Algorithms play a huge role in the content people watch. Netflix, for example, has said that approximately 80% of subscribers trust the platform’s recommendations. But as artificial intelligence technology advances, it may play an increasingly important role in what films and TV shows are made. 

Bloomberg columnist Trung Phan recently wrote about AI’s potential in evaluating film and television projects’ commercial viability. The following is an edited transcript of Phan’s conversation with “Marketplace” host Kai Ryssdal about what he learned by having his own script analyzed. 

Kai Ryssdal: Tell me about this screenplay you wrote called “The Lose.”

Trung Phan: OK. So, I write now publicly on the internet quite a bit, including a column at Bloomberg. But 10 years ago, I was living in Ho Chi Minh City [Vietnam], and I had dreams of being a screenwriter. And I managed to put a comedy script together and sold it to Fox. The log line for that film was “The Fugitive” meets “Harold and Kumar,” set in Southeast Asia.” (Laughter) It was just ahead of its time — they weren’t ready to make it. TL;DR it didn’t get made, and now we’re about a decade later.

Ryssdal: All right. So how did it come to pass that you wound up writing a column about it?

Phan: So, separately, I wrote about the story of writing and selling the script ages ago. And the CEO of an artificial intelligence company called Corto AI happened to read my newsletter, and he’s like, “Hey, Trung, I read this article that you wrote about this old script of yours. And it just so happens that my company has technology that uses artificial intelligence to scan screenplays.” And the way they describe it is they look for the “narrative DNA” of a screenplay and they can basically tell you why a film could or cannot succeed. I already knew that my film couldn’t, but I wanted to find out.

Ryssdal: Yeah, being a glutton for punishment. … So they run your script through the algorithm. How do they know what they are looking for?

Phan: So the CEO told me that they have a database of about 700,000 scripts. And I guess in the AI industry, in machine learning, a lot of what you do is you kind of tag certain items. So, as an example, in script writing you can usually tell when a screenplay transitions from the first to the second to the third act. So basically, they have this giant catalog, and they wanted to run my script against their giant catalog.

Ryssdal: So they come back with some report, and they tell you what?

Phan: So the report comes back. I’m not going to bury the lede here. They said, “Your film is not commercially viable.” So they come back with something I already knew. Having said that, though, they correctly identified kind of the genre of the film. One of the top comparisons that they pulled up was “The Hangover [Part] II,” which was set in Bangkok. But they said two things specifically about my script that made it not supermarketable. They have these two scores that they calculate. One is called “interestingness,” and another is called “uniqueness.” So what interestingness does is it looks at the range of characters in a script. Obviously, most films will have the protagonist and the villain. But then other films — really good films — will have a lot of good secondary and tertiary characters. Apparently, my script didn’t have those. Whereas, if you take a movie like “The Godfather,” you probably have half a dozen or 10 characters that you see how they progress. But to my credit, that’s a three-hour movie, and I’m not Francis Ford Coppola. 

Ryssdal: And it’s worth pointing out here, actually — and this is my favorite part of it — that they also recommended that if this film did get made, if the studio put Chris Pratt in it, it might work.

Phan: Yeah, it basically chose Chris Pratt as the silver bullet for this film. Like, we looked through the entire 700,000 film and TV [show] database and, based on the interestingness and the lack of uniqueness, the other score, your only chance now is — you know, you gotta get Chris Pratt in the driver’s seat or this thing’s done.

Ryssdal: If Chris Pratt is a “Marketplace” listener, maybe we can hook you up. But wait a minute. You mentioned “The Godfather,” right? And so here’s what I want to go bigger picture on Hollywood and AI: “The Godfather” almost didn’t get made like 10 different times, right? Oh, and so now we’re trying to put AI into this unbelievably subjective “what makes a good movie.” And I just wonder what you think of that. Set aside your own bad experiences with your film, but come on, man, how do they know?

Phan: I am 100% on the same page with you on this. The one thing I will say is that AI is advancing just so incredibly fast, I’m hedging a little bit, but will I be completely out of a job as a writer in 10 years? Maybe, so there’s that.