Everyone Is Already Using AI (And Hiding It)

by

In an evening on the Eastside of Los Angeles, around 200 people assembled inside an old soundstage, now repurposed, to commemorate the launch of Asteria Film Co., a fresh AI studio among the almost hundred currently active in Hollywood. Founded by entrepreneur Bryn Mooser and actress-writer Natasha Lyonne, Asteria is the latest venture in a lineage of creative pioneers who have revolutionized filmmaking throughout history. At 45, with a chiseled jaw and salt-and-pepper beard, Mooser embodies the rumpled yet refined pitchman persona, as evidenced by Cartier’s advertorial titled “The Entrepreneur.”

Guiding me through the studio, Mooser showcased its 25,000 square feet of soundstages and workshops that were initially constructed in 1916 by Mack Sennett, a groundbreaking filmmaker who innovated using scenic backdrops. In the lobby, he pointed to an architectural model encased in glass, representing the studio’s original appearance without a roof, as it was before electricity became common. Mooser viewed Asteria as a continuation of this legacy of creative trailblazers, reminding us that figures like Walt Disney and George Lucas were indeed technological visionaries. “The tale of Hollywood,” he said, “is the story of technology.

In a secluded area of the studio, we encountered Lyonne amidst a row of antiquated televisions screening peculiar clips produced by Asteria’s AI system. The content was disquieting, featuring robots with expressionless faces typing aimlessly in vintage offices and mannequin heads floating in space. At that moment, Lyonne was sipping on sugar-free Red Bull, sporting a chic black velvet jacket with a deep v-neck, having recently returned from Seattle where she spoke with science-fiction author Ted Chiang about the limitations of AI in creating superior art. In Hollywood over the last few years, it has become evident to Lyonne that many individuals are not entirely honest regarding their use of technology. “When I’m directing an episode,” she explained, “I like to delve deep into details and specifics. And you often discover that they’re labeling it as machine learning or something else, but in reality, it’s AI.” She had started conducting her own investigations, immersing herself in the works of Oxford scholar Brian Christian and philosopher Nick Bostrom, who warns about the potential risks AI poses to humanity’s future. However, she came to believe that it was no longer feasible to “put the genie back in the bottle.” “It’s wiser to get your hands dirty than pretend it’s not happening,” she said.

The Hollywood Issue

The Great Realignment

See All

 

It appeared that most people at the gathering shared a common viewpoint. Many acquaintances and close confidants of Lyonne were present: Janicza Bravo, Clea DuVall, and Tessa Thompson among them. As I moved through groups of studio executives in formal attire and filmmakers in casual clothing, I overheard the term “inevitable” being used five times. A man wearing a cowboy hat, who identified himself as a creative director, commented, “It’s going to happen regardless of our feelings.” When I sought Tim Heidecker’s opinion on the matter, he said his thoughts were too complex for a party discussion. I came across an agent from CAA sitting in a nearly empty row of antique theater seats. At CAA, she explained, discussions about AI are frequent. In 2023, the agency had initiated a project, called the CAA Vault, to preserve and control the rights to their clients’ likenesses. “Everyone‘s using it,” the agent said, “they just don’t discuss it openly.” Close by, a former Netflix executive and current independent producer enjoyed a glass of red wine. She chose not to inform some of her colleagues about attending the party that evening. A director she collaborates with recently warned her that if anyone at her company used AI in any capacity, he would resign. “If they catch me expressing interest,” she said, “they’re going to be furious.

Recently, Hollywood has been grappling with challenges that some experts refer to as an “existential predicament,” a collapse, and a “downward spiral.” Studios have been producing fewer films, and viewership of these productions has decreased. Layoffs have become commonplace. The impact of AI on this situation is debatable; while it poses potential risks for the industry’s demise, it also offers possible solutions. AI tools can now create various elements of a film, such as scripts, footage, soundtracks, and actors. However, there are concerns about these AI models since they were trained on copyrighted material obtained without permission or payment.

