What Happened
On September 18th, 2024, Lionsgate, one of Hollywood’s smaller entertainment producers and operator of the Starz television network, inked a data sharing deal with the generative AI video platform Runway.
According to Michael Burns, Vice Chairman of Lionsgate Studio, Runway’s new model, when trained on Liongsate data ,will allow the studio to save “millions and millions of dollars” on things like pre-production storyboarding and post-production special effects work. Lionsgate has a film and television library of more than 20,000 titles, including the “Hunger Games” franchise, “Knives Out,” and “Mad Men.”
Runway, in return, gets access to Lionsgate’s entire digital corpus of films, which it will use to train a new custom model based on those assets. Just imagine a prompt like “create a 5-minute trailer for a movie about…” and a few minutes later, a whole movie story arc is created and pitched. Or testing out dozens of different AI-generated background settings for a sequence already shot against a green screen.
The use of genAI tools for final production assets is already here. For example, the Oscar winner “Everything Everywhere all at Once” used Runaway’s tools to execute some of its dazzling and surreal special effects shots. But this partnership marks the first major collaboration between Runway and a Hollywood studio to access a complete library of content.
Runway, based in New York City, has raised $237 million in venture capital from investors like Google, Nvidia, and Salesforce Ventures.
This deal is about Lionsgate using generative AI for its own internal purposes, rather than letting you generate your own sequel to the “Hunger Games” or “John Wick” franchises.
Why It Happened
To live up to their lofty valuations, every generative AI startup needs big bucks enterprise deals — rather than just selling $12-a-month subscriptions to consumers. Note that this deal is about Lionsgate using generative AI for its own internal purposes, rather than letting you generate your own sequel to the “Hunger Games” or “John Wick” franchises.
And given generative AI platforms’ questionable track record of acquiring their training data on a legal basis, signed agreements like these may provide them with some legal cover, or at least the impression of trying to be an upstanding citizen in the creative community.
Runway earlier signed a data access agreement with stock photo rights holder Getty Images last December. The motive in that case, too, was to inoculate Runway (and Runway’s enterprise customers) from legal troubles downstream by licensing the images used to train their model. Unlike the Lionsgate deal, that one allows you, and other end users, to benefit from that content as part of your generative AI output.
On the other side of the table, publicly-traded Lionsgate has production costs to control, and an image to manage. It got a black eye this August for allegedly using a chatbot to generate faux critics’ reviews for its forthcoming Francis Ford Copolla directed film, “Megalopolis.”
Bottom line: both Lionsgate and Runway want to center and prioritize ethical and legal uses of the new generative technology, gain efficiency from it, and ideally avoid lawsuits, bad press, or disputes with creative talent and unions. We’ll see how that goes…
Why not leverage everything you’ve created in the past to help accelerate the process of creating something new — even if you still will require humans to evaluate and improve it?
What Happens Next
Two threads are playing out here:
First, we’ll see more of these “custom-trained AI platform” deals happen in all kids of creative industries, from consumer product design to movies to publishing. Why not leverage everything you’ve created in the past to help accelerate the process of creating something new — even if you still will require humans to evaluate and improve it? Many companies already use “internal ChatGPT” chatbots that leverage their data, such as customer proposals, to create new versions.
The second thread is platforms like Runway or ChatGPT licensing copyrighted content to use as part of their generative AI offerings for all users. For example, in publishing in the past few months, Wiley announced that it had already earned $44 million in new revenue from its AI licensing, and back in April, OpenAI announced it had done a deal with the Financial Times. The underlying driver for the AI model creators is satisfying their voracious appetite for training data, especially copyright-cleared data that is free of legal entanglements.
Lastly, it’s safe to predict that individual artists and creators won’t be happy with any of this — especially if some of the “custom-trained AI platforms” begin using generative AI to crank out new product — say, YouTube shorts based on a hit movie franchise. Even with the resolution of the Screen Actors Guild and Writers Guild of America strikes in 2023, and new regulations such as those signed this month by Gov. Gavin Newsom of California, intended to protect actors’ digital likenesses, it’s clear that disputes over the ways that generative AI can be used in creative industries have only just begun.
