Warner Bros. Signs Deal With Artificial Intelligence Analytics Company NPR's Audie Cornish speaks with James Vincent of The Verge about a new Warner Bros. deal with a company that uses artificial intelligence to predict movie success.
NPR logo

Warner Bros. Signs Deal With Artificial Intelligence Analytics Company

  • Download
  • <iframe src="https://www.npr.org/player/embed/795366576/795366577" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
Warner Bros. Signs Deal With Artificial Intelligence Analytics Company

Warner Bros. Signs Deal With Artificial Intelligence Analytics Company

Warner Bros. Signs Deal With Artificial Intelligence Analytics Company

  • Download
  • <iframe src="https://www.npr.org/player/embed/795366576/795366577" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript

NPR's Audie Cornish speaks with James Vincent of The Verge about a new Warner Bros. deal with a company that uses artificial intelligence to predict movie success.

AUDIE CORNISH, HOST:

Artificial intelligence has made its way into the stock market, our phones and our shopping carts. Movie theaters are next. Warner Brothers has signed a deal with a company called Cinelytic. It's a startup that claims its algorithms can predict how successful a film will be. And Cinelytic isn't the only company in the field. James Vincent has written about a few of them. He's a reporter for The Verge with an amusing professional bio - I cover machines with brains despite being a human without one.

James Vincent, welcome to the program.

JAMES VINCENT: Hello. Thanks for having me.

CORNISH: All right, how do these companies do what they say they can do?

VINCENT: So they work like pretty much every other machine learning application out there, in that they are fed a lot of data, and then they look for patterns in that data. It's kind of as simple as that. Cinelytic, for example, will look at the historical data related to a film.

So I had a sort of tour of the software from the CEO of Cinelytic, and he kind of showed me the interface for how the software works. And it's - you know, it's exactly like looking at sort of the Wikipedia page for a film. You have description about what it's about, who it's starring, where it's going to be released. And you can just literally swap in some of these elements. So if you want to see if, you know, Scarlett Johansson is going to work for the lead role, you can just swap her in, and you can see what that does to the predicted revenues in different countries.

CORNISH: Any evidence it works?

VINCENT: Yeah, so that's the really, really tricky question. If