The writers and actors unions ended their strikes in 2023, resulting in new contracts that included safeguards against AI misuse. These contracts prohibit studios from using scripts written by AI and from digitally replicating actors without explicit consent and compensation. However, they allow for certain uses, particularly in the realm of generative video: Studios can use models to create synthetic performers and other types of footage (including visual effects and entire scenes) as long as a human is nominally in charge, and unions have an opportunity to negotiate terms.

Despite this flexibility, the industry faces a thicket of legal uncertainty, particularly regarding how these systems were initially trained. Over 35 copyright-related lawsuits have been filed against AI companies so far, the outcomes of which could determine whether generative AI has a future in Hollywood. As one producer stated, “The primary concern throughout Hollywood is that you’ll make a blockbuster, and then you’ll spend the next 30 years in court.

In spite of lingering apprehensions, major film production companies are pressing on with their plans (though few have publicly announced their strategies). Meanwhile, artist-led studios designed particularly for filmmakers have gained traction in the industry. For instance, Asteria debuted its model this year and boasts a unique selling proposition. Lyonne and Mooser claim that Asteria’s model was trained exclusively on licensed content, positioning it as the first “ethical” studio. Prior to Asteria, Runway — a tech and media start-up — had already made a mark in Hollywood, with its models being among the most frequently employed. Moreover, Runway was the pioneer in striking public partnerships with film studios. These innovative studios are refining advanced techniques for integrating generative video (capable of producing an image or video upon a prompt and potentially revolutionizing movie production).

This spring, Darren Aronofsky announced a partnership with Google’s DeepMind. James Cameron teamed up with Stability AI, one of the tech companies making inroads in Hollywood. “In the New Year, many people in the studios woke up and said, ‘Okay, in 2025 we need to make a difference,’” said Prem Akkaraju, the CEO of Stability AI. “Because production is down, profitability is down, attendance at theaters is down. It’s harder to make a movie today than it ever has been.” Cameron put it more bluntly on a tech podcast recently: If audiences want more blockbusters, he said, “we’ve got to figure out how to cut the cost of that in half.” Erik Weaver, a producer and technologist who regularly talks with studios on how to use AI, had observed a “radical shift” over the past few months. A few weeks ago, in a meeting with a major studio, an executive told him that “almost every single one of our productions is coming to us and saying, ‘How can I use AI to get back on budget?’” Weaver added, “These filmmakers need $30 million to make their movie and they have about $15 million. They only have so much money, and they’re getting desperate.”

In 2018, the idea of Generative AI was primarily confined to academic and tech circles, but Cristóbal Valenzuela, a 28-year-old self-taught programmer and artist who had recently graduated from an unconventional art-and-tech program at NYU, saw potential. For his thesis project, he developed a toolkit that enabled artists to generate images, text, and video using AI. By typing a sentence into an app, he could create brief movies. The results were far from perfect and didn’t have immediate commercial value, but the concept struck a chord with the artistic and design community. It wasn’t long before Runway started making inroads in Hollywood, not with studio executives, but with those working behind the scenes. Visual effects artists were among the earliest to experiment with these tools. Traditional visual effects techniques are known for being painstakingly slow, so the allure of quick, generative shortcuts was instant.

By summer 2023, the performance of Runway’s model noticeably escalated, coinciding with the surge of interest in generative AI in the general public. Among other capabilities, Runway was capable of simulating lifelike explosions and crafting enchanting backdrops within minutes. Later that year, Valenzuela traveled to Los Angeles for a discussion with Michael Burns, Lionsgate’s vice-chairman, who oversees productions like The Hunger Games, Twilight Saga, and John Wick series. Previously, Burns had interacted with major tech firms, but they lacked filmmaking knowledge, as Burns put it, “They are strangers in this land.” However, Valenzuela was developing tools tailored for filmmakers. As Burns pondered, he envisioned that with Runway, filmmakers could “produce movies and television shows we wouldn’t have made otherwise. We might not be able to afford it for $100 million, but we would make it for $50 million due to AI.” He proposed to Valenzuela, “Perhaps we should acquire you.” To which Valenzuela responded, “Oh, no, no, we’ll acquire you.” True to his words, by this spring, Runway’s value surpassed Lionsgate’s market capitalization, reaching $3 billion.

