When it comes to making movies, AI has always been seen as a lazy shortcut for people who don't want to get their hands dirty or pay a crew to make something unique and special.
But still, the promise of just instantly getting what you want was too much for some to pass up, like Lionsgate, which entered a partnership with Runway AI last year with the hopes that AI could help them revolutionize their studio by feeding it their library.
Well, how has that been going?
According to a recent report from The Wrap, it’s not going well at all. In fact, Lionsgate has hit the same creative limitations and logistical headaches that humans hit when they make a movie, and then they also hit those very specific AI legal problems as well.
Let's dive in.
AI Needs More Than What Lionsgate Can Offer
The core of the deal was that Lionsgate would give Runway AI its entire film library to train an exclusive, custom AI model. The studio would then use this bespoke tool to generate new content based on its own IP.
It was an ambitious and great idea; to pay a lot less to just pump out content. Isn't that why we all got into the movie business?
Even though I found this to be foolish, I understood the financial reasons behind the decision. If you could actually just pump stuff out based on what you already had, it could be a huge money source for you.
But it turns out that was a promise that was too good to be true.
The entire Lionsgate library isn't nearly enough data to build their own AI model, and that's all Runway AI could legally license to use, so they got stuck.
Building a functional, high-quality generative video model requires a truly astronomical amount of information. Way more than in a movie studio's library.
And it's not just because Lionsgate is a small studio, the report we read said that even something the size of Disney would struggle to do this as well
The reason these LLMs succeed is that they have access to the entire internet to steal from. When you shrink it down, it has less to pull from and therefore cannot compute.
Think about it from a practical standpoint. If you want an AI to generate a specific lighting effect, it needs to have seen thousands upon thousands of examples of that effect in different contexts. Even though a movie studio may have thousands of movies, it pales in comparison to the millions, if not billions, of real-life footage or movie effects you'd find online.
A model trained on a limited dataset would inevitably have far more limited generative capabilities. And if it can't generate something real and accurate, it's completely worthless to a studio.
'John Wick 2' Credit: Lionsgate
The Unanswered Question: "Who Gets Paid?"
Even if the technology worked perfectly, Lionsgate immediately ran into a second, more complex wall: the law and the unions.
Let's say their own LLM worked. Well, if they just made a new John Wick movie, who would they pay for it?
See, they would have to pay the actors and "other rights participants." That's just the way these laws are written. And if you're using images, likenesses, or characters, people need to be cut checks.
- Do the screenwriters get a check for the story structure and dialogue the AI is mimicking?
- Does the director get compensated for the visual style and pacing being replicated?
- What about the gaffer whose lighting designs created the mood the AI is now generating on command? Or the production designer whose world-building is the foundation of the new visuals?
Simply owning the final IP doesn't erase the contributions of the hundreds of artists who created the source material you used to train the LLM, and there is no clear framework for how to pay them for AI-generated derivative works.
Lucky for craftsmen, unions and laws have them protected.
So, What Is It Good For?
Despite the setbacks, Lionsgate's official statement maintains a positive outlook. Peter Wilkes, the studio's Chief Communications Officer, told Gizmodo, “We view AI as an important tool for serving our filmmakers, and we have already successfully applied it to multiple film and television projects to enhance quality, increase efficiency, and create exciting new storytelling opportunities.”
So what are those "opportunities"?
It's hard to tell, but most likely it's just VFX in post and possibly some other enhancements in the edit. People have talked about AI making trailers, but we have yet to see it being used that way at the studio level.
AI will continue to advance, but for now, it's not moving as fast as promised to live up to the expectations.
Summing It Up
The narrative of AI "replacing" creatives was scary, but a lot of smoke and no fire, for now. It turns out we don't have the tech to keep up with expectations, and while things are advancing, this is going to take years to move forward.
For now, the dream of an automated content machine is just a dream.
Long live human creatives.
Let me know what you think in the comments.
