks they are not interested in AI.
Free on YouTube
Number 3: Coded Bias
2020 · dir. Shalini Kantayya
One researcher discovers facial recognition cannot see her, and follows it all the way to Congress.
Joy Buolamwini found that commercial face-recognition systems failed on darker-skinned faces — including her own, which some systems detected only when she wore a white mask. Kantayya turns that finding into a structure: one discovery, one researcher, one escalating fight. It remains the clearest documentary account of why a system that is accurate on average can still be unusable for the people it misclassifies.
Free on PBS in the US
Number 4: iHuman
2019 · Norwegian · dir. Tonje Hessen Schei
AI as an arms race between states and corporations, not a consumer product.
Where most films treat AI as technology, iHuman treats it as power — surveillance infrastructure, autonomous weapons, and the competition between China, the US and Europe to own both. It is the least reassuring film here and the necessary counterweight to founder-led narratives. If AlphaGo shows what the technology can do, iHuman asks who is holding it.
Number 5: The Social Dilemma
2020 · Netflix
Wikipedia — The Social Dilemma
The recommendation engine as the first AI to reshape a society.
Ex-Google and ex-Facebook insiders describe how optimising for engagement produced outcomes none of them intended. The dramatised segments are heavy-handed and the film is better on diagnosis than on remedy — but it documents the one deployed AI system that has already altered mass behaviour, which makes it more concrete than most speculation about future harms.
On this siteGoogleNumber 6: Do You Trust This Computer?
2018 · dir. Chris Paine
Wikipedia — Do You Trust This Computer?
An existential-risk film made before the argument went mainstream.
Released when AGI discussion was still a fringe position, it now reads as a record of how that argument sounded before the frontier labs adopted it. Some of it has aged into relevance and some into overstatement; both are instructive. Useful mainly for seeing which of 2018's predictions survived contact with the 2020s.
Number 7: The Thinking Game
2024 · dir. Greg Kohs
Hassabis and DeepMind, from game-playing to protein structure.
By the same director as AlphaGo, and effectively its sequel: the road from beating humans at Go to solving a fifty-year-old problem in biology with AlphaFold. It is the strongest available case for the scientific rather than commercial argument for AI, and the best film for anyone who suspects the whole field is chatbots.
Free on YouTube
Number 8: Lo and Behold, Reveries of the Connected World
2016 · dir. Werner Herzog
Wikipedia — Lo and Behold, Reveries of the Connected World
Herzog asks the internet whether it dreams of itself.
Idiosyncratic, philosophical and structured as ten loosely connected chapters rather than an argument. It is not a primer and will frustrate anyone looking for one, but Herzog asks questions no technologist would think to ask, and the interviews land differently for it. For the already-interested.
Number 9: The Great Hack
2019 · Netflix
Cambridge Analytica, and data as a political instrument.
Adjacent to AI rather than about it — the technology at issue is psychographic targeting, not machine learning per se. It earns its place because the surveillance-and-persuasion thread running through Coded Bias and The Social Dilemma is unintelligible without the Cambridge Analytica case as its reference point.
Number 10: The Age of A.I.
2019 · Series · 8 episodes · hosted by Robert Downey Jr.
A broad, optimistic survey with very little friction in it.
The widest coverage on the list and the shallowest: each episode tours an application area without seriously engaging the objections. Included because breadth has value as an entry point, and because a reader should know what the well-produced optimistic version looks like before deciding it is the whole picture.
Free on YouTube