Lionsgate decided to share some of their film archive with Runway for the purpose of training a unique model on their exclusive collection of films. During my visit to Burns in his office with a view of Santa Monica’s residential area, he was still astounded by the rapid pace of evolution. “The speed at which it’s changing is mind-boggling,” he remarked. “I’ll discuss with Cristóbal what my team requires next, and he replies, ‘We’ll have that ready next week.'” Earlier in the day, Burns had requested Valenzuela to create a trailer for a film they hadn’t yet filmed. The idea was to present it at a film festival, but first they would sell the film using AI-generated scenes based on the script beforehand. Although the production cost could be as low as $10 million, the final product would resemble a $100 million movie. “We’re going to make explosions to make it look more grandiose and cinematic,” he said.

The partnership with Runway is adaptable and open for exploration. Burns explained, “We’re brainstorming the realm of possibilities.” In essence, they want to experiment and see what works best. Currently, they are contemplating various strategies, such as repackaging and reselling Lionsgate’s extensive library content. This could involve altering the tone, format, or rating to create a more child-friendly version, or even transforming a live-action film into an animated one. He used an example from Lionsgate’s action franchise collection: “Now we can suggest, ‘Make it anime and make it PG-13.’ Within three hours, I could have the movie.” However, he noted that additional costs would be incurred for actors and other rights holders, but he is confident that they can manage this, allowing them to resell the content afterwards.

He provided another illustration by stating, “We’re currently debating whether to produce a movie, and here’s an instance: there’s a scene involving 10,000 soldiers on a snowy hillside with plenty of horses. Filming this in the Himalayas would take three days and cost millions, but using Runway, we can replicate it for only $10,000. Despite doubts about whether the movie will be made, the financial calculations are favorable.

As a passionate cinephile, I’ve noticed that generative-AI models like Runway are seldom seen in the limelight of mainstream films and TV shows. However, in the genre-defying masterpiece “Everything Everywhere All at Once,” a VFX artist cleverly utilized Runway’s green-screen tool to swiftly eliminate unwanted background elements from scenes. Furthermore, Runway played a significant role in Amazon’s biblical epic “House of David,” contributing to numerous shots, including the mesmerizing 90-second fantasy sequence that graces the sixth episode’s opening. The executive producer Jon Erwin shared with me that such stunning visuals would have been beyond their reach without Runway, given budget and time constraints. When I inquired about other films using Runway, Valenzuela, who seems to be deeply involved with every major Hollywood studio, was somewhat tight-lipped, merely saying, “I’m afraid I can’t disclose that.

A studio executive shared with me that many studios are hesitant to openly discuss their use of AI due to fears related to unions. Duncan Crabtree-Ireland, SAG-AFTRA’s chief negotiator, noted a cautious approach taken by studios following the strikes. The union successfully negotiated a requirement for regular meetings with each major studio regarding the use of generative AI every six months. Crabtree-Ireland mentioned that he has noticed studios making moves beyond the contractual demands, establishing internal ethics committees and legal reviews. For now, it appears that AI is being used primarily to improve editing, sound quality, and visual consistency rather than replacing performers. However, Crabtree-Ireland hinted that this cautious approach might only be a temporary strategy while studios wait for the public’s concern over AI to subside and for laws to catch up.

