This AI startup says it can tell if a script would make a hit movie

When Quilty entered the market earlier this year, the AI ​​startup promised that its tool could accurately predict a movie’s success just by reading the script. When people actually got it An opportunity to experiment However, they remained skeptical about Quilty’s product. Even with all the data in the world, I predicted the scenario Christywhich would continue A flop at the box office, it will outperform the screenplay Sinnerswhich became an Oscar-winning film.

As many AI executives have offered before, Quilty’s founders believe it could help “democratize” their industry by giving up-and-coming creators access to helpful tools — a great Quilty score might be an engagement with a product, and a low score might be a sign of the need for more reviews. But for now, Quilty is little more than a mish-mash of pre-existing AI systems, and the company has yet to prove that its technology has the flair or analytical capabilities to determine future success (let alone proven success).

Founded by film producers Simon Horsman and Daniel WoodQuilty uses artificial intelligence to analyze scripts and create detailed reports on the project’s chances of success. After an unproduced script is fed, Quilty gives it a score ranging from 0 to 100 that reflects the potential project’s narrative quality, its commercial viability, whether it will resonate with audiences, and how much production it is likely to cost. The platform sells the idea that it can give users a glimpse into the future as they try to get their films/films greenlit. Horsman Wood believes Quilty is about to become an integral part of how traditional production studios operate.

When I recently sat down with Horsman and Wood, they were adamant about their desire to “keep humans in the loop” rather than automate the entire pre-production process. While first building their company, Horsman and Wood solicited feedback from a number of other innovators who often expressed concerns about AI’s potential to negatively impact jobs and leave human workers unskilled.

“We agree with a lot of the negative sentiment toward AI, but what we’re trying to do is empower human creativity,” Horsman told me. “Quilty is really about development and giving users — whether they’re a writer, producer, buyer, financier or studio executives — as much information as possible to make an informed greenlight decision.”

Instead of giving users access to a single dedicated AI model that provides feedback on scripts, Quilty combines a number of widely available AI tools to bring different types of analytics into the process. All users have to do is upload their scripts to the platform, and after a few minutes, it produces a report detailing things like an estimated budget, outlines of important story beats, and character analyses. The service costs $50 for each individual analysis, but you can also purchase multiple analyzes at a discounted price.

The idea for this kind of segmented analytical workflow first came to Wood — who is also CTO at Quilty — a few years ago when he was sued over a real estate matter. Instead of spending money on a lawyer, Wood unleashed ChatGPT, which promptly told him: “I’m not a lawyer; go find someone else to help you.”

“Then I went to Gemini, which worked better for a while, because I had a bigger contextual window,” Wood said. “But then I was on X, and He thought Elon Musk was stupid When you talked about how your puppy had the best lawyer’s degree ever for the AI ​​model, I thought to myself, “Let me check that out.” (Wood did not detail how this legal dispute ended.)

This experience has allowed Wood to better understand how similar consumer AI models excel at different tasks. Wood’s personal use of artificial intelligence has informed Quilty’s approach to measuring a text’s potential success. Because “Gemini is great at structure and styles,” Quilty uses that tool to help create breakdowns — documents that summarize all the elements of a movie or show’s production into comprehensive lists. For financial modeling, the company places its trust in an example Deep Sick Which is hosted on servers located in the United States. For narrative/character analysis, Quilty uses a combination of Claude and ChatGPT.

Wood told me that the company relies on context stimulation—a process in which additional contextual data is provided—in order to generate high-quality, hallucination-free output. Quilty does not personally train any of the models it uses to create its film reports/scores. But Wood insisted that this is a strength rather than a weakness because it makes it easier for Quilty to integrate new and improved models into its workflow when they become available to the public.

“When Claude Mythos comes out and I can see that it’s a better MBA, all of a sudden, my whole program becomes better,” Wood said, referring to The powerful new model This is only available to a small group of organizations for cybersecurity purposes. “If some Chinese model suddenly becomes better than all the U.S. border models, why shouldn’t I use those instead?”

While the modularity of Quilty’s technology stack may make it more flexible in terms of comprehensive updates, it also makes it somewhat difficult to fully understand how the platform takes a script and comes up with a set of metrics that allegedly measure intangible things like how an audience engages with it. may A reaction to a movie that doesn’t actually exist yet. Forecasting has been an essential part of film development since the birth of Hollywood, but this work has traditionally been done by human workers with a precise understanding of audiences.

No AI company has been able to develop a model that can replicate human thought processes or the imprecise way we form our opinions about art. But the founders of Quilty believe their “emotion engine” is the next best thing when it comes to evaluating texts because of the way it integrates… Tools like VADER (Valence Aware Dictionary and Entiment Reasoner) – An open source program that measures the extent to which text appears positive versus negative.

Horsman and Wood are also firm in their belief that Quilty can pinpoint how a project “addresses the cultural moment” and provide reliable box office forecasts. They pointed out Revenge of the nerds An example of a popular old film would get a lower Quilty score precisely because of the way it is filmed It attempts to portray sexual assault in a comedic light – something that modern viewers might view as being in poor taste.

When Horsman asked Wood why Quilty surrendered Christy (which eventually grossed about $2 million) Higher degree from Sinners (which grossed $370 million), they insisted that the platform’s verdict “boiled down to the fact that Sidney Sweeney is really, really popular.” They said that on paper, Sweeney’s star power is coupled with the fact that biographical boxing dramas are cheaper to produce than fantasy/action films like Sinners to make Christy A safer bet. But this situation highlights how unreliable Quilty’s reasoning is. Horsman and Wood admitted that there are some situations in which Quilty cannot predict factors that might affect a film’s financial performance or the way it is received by audiences.

For example, Quilty could not have predicted this Elijah Bynum Ahlam magazine (which Horsman produced) would end up derailing the actor Jonathan Majors’ notable fall from grace In 2023. Likewise, nothing about it Minecraft movieThe script was referring to the platform Chicken Jockey phenomenon He would become part of the massive success of the film. Horsman and Wood told me that they eventually want Quilty to be able to see these kinds of things coming, but it’s hard to imagine how that could come to fruition.

Despite all the hype, what Quilty is selling is indirect access to a variety of large language models that are asked to predict the future as they relate to unproduced pieces of art. It would be truly amazing if any of these AI tools worked as Quilty claims they can. But most of them are just sophisticated pattern recognition/simulation machines, and are far from being able to understand what humans find amusing.