Since the launch of ChatGPT thrust artificial intelligence (AI) squarely into the public consciousness, the anxiety and excitement about it within the film sector has risen to feverish levels.
Many in the industry already feel immensely threatened by AI. The Screen Actors Guild (SAG-AFTRA) and Writers Guild of America (WGA) strikes in Hollywood — ongoing at the time of writing — have been partly caused by the fear of how generative AI tools could potentially cost actors and writers their jobs.
‘Deepfake’ technology is becoming ever more sophisticated, and actors can no longer control their voices or images. AI tools have already cloned the voice of Harry Potter star Emma Watson and used it to read Adolf Hitler’s Mein Kampf, while in the animation industry there was consternation this summer when it emerged the credits for Disney+’s Marvel series Secret Invasion had been designed not by humans but by AI.
Independent filmmakers across Europe have expressed concern that the streaming platforms’ AI-driven recommendation algorithms are pushing their movies to the margins. At the same time, others have welcomed the potential time- and labour-saving opportunities, with the British Board of Film Classification confirming it is testing AI for content classification — something the organisation’s president Natasha Kaplinsky described as “an incredibly exciting transition for the BBFC”.
European AI companies such as Switzerland-based Largo.ai and Danish outfit Publikum are both reporting a surge of interest in their tools from European filmmakers and financiers, while longer established companies such as Los Angeles-based Cinelytic, founded in 2015 and which has US studios such as Sony and Warner Bros among its clients, has seen enhanced interest in its predictive analytics and project management tools. (Cinelytic recently boasted its predictive box-office analysis tools, which it claims have been used for greenlighting decisions, have been tracking “at 96.3% accuracy for 2023”.)
“ChatGPT has changed the conversation,” says Rikke Flodin, a partner at Publikum. “People are curious now about how they can use AI.”
“The last six months is different than before,” echoes Sami Arpa, CEO and co-founder at Largo, of changing industry attitudes. “We always had interest but there was a big section of people who were sceptical about the use of AI.”
Anxiety around AI is prompting policy action. The UK government is formulating plans to regulate it while the European Union is preparing its own AI act, which it bills as “the world’s first comprehensive AI law”. The final form of the law is expected to be agreed by the end of this year. As governments look to bring in regulations, European film funds have also been scrambling to assess both the opportunities and the threats posed by AI. Copyright, IP, unconscious bias and employment protection are among the subjects they are investigating.
Koen Van Bockstal, CEO of Flanders Audiovisual Fund (VAF), has devised an AI questionnaire that all producers applying to the fund will now be obliged to complete. The questionnaire demands to know whether those applying to the VAF’s film, media or games funds used AI, and, if so, how they went about it — including the deployment of graphic AI tools such as Canva or Galileo AI, text programmes such as ChatGPT or Largo.ai, or whether they have scraped internet data. Van Bockstal insists there is no agenda behind the questions beyond gathering information at this stage.
“I don’t want us to fall into the trap of having a moral opinion or a prejudice before we know what we are talking about,” Van Bockstal explains of the questionnaire, which he is lobbying for other European funders including the British Film Institute (BFI) to adopt. “When we have sufficient information, then we will be able to come up with rules and regulations.”
One issue flagged by Van Bockstal (although often overlooked in discussions on the topic) is environmental — the vast amount of energy that AI-generated tools consume. The VAF CEO claims that “AI in the next five years will consume as much energy as we need for the entire world; that is a challenge in terms of climate change”.
Blunt instrument
One of the paradoxes about the increasingly heated debate around AI within the European film industry and beyond is that relatively few filmmakers and financiers are using it in a significant way, especially when it comes to production. “[AI] is still a fairly blunt instrument in its current form,” TV writer Lisa Holdsworth, chair of the Writers’ Guild of Great Britain, recently told Sky News.
“A lot of film is execution dependent — a piece of AI can’t really figure that out,” says Phil Hunt, managing director of prolific London-based production and financing company Head Gear Films, on why his company has actively avoided AI so far. Hunt does not believe, for instance, that an AI tool would have spotted the potential in recent low-budget Australian horror hit Talk To Me, which is sold by Head Gear’s sister company Bankside Films and has been a summer hit in the US following a buzzy Sundance premiere in January. “That is something a piece of AI would never have seen, ever,” he insists.