The numerous dialogues with workers across various ranks hinted at a distinct narrative, one of secret experimentation and the ability to disclaim responsibility. Roma Murphy, a writer and co-chair of the Animation Guild’s AI committee, had heard about rogue employees at studios – lower-level staff under tight deadlines – requesting colleagues to utilize AI without formal approval. An anonymous animator shared an example where a costume designer used AI to create concept images, which were then given to an illustrator to rework, effectively washing the AI’s involvement. As the animator put it, “They’ll essentially launder the AI-generated content through an artist.” Reid Southen, a concept artist and illustrator who has worked on films like The Hunger Games and The Matrix Resurrections, conducted an informal survey among professional artists, asking if they had been asked to use AI as a reference or to enhance their completed work. Almost half of the 800 respondents admitted to having done so, including Southen himself. “Work opportunities have dwindled,” Southen shared with me. Southen, who has worked in film for 17 years, stated that his income had dropped by nearly half over the past two years – more than it did during the early stages of the pandemic when the entire industry halted. It’s becoming common for producers to exclude artists from the process entirely. “I can confirm,” one producer said, “that some producers are developing shows and they need art to present their ideas.” Typically, they would hire an artist to create the art; now, they simply input prompts. As a studio executive put it, “If you’re a storyboard artist, your job is gone. That’s over. Because the director can tell AI, ‘Here’s the script. Storyboard this for me. Now change the angle and give me another storyboard.’ Within an hour, you’ve got a dozen different versions of it.” However, he added, if that same artist learned to use generative-AI tools effectively, “he’d have a promising job opportunity.

Even a skilled artist would question the quality, but we’re all struggling to meet deadlines, stay within budget. In desperate situations, it can be tempting – even if the outcome might be questionable in terms of quality and ethics.

The reason studios are taking their time to fully incorporate certain AI technologies into their workflows is because the technology currently available isn’t advanced enough. During a visit to Joel Kuwahara, co-founder of Bento Box (the studio behind shows like Bob’s Burgers), in March, it was evident that not all Hollywood animators are convinced about the role of AI in the creative process. However, Kuwahara, a technology enthusiast who has been in the animation industry for over 30 years, sees potential in AI as a productivity tool. He was part of the team that digitized The Simpsons in the mid-’90s and believes that animation, as an industry, could benefit from AI. However, Kuwahara admitted that most of the AI tools he has tested so far have failed to deliver on their promises. For instance, a recent AI storyboard generator he tried was disappointing, as it produced a new house instead of simply adjusting the camera angle by ten degrees. Despite decades of automation in the industry, creating a storyboard for a half-hour animated show on network television still takes four to six weeks. Kuwahara explained that while some people may suggest shortening this time to three weeks, it’s not feasible because a board artist needs time to think and consider each scene, and rushing the process would result in a subpar board.

In an age dominated by abundant Netflix content and numerous Marvel spin-offs, there’s a debate about whether quality continues to hold significance for studio executives or viewers. I had a conversation with a VFX artist and technology consultant who advises studios on integrating AI into their production processes. He preferred to remain unidentified due to the confidential nature of his work, stating “I’m bound by numerous non-disclosure agreements.” Recently, he mentioned that when a director requested minor effects such as a wisp of smoke or a flash of fire, he would generate it using AI technology. While the results weren’t as visually stunning as they would have been with conventional VFX software like Houdini or Maya, he believed that most viewers wouldn’t perceive the difference. “There is indeed a loss of quality,” he admitted. “But that loss is only apparent to those who truly appreciate it, much like the subtle nuances in fine wine.

Following discussions with studios and other AI firms during conferences, the VFX artist had reached a realization: Occasionally, they were marketing potential future capabilities instead of completed products – assurances that in a short period (a few months or years), the technology could produce complex action scenes or entire performances. Lately, some of these meetings have left him feeling uneasy. “I’ve heard executives say, ‘Can we feed an AI a season’s worth of scripts and it can produce a show in a day?’ And I can’t help but wonder, Is that really your goal? While the people below them, who are also part of these discussions, leave those meetings feeling like they need to wash off.

