The Latest AI Documentary Asks: Just How Scared Should We Be?

It’s not easy to get an interview with Sam Altman—just ask Adam Bhala Lough, the filmmaker behind the recent documentary Deepfaking Sam Altman.

Lough originally planned a feature exploring the potential and perils of AI that would center on a conversation with the OpenAI CEO. But, after having his inquiries ignored for months, he opted instead to commission a chatbot that mimicked Altman’s speech patterns and approximated his facial expressions by way of a digital avatar.

The real Altman did sit down, however, for the new feature The AI Doc: Or How I Became an Apocaloptimist, which hits theaters March 27. So did Dario Amodei, the CEO of Anthropic, and Demis Hassabis, a cofounder and CEO of Google’s DeepMind Technologies. (Though the filmmakers say they requested interviews with Meta’s Mark Zuckerberg and X’s Elon Musk, neither made an appearance.)

It’s an impressive level of access for codirector and documentary protagonist Daniel Roher, whose 2022 documentary Navalny, about the Russian opposition leader Alexei Navalny, won an Academy Award. The problem is that once they’re on camera, Altman et al. say little we haven’t heard before—and they skate by on glib answers concerning their responsibilities to the rest of their species. When Roher asks Altman why anyone should trust him to guide the rapid acceleration of AI, given its extreme ramifications, Altman replies: “You shouldn’t.” The line of interrogation ends there.

The AI Doc is framed by Roher’s anxiety over the impending arrival of his son and first child with his wife, filmmaker Caroline Lindy. He wonders what kind of a world his boy will inherit and whether the rise of artificial intelligence will preclude the experiences that develop us into self-sufficient adults. In Roher’s first several interviews, all his worst fears seem to be confirmed. Tristan Harris, cofounder of the nonprofit Center for Humane Technology, delivers one of the worst gut punches: “I know people who work on AI risk who don’t expect their children to make it to high school,” he says, invoking a scenario in which the technology demolishes the very infrastructure of traditional education.

Despite the sense of mounting panic, Roher and codirector Charlie Tyrell present an admirably robust crash course in AI and the biggest questions it poses, helped along by Roher’s insistence on defining terms in plain language rather than startup buzzwords. Visually, the film is charmingly human, featuring colorful drawings and paintings by Roher, while whimsical stop-motion sequences hint at the influence of producer Daniel Kwan, the Oscar-winning codirector of Everything Everywhere All at Once. The vibrant creativity amid portents of doom provides some of the hope that Roher is desperately seeking.

Yet later interviews with Silicon Valley techno-optimists promising AI that conquers diseases and climate change—followed by the CEOs striking their usual balance between hype and the tones of sober caution—pass without much interrogation of grandiose claims. There is barely a moment spent considering why or how we should expect the current crop of fallible large language models to give rise to the mythical “artificial general intelligence” (AGI) that would outstrip human cognition. There are, at best, euphemistic acknowledgements (from venture capitalist Reid Hoffman, for example) that any benefits will come along with unspecified harms.

Even when the top players say that the near-term implications of AI are as significant as the advent of nuclear armament, they are defaulting to a familiar playbook, presenting their products as singularly consequential one way or another—hinting that only they can be trusted to advance them.

The documentary accurately conveys how the unregulated AI gold rush is driven by the perverse incentives of a global market and a struggle for domination. It observes how this mania concentrates wealth and power in the smallest possible circle of elites. Strange, then, that The AI Doc eventually carves out a gotta-hear-both-sides position in which the general public, not the executives under the microscope, are tasked with steering the AI revolution in the right direction. It’s stranger still considering that Roher has produced lacerating critiques of the AI economy on the press circuit, blasting it as a “Ponzi scheme.”

As he prepares to be a father, Roher has a touching conversation with his own dad, who advises him that while there are historical forces he can’t control, he’ll be a great parent no matter what—and that every generation has dealt with the existential distress of bringing life into an era of instability.

Nevertheless, Roher and Tyrell call viewers to action, concluding the film by suggesting that ordinary citizens can pressure governments and corporations to ensure that AI evolves along the safest, narrowest path toward prosperity for all. The sequence is set to footage of other grand projects, including the construction of the Golden Gate Bridge, as though this piece of architecture were shaped by collective opinion.

After a screening of The AI Doc at Los Angeles’ Academy Museum on Monday, Tyrell, Kwan, Harris, and producer Ted Tremper held a brief Q&A, with each reinforcing the idea that the feature was a productive step toward raising awareness of AI as a critical issue. “We're excited to continue this conversation,” Kwan said at one point. “This is just the beginning, and I know that this movie will never be able to encompass everything.” But he foresaw that the film would encourage audiences to “link arms with us and step confidently into the darkness as we try to figure out what we do together.”

The documentary’s vision of positive change, though, is hazy, perhaps clouded by both the necessity of a rosy ending for Roher’s expanded family and the delicate suspension of skepticism whenever a billionaire enters the frame.

In this narrative, these executives are apparently just along for the ride like anybody else, their status a mere accident of fate—which sets them up for a modest shrug whenever they admit they don’t totally understand what goes on inside the AI models they have already deployed at scale. As long as we’re so preoccupied with whether these programs may soon possess consciousness or intent, we might want to treat these people as though they, at least, have agency.