The Walt Disney Company’s multi-million dollar artificial intelligence initiative was dealt a blow this summer when it lost a key leader.
The company fired Ben Stanbury, a former head of technology at Amazon Studios whose job as Vice President of AI and Machine Learning was figuring out how Disney could best use AI, with a focus on creating efficiencies in production and post-production, according to two people familiar with the matter.
Disney cited performance issues for his dismissal, but an individual close to the situation said Stanbury was put in an impossible situation. “The pressure to deliver on a short timeline, on a goal set that was unrealistic and with a tech that just isn’t ready” for the giant demands of Disney, set Stanbury up for failure, the individual said.
The incident marked a setback for Disney, which has poured $50 million into the AI initiative, according to a company insider.
But it also underscored the broader cultural challenges facing Hollywood and traditional media companies when it comes to AI. Even as other industries — from technology to publishing — proceed to incorporate AI to cut costs and replace jobs, media companies have bumped up against numerous roadblocks including actors reluctant to cooperate with AI models, animators and post-production experts pushing back on change, technological limitations and legal questions.
All these elements combine to slow down Disney — and others — at a time when AI companies are racing forward with little care for an “ethical” approach that considers details like copyright.
“Internally, it’s a cultural resistance, not just within Disney, but all of Hollywood,” said Raymond Wang, CEO and analyst at Constellation Research, which works with media companies.
Disney declined to comment for this article. Stanbury did not respond to a request for comment.
***
With the hype over AI exploding over the last three years, Disney set out to be first in class in innovating with the technology.
A year ago, the entertainment giant established the Office of Technology Enablement and tapped Jamie Voris, the chief technology officer of Walt Disney Studios, to run the unit. He and his team would focus on the potential use of technologies like AI and mixed reality to win efficiencies that would boost the bottom line.
The team also had a second mandate: figure out how to do this ethically and responsibly.
On the face of it, that priority made a lot of sense. The industry has increasingly embraced the ethical use of AI, or using models trained on licensed data and prioritizing the recognition of rights and IP, and protecting IP from AI exploitation has consistently been a focus for Disney.
But those high minded intentions became a way for the resistant parts of the company to deprioritize any AI efforts, preventing the team from making any tangible headway with the technology, according to Wang.
For Disney, the problem was twofold. From a cultural perspective, many in the acting and animation communities within the company oppose AI and have been vocal about the issue (“Black Widow” star Scarlett Johansson and director James Cameron are among many who have warned about its dangers), requiring the OTE group to tip-toe around their concerns, according to people familiar with the company’s AI work. With leadership not looking to ruffle feathers with talent, there’s been no champion to get buy-in on AI from the disparate parts of the Disney corporate empire.
Then there were the technical hurdles. For all the promises that generative AI offered, the team realized the technology wasn’t quite ready for prime time. After a year of work, Disney found the AI systems didn’t meet its standards to replace the pricier aspects of production, like visual effects, leaving CEO Bob Iger frustrated with the lack of progress, according to the insiders.
Disney isn’t the only company whose expectations for AI overshot reality. Lionsgate also had big ambitions with its partnership with AI startup Runway, but the reality has been a far more limited application of the technology.
“For some businesses, good enough is good enough,” said Chris Ross, an analyst at Gartner. “In the entertainment business, the bar is higher.”
It’s perhaps why when Iger was asked about AI’s future role in the company during last week’s earnings conference call, he struck a collaborative tone.
“We’ve been in some interesting conversations with some of the AI companies, and I would characterize some of them as quite productive conversations as well, seeking to not only protect the value of our IP and of our creative engines, but also to seek opportunities for us to use their technology to create more engagement with consumers,” he said.
An internal culture clash
The unrestrained use of AI to copy actors or recreate elements or characters from existing shows and movies is a nightmare scenario for Hollywood. It’s why Duncan Crabtree-Ireland, national executive director of SAG-AFTRA, estimates that as many as 15% of its 160,000-strong union “would prefer that AI not exist.” Nicolas Cage has described AI as a “dead end” for actors. Guillermo del Toro said he would “rather die” than use the technology in his films.
