Four years tackling a question Hollywood can’t dodge. Here’s how he did it.
Ben Affleck did not approach artificial intelligence like a wide-eyed futurist. He approached it more like a director walking onto a set at dawn – coffee in hand and scanning his surroundings for what could go wrong.
When he first watched generative AI in action, it appeared to him as equal parts promise and risk. As he later described it, the technology represents “a real opportunity and a real authentic danger” in an industry he’s dedicated his life to.
In Hollywood, AI-related risks continue to have real consequences for working people, with concerns narrowing in on the future of jobs, payroll, and whether a creator’s work might end up being used without permission.
The debate has been shaped by fears of replacement and cheap imitation, which Affleck says he understands because he has felt those fears too.
But instead of watching from the cheap seats, he decided to step onto the soundstage and into the machinery.
The moment he decided to build
If AI was going to seep into filmmaking, shouldn’t it be built where filmmaking actually lived? Under the same hard constraints that crews fought every day, like budgets, schedules, consent and credits?
Affleck began to think about the problem less like an actor reacting to a new technology, and more like a builder asking how it should work.
If AI was inevitable, perhaps the more important question was whether filmmakers themselves could shape the tool before it began reshaping the craft.
So in 2022, Affleck started building quietly and out of sight. The stealth startup – called InterPositive – was later revealed once it was acquired by Netflix in a deal that could be worth up to $600 million.
But the startup’s premise reflected that same filmmaker-first idea Affleck had from the start. If AI was going to enter Hollywood, it should be trained inside the realities of production, not scraped from the internet.
While it wasn’t chasing the headline fantasy of “type a prompt, get a movie,” Affleck had made it clear he had little patience for that idea.
“People mostly think of [AI] as making something from nothing […] that’s not what this is.”
Teaching a machine how filmmaking works
Instead, InterPositive started where film was actually made – on a controlled soundstage, with the muscle memory of a full production. Pulling together a small team of engineers, researchers, and creatives, the company filmed a proprietary dataset.
It included raw material intended to teach a model not only what images looked like, but how shots behaved inside a scene, sequence and story.
InterPositive’s model was built to relieve the bruising reality of production, like missing coverage, imperfect lighting, continuity headaches. Basically, the kinds of problems that rarely appeared in demos but routinely surfaced in edit bays late at night.
Reuters described the goal as training the AI to understand “visual logic” and “editorial consistency,” while keeping cinematic rules intact under real-world pressure.
Affleck also insisted the system be built with restrictions meant to protect creative intent, with guardrails designed to keep decisions in artists’ hands, rather than handing the wheel to an algorithm.
Who gets to shape what comes next for Hollywood
And after all that hard work, it’s no surprise Netflix wanted in. Because what InterPositive not only a tool, but a way for Hollywood to shape AI before AI reshaped Hollywood.
Netflix’s acquisition of InterPositive will see the the entire team brought inside the company, including Affleck as an advisor. While the financial terms of the deal were not disclosed, Netflix had framed the intent in very human terms.
New tools, the streaming giant argued, should expand creative freedom at the hands of writers, directors, actors and crews.
And in an industry that ran on authorship – the fragile, hard-won sense that a scene was made by people – Affleck is not asking whether AI is going to enter Hollywood. He’s asking who gets to write the rules when it does.
Editor’s Note: This article has been updated to reflect the latest financial figures relating to the deal.