In 2023, several artists filed lawsuits against Runway and numerous other AI-based art generation companies, with more similar cases emerging. These companies have primarily defended their actions by claiming that using copyrighted content during model training falls under fair use, as long as the generated output does not replicate distinct features of the original work. During our discussions, Valenzuela compared individual works or images to grains of sand on a beach – too insignificant and numerous to significantly impact the outcome. When I raised the example of a user instructing an AI to create a video in the style of Wes Anderson, he emphasized that the ethical dilemma lies with the enthusiast, not Runway. “You can still achieve that without AI systems – produce a film, color-grade it in the style of Wes Anderson, and market it accordingly,” he explained. “However, the one who might face consequences won’t be the camera, editing software, or computer. It’ll be you.

As a passionate film enthusiast, I’ve always been troubled by the idea of companies exploiting artists’ work for profit. Just like Haley has expressed her wish for a month-long film festival at Film Forum during her funeral, I too feel strongly about ethical practices in the movie industry. When Haley and I first discussed starting an AI movie studio about a year and a half ago, it was me who proposed we aim for something ethical. “Why can’t we create a clean model?” I remember asking Haley. “That would address most of our issues,” I added.

As a film enthusiast, I’ve always been fascinated by the realm of technology and its potential, particularly in the field of artificial intelligence (AI). For years, it has been asserted that creating a generative AI model demands internet scraping due to the enormous data required to teach these systems how to depict our world – visually, linguistically, emotionally. I’ve always wondered if, with enough funding and talent, it could be possible to bypass this limitation.

Having established several tech companies before, I’m no stranger to convincing influential figures to join my ventures. My mentors include industry giants like Bob Iger and Elon Musk, with whom I maintain a somewhat complex relationship. With AI as my next project, I began reaching out to studio heads, agents, and executives in 2023, proposing an ambitious idea: the “Pixar of AI.”

This vision encompassed a unique collaboration between technologists and filmmakers, with a strong emphasis on animation. The goal was to push the boundaries of what AI could achieve, merging cutting-edge technology with storytelling prowess to create something truly extraordinary. Conversations with industry veterans like Iger and former Disney chairman Jeffrey Katzenberg were instrumental in shaping this vision. It’s an exciting journey, and I can’t wait to see where it leads!

During a discussion with Vinod Khosla, a billionaire venture capitalist who has been involved in a lengthy dispute with California over access to his private section of the Pacific coastline, he casually mentioned that Steven Spielberg is one of his close friends. He shared with Mooser that Spielberg frequently queries him about how artificial intelligence (AI) will impact the entertainment industry. “I don’t have the answer,” Khosla said, “but I think you might. So if I were to invest in you, I could introduce you to him, and you could enlighten him on the subject.

Khosla made the choice to put money into the project. Similarly, Hemant Taneja, CEO of General Catalyst and a globally recognized advocate for AI, also chose to invest. He viewed Asteria as not just a public relations endeavor for artificial intelligence, but also as a means to alter society’s perception of it. As he put it, “If you aim to revolutionize an industry, you must captivate hearts and minds in society. Telling compelling stories is essential to this.”

For years, the general portrayal of AI has been dystopian – think killer robots and science fiction catastrophes. Taneja believes that such narratives need to be reshaped: “We’re aiming to demonstrate the tangible benefits of AI. It’s not just a potential risk; it’s an opportunity. AI has the power to make everyone extraordinary.”

He appreciated the simple concept of Asteria because he felt it would help establish trust among those he needed to persuade.

Taneja brought together Mooser and the Moonvalley team, a collective of ex-Google DeepMind engineers, to construct the model they named Marey. This group discovered a method for training models using minimal data directly from filmmakers, libraries, archivists, and emerging AI data brokers. As Mooser explained, “Our timing was perfect; if we had attempted it six months earlier, I’m not sure it would have been feasible. It was still quite costly.” However, the specifics regarding how Moonvalley funded and obtained a large enough dataset remained unclear, as their representative declined to disclose such information, citing confidentiality reasons.

