Ben Affleck says he got private demos of Google and OpenAI’s AI video models by simply calling up and asking
The actor and filmmaker has revealed that he got private looks at Google and OpenAI ‘s AI video models by calling the companies directly and asking to see what they were building. The Oscar winner also revealed that he writes Python, understands how neural networks read images and has spent years working with GPUs.
He picks up the phone and says his name. Sometimes people say yes, to my astonishment,” Affleck said on GQ’s One More Question series. The tip came from visual-effects artists he knew, who told him both companies were doing interesting work with transformers in video. Affleck said he has been into computers since he was young, and the interest grew as filmmaking moved from analog film to digital.
Ben Affleck has a simple way of getting into the world’s top AI labs. “I’ve learned that I can actually just call up and go, ‘Hey, it’s Ben Affleck. Can I come see what you’re doing? Machine learning has been part of the visual-effects workflow for years, which is how he started coding. He is modest about the results, calling them “pretty s***ty Python scripts. His grasp of the tech runs deeper than that. He explained how convolutional neural networks turn an image into a tensor, a set of numbers covering the batch, the frame and the red, green and blue values of every pixel.
After seeing realistic video generated with a few keystrokes, he called Damon and told him they were finished, and that they should make as many films as possible while they still could. “You’re never just going to type your movie out, that’s just never going to happen,” he said. Speaking at an event in Los Angeles last weekend, Affleck said most AI models in use had been trained on the work of his peers, which made him uncomfortable.
The network then uses edge detection and feature extraction to work out details like where a window ledge sits in a shot, making it easier to remove a green-screen background and replace it. He picked up much of this at a small visual-effects company he ran before co-founding Artists Equity with Matt Damon . The early demos rattled him. The panic lasted “a day or two. Then he sat down with engineering teams and went through the data. His takeaway was that the tools were brilliant but built by people who “had no understanding of the domain expertise of filmmaking. AI video generators, in his view, still lack the consistency and control a director needs. The gap he spotted between the engineers and the film set became the starting point for InterPositive, his AI company for post-production. So he raised money and spent eight months shooting his own footage to train the tools.
Netflix bought the company earlier this year in a deal reported at $587 million (roughly Rs 5,000 crore), and Affleck joined the streamer as a senior advisor. He has already put the tools to work on Animals, his kidnapping thriller with Kerry Washington that releases on Netflix on October 9. Where exactly, he won’t say. He has also signed on to the Creators Coalition on AI, which is pushing for guardrails to protect creative work.
Pointing it out, he said, would be like a magician explaining a trick when filmmakers are trying to create the illusion of reality. “Anybody who’s told, ‘Hey, this is going to replace you in your livelihood,’ is f***ing concerned,” he said, adding that he is baffled anyone finds that surprising. InterPositive starts with open models built on publicly available footage and trains them for cinematic tasks. Filmmakers can then train the tools further on their own productions, and that footage stays theirs. A few phone calls have given Affleck a closer look at AI video than most of Hollywood gets, and he has come away less worried than many of his peers, though not dismissive of their fear. You use AI every day. Now get your AI Quotient. Take the AIQ test.