But in reality, that meant the OTE had to navigate the core creative class of Disney, which was vocally opposed to the technology. It proved to be too much of a challenge for Stanbury, who is now at a stealth startup focused on AI-driven content creation.
“It became more about what they won’t do vs. what they could do,” Wang said.
It became more about what they won’t do vs. what they could do.
– Raymond Wang, Constellation Research
It wasn’t just with actors and directors. One of the areas where AI can provide an immediate benefit is with animation, as an AI-generated cartoon is more palatable than an AI-generated human. But Disney animators were just as opposed to the technology as actors, the person said.
A few hours after Iger’s comments in which he teased the possibility of users generating short-form AI content on Disney+, “Owl House” creator and former Disney animator Dana Terrace tweeted, “Unsubscribe from Disney+. Pirate Owl House. I don’t care. F–k gen AI.”
That reaction is exactly why Iger has had to walk a fine line between pushing AI and maintaining relationships with key talent. But that meant the initiative lacked any leaders willing to make AI a priority.
The reality vs. perception gap
Cultural hurdles aside, Disney was beset by the classic problem of generative AI: an overly optimistic perception of what it could achieve vs. the reality that the technology is still limited in many ways.
“There’s a big expectation gap with AI in general,” Ross said.
Indeed, a September report issued by software tools maker Atlassian found that 96% of CEOs have failed to see any meaningful return on investment from the technology.
In order to achieve a competitive advantage, Iger pushed the OTE on a homegrown effort for AI video generation, according to a person familiar with the company’s plans. Disney’s thinking was that having its own model would give it a competitive advantage if other studios used publicly available options like Runway, Adobe’s Firefly or Moonvalley’s Marey, among others, the person said.
This came at a time when the industry was buying into the hype of AI and making big bets.
“A year ago, there was a lot of money being thrown into AI as an end-all-be-all solution,” said a VFX producer who wished not to be named because of his relationships with different media companies.
But as studios look to use gen AI to bolster VFX work on projects in the pipeline, it’s clear that the technology isn’t ready. The producer noted that AI struggles to create consistent imagery long enough to last the length of a movie, with even advanced models like Sora working most effectively for short TikTok-like clips. AI also struggles with delivering video at a high enough resolution or right high-dynamic range color format, among other technical requirements, the producer said.
“After all these millions of dollars spent, they’re finding out you can’t deliver what we’re used to delivering over the past two to three decades,” the producer said. “It just doesn’t meet the standards.”
In addition, working with AI models means generating imagery by using prompts, a less precise method than working with traditional CG imagery, where individual elements can be worked on.
As an example, Coca-Cola earlier this month released an AI-generated commercial and an accompanying video detailing the effort put in to make it work. But the spot still looked wonky, and elicited quick and vocal backlash.
A big part of the problem is the difficulty involved with training an effective AI model, particularly if you have little content to train it with. Another hurdle was Disney forbidding the team from using its own vast library of films and television shows to train the model, the person said, over concern about the potential legal liability.
That stands in contrast to Lionsgate’s deal with Runway, where the studio’s library of content was used to train a custom Runway model for the studio. But as we reported, that breadth of content wasn’t sufficient to create a model that could spit out a whole film.
Still room for AI
AI is already being utilized at the big studios, but in smaller, less flashy ways. Universal VP of Creative Technologies Annie Chang said at TheWrap’s TheGrill business conference that production executives are using AI to break down scripts and organize them into efficient shooting schedules, or using generative AI to create rough visual approximations of ideas and concepts.
Several executives have talked about the use of AI as a tool to speed up pre-viz, or the pre-visualization of scenes. Or it can be used to cheaply and effectively dub content into different languages. These are features that Disney is likely working on, if it hasn’t already implemented them.
The problem is none of these applications of gen AI are particularly sexy — or save a lot of money.
The big swing of AI actually creating scenes or generating eye-popping visual effects with a quick written prompt requires even smarter and more powerful models, and those working in the technology acknowledge that it’s not quite there yet.
That doesn’t mean those models and the techniques won’t get better and smarter over time as new innovations emerge that will properly harness the power of AI. But as Iger suggested with his comments, they probably won’t be coming out of Disney.