Previously known as Moonvalley, Asteria’s current corporate parent has managed to amass over $100 million in funding thus far. Impressive teams within the visual effects industry have been gathered by this company, according to insiders. Paul Trillo, a director recognized for his innovative and experimental work, has come on board as a strategic partner and is instrumental in shaping Asteria’s unique hybrid-animation pipeline that combines traditional filmmaking methods with artificial intelligence. Benjamin Lock, the VFX supervisor for ‘Rogue One: A Star Wars Story’ and ‘Ready Player One’, now leads Asteria’s visual effects endeavors. An anonymous VFX artist who has watched Asteria’s filmmaking process remarked that it is not simply a matter of pressing buttons and waiting. Instead, they stated, “They’re actually doing, like, 20 things to get the image out.” The artist added, “When you see a Paul Trillo piece, it looks effortless, but I’ve seen the steps. It’s not easy.

In a contemporary music video, there’s a sneak peek at how Asteria is merging traditional hand-drawn artistry and AI technology. For this venture, they partnered with L.A. artist Paul Flores who created 60 unique images by hand. These were employed to develop a custom AI model in Flores’ distinctive style, enabling the generation of countless more images. Subsequently, a 3D creation tool was utilized to construct a digital city, dubbed Cuco Town, which facilitated smoother camera movements and repeated settings. A team of twenty individuals eventually transformed Flores’ sketches into a vibrant, multi-layered animated short with an impressionistic, painterly aesthetic.

Similar to Runway, Asteria is holding talks with almost every significant studio and has collaborated with several of them, though none of these partnerships have been made public yet. They also have a variety of projects in the works, such as an adult animated series called The Odd Birds Show. Currently, Lyonne is working with Asteria on a sci-fi film she’s developing alongside filmmaker Brit Marling and futurist Jaron Lanier, who is known for his work in virtual reality. The spark for Lyonne’s interest in futurism was ignited during the production of Russian Doll , her critically acclaimed Netflix series about a woman who keeps dying and reliving the same day. She explained, “Russian Doll III in a way is what I’m doing here.” She was tight-lipped about the film’s screenplay, only revealing that she envisioned it as half ’90s indie and half “Chutes and Ladders” open-world gaming. She hasn’t utilized AI during production yet but intends to use Marey to create the visuals for the movie’s second half. However, her attempts to utilize it for screenwriting were unsuccessful as the results fell short of her expectations. “I have not found a way at all to effectively use it in writing,” she said. “I long for it to work.

As a passionate cinephile, I find myself drawn to the realm of AI, not for its speed or vast potential, but for its promise of independence. I’m not aiming to be a tech mogul; instead, I’m a filmmaker yearning to resist the dominance of tech giants over the future of our cherished medium. I envision a future where independent filmmakers can harness AI tools to reclaim authorship from studios, bypassing the constraints that come with the broken funding system.

In essence, we need a modern-day manifesto, akin to the Dogme 95 movement of the ’90s that sought to free cinema from technology’s grasp. If we can embrace this artist-first approach before it becomes the industry norm, that’s something I would be eager to contribute to. It’s almost as if we are determined not to fade into oblivion without a fight.

However, not everyone shares my optimism about democratization. The fear is that powerful tools in the hands of even more powerful companies could lead to an endless stream of low-cost content. But I remain hopeful and ready to navigate this exciting, challenging terrain.

Recently, Natasha had a conversation with David Lynch, a renowned filmmaker from the past who was an early adopter of digital cameras. As it happened, they were neighbors before his passing. Last year, she asked him about AI. Picking up a pencil, he said, “Natasha, this is a pencil.” Just as everyone has access to a pencil, so too will everyone with a smartphone be using AI, whether they realize it or not. He explained, “It’s not just having the pencil; it’s how you use it.” Do you understand?

Read More

2025-06-04 14:56