Hunt’s perspective is reflective of scepticism among independent film practitioners — and at least partly endorsed by Hollywood filmmaker James Cameron, who dealt with AI nearly 40 years ago in his 1984 film The Terminator. In a recent interview with CTV News, the director of Titanic and the Avatar franchise acknowledged that advances in AI technology pose a threat to humanity, but he also expressed extreme scepticism about the current creative abilities of AI programmes.
“I just don’t personally believe that a disembodied mind that’s just regurgitating what other embodied minds have said — about the life that they’ve had, about love, about lying, about fear, about mortality — I don’t believe that’s something that’s going to move an audience,” said Cameron.
Largo boss Arpa counters that AI can effectively identify audiences and markets based on script and cast. It can help independent producers to package their films, budget them at the right level and make them attractive to funders, sales agents and distributors. The company has content- and genre-mapping tools, casting tools, financial forecasting tools and even ways of using AI to measure the potential emotional impact of projects.
“Analysis of the content, predicting which actor will fit for which character or which type of audience will be interested in the content and what will be the financial success of the content,” says Arpa, citing the benefits of AI soothsaying for market purveyors. “The chance to find investment for the film is increasing significantly — we can tell from the feedback we are getting.”
Arpa declines to mention titles on which the firm has worked because it “creates conflicts”. However, the UK’s Pinball Films, France’s Cinema Defacto and Belgium’s Versus Production are among companies known to have tested Largo tools, albeit with varying results. Sometimes, it seems, directors do not even realise their producers have been turning to AI. “We work with the producer and then the director is calling us saying, ‘I didn’t work with you, why did you mention my film?’”
One insight Arpa offers is that for an indie project, spending half the budget to secure an A-list actor is not necessarily money well spent. “So many of our assumptions for the film industry are not correct anymore — and some of them were not correct from the start,” he says. “Many independent producers try to bring A-list actors to their film, fully convinced that is how they can achieve success. But when we look at the past data, we cannot find a strong correlation between the popularity of the actor and the success of the movies.”
Industry suspicion
Largo provoked controversy this summer when US actress/producer Charisma Carpenter expressed indignation the company was trying to use AI casting tools to usurp the role of traditional casting directors. Arpa denied this, issuing a statement that the company backed the SAG-AFTRA and WGA strikes and wanted to “support creators”, not undermine them. Nonetheless, the spat highlighted the suspicion with which many filmmakers and actors regard AI.
For its part, Largo’s rival AI company Publikum, which has been active since 2021, combines AI analysis with anthropological insights. “AI is the tool that will give us the large overview, and the anthropology is where we go into depth and figure out what is the emotional connection between the story you are writing and the feelings you want to trigger in your audience,” says Flodin. She notes that Publikum’s AI tool is not being used to make artistic decisions on behalf of filmmakers but to give them additional insights. “It is completely up to the producers and the producers’ team to decide which of the insights they would like to use.”
Publikum has worked with the Danish Film Institute, Norwegian Film Institute and Netherlands Film Fund among other organisations on projects they have supported. Its insights have been used in several auteur-driven European arthouse films, among them Kristoffer Borgli’s Sick Of Myself, which screened in Cannes’ Un Certain Regard in 2022, and Ena Sendijarevic’s Sweet Dreams, produced by Dutch outfit Lemming, which just had its world premiere in Locarno.
“We focused on the different ways people in Europe and North America talk about colonialism and power relations,” producer Erik Glijnis explains on Lemming’s use of Publikum’s AI in support of the international launch of Sweet Dreams. “We used the AI to get an overview of the conversations and sentiments, and used these results during the development process and the positioning of the film.”
UK filmmaker Ben Mole consulted Publikum on the development of his sci-fi drama Save Point, about a gamer trying to rescue her father on a distant planet. Mole recently won the Future of Film Incubator prize and was put in touch with Publikum as a result. “They worked with us on the script and isolated half-a-dozen thematic plot and character [points] — and then their AI bots mine the internet to find what people are talking about surrounding the themes,” Mole explains. “The other thing they do is present focus-group questions to individuals on Zoom and then their answers are fed into a different AI package. It’s not just the transcript of answers that the AI looks at, it’s the pauses, inflections, facial [tics] and the way people answer.”
