Ben Affleck, who founded an AI film startup that Netflix bought for about $587 million, set off a wave of online surprise this week after clips showed the actor and director explaining artificial intelligence like an engineer. In a GQ interview that spread widely on October 6, he walked through convolutional neural networks, transformers, tensors and GPUs.
The main clip came from GQ's "One More Question" series, with interviewer Zach Baron. Affleck said he can write "pretty shitty Python scripts." He described how older convolutional neural networks were "the precursors to what the transformer can do," which runs much more computation at once. He explained how such a network reads a tensor, a grid of numbers for an image, to do "edge detection or feature extraction" and find, say, a window ledge so a green screen can be swapped out. Baron joked that he appreciated being spoken to "like I understand any of it."
Affleck went further at the Bloomberg Screentime conference, where he told Bloomberg's Lucas Shaw, "I learned Python." He described taking open video models and retraining them by unfreezing their weights, then training a final "cinematic" layer so a general model learns what film looks like.
The fluency has roots. Before his studio Artists Equity, Affleck ran a small visual-effects company and worked often with GPUs. In 2022 he quietly founded the startup, InterPositive, which Netflix bought for about $587 million. The company disclosed that figure in a July securities filing for a deal completed in March. The 16-person team joined Netflix, and Affleck became a senior adviser.
The tool does not make movies from text prompts. It builds a model from a production's own daily footage, then helps in post-production with tasks like relighting, color and visual effects. Affleck said he was uncomfortable that many video models had trained on work by people he knows. So his team shot its own footage for about eight months to create a clean dataset. That material became the layer that teaches the model cinema, and each production keeps the model trained on its own shots. He used the technique in his crime thriller Animals, which Netflix releases on October 9.
Beyond the facts, reactions split along a familiar line. Investor Michael Burry, a noted skeptic of AI hype, praised the explanation on X as "familiar and on point." Others were sharper. Online commenters noted a gap in the ethics argument: the open models Affleck fine-tunes were themselves trained on other artists' work, so his clean dataset sits on top of data he did not vet. A Substack analysis argued his wider take on consumer AI tools was thin.
What makes the moment worth watching is not Affleck's coding but that a major star explained machine learning accurately to a mass audience, at a time when Hollywood's AI debate is usually framed as fear. If that framing shifts toward craft and consent, studios and unions will still have to settle the harder question Affleck's own method raises: whose data trained the model underneath.