“It’s almost like [Publikum’s] role is to get the questions right and then the AI goes off and does the work,” adds Mole, who tweaked his script as a result.
Publikum charges around $19,700 (€18,000) for its analysis. “It’s worth it if you can afford it,” Mole suggests of what might be a hefty price for a low-budget indie picture.
However, many others remain sceptical. In 2020, the European Producers Club (EPC), the Paris-based network of leading European producers, partnered with Largo. The idea was that EPC members could trial its forecasting tools on their new projects. These seek to predict, for example, how particular actors are received in territories where the producers are hoping to sell the film and at what budget level a movie is likely to be successful.
“When I was first studying what AI can bring to the sector, I was looking at that aspect and I thought it might be useful to use it for the forecast and the analysis of the scripts,” says EPC managing director Alexandra Lebret. “[But] it was very difficult to convince our members to use it even if I was telling them Largo is just doing what a sales agent is doing — giving you an estimation based on the script and the cast. When you have that, you can go and see a financier and be in a stronger position.”
While a couple of EPC producers embraced the Largo tools, most were dismissive. “My experience wasn’t great,” says a UK producer who took part in the experiment. “To me, the key thing is that we need to know what is the set of data that the AI has been trained with. The set of data that Largo has trained its AI [with] to me felt obsolete and felt like a set of data created by people who were not producers.”
The data, the producers complained, was out of date. They found their projects being compared against US movies and TV shows from many years before. “The results we were getting were completely ridiculous,” the producer remembers. “We ran a project in development through it… the granularity of the information wasn’t there. To me, it felt like a waste of time.”
“It needed to have more European shows and films to be relevant,” Lebret agrees of these producers’ reservations. She points out, however, that Largo’s tools have “improved a lot” since then and EPC is planning a new trial with the company. “I believe it can be a good choice for the producers and they should embrace AI,” she says.
Feeding the data
For Lebret, one of the biggest current concerns surrounding AI is how algorithms are being used by streamers to guide viewers. “AI creates something which is not neutral anymore. That’s the case still for the recommendation system of Netflix and the streaming services. That’s something that has to be on the table [in regulations]. For me, the highest risk in AI is that recommendation system.”
The issue here is cultural diversity and the worry that European producers’ films are not being promoted properly on the major streamer platforms. “There is always a human behind the algorithms. It’s never [just] the audience choice,” Lebret notes.
AI companies argue they can use their own systems to counter bias and oversights. “AI, if it’s used correctly, can bring great advantages in terms of diversity,” insists Arpa. “AI is biased in a manner that reflects the bias in society, but AI systems can be used to correct these biases in a very fast way. It’s just a matter of how you feed the data.”
Another vexed issue is the lack of transparency among AI providers and their potential infringement of copyright as they scrape data from multiple online sources. There are sharply divided opinions about what is and what isn’t permissible. “Under EU law, AI cannot constitute as an author, so AI-produced content does not have an author. The data input which AI uses to come up with its content is called TDM, text and data mining. This is not illegal, because it’s not seen as copying work but as taking inspiration from it,” says Lebret.
There are European directives through which creators can prevent their work being used for TDM — but it is a grey area. The media has been awash with recent stories about novelists, authors and artists complaining their work is being cannibalised by AI. Two authors, Mona Awad and Paul Tremblay, recently started a lawsuit against OpenAI, the company behind ChatGPT, claiming their work was being used to “train” ChatGPT and their copyright was being breached as a result.
“If you say from tomorrow that AI can’t learn from copyrighted materials, Google cannot work,” Arpa says, warning of potential consequences if restrictions on AI are too draconian.
AI has yet to take root in the UK and European indie film sectors, for all the current alarm about the threats it poses. For many, the old-fashioned ways remain the best. “The best way of making decisions is talking to the market itself,” says Head Gear’s Hunt. “If you speak to sales companies, you get a very clear sense of what’s working in the market and what’s not.”
Predictably, the AI companies have a very different perspective. They believe the AI revolution is already underway and it is time for the indie sector to embrace the tools available.
No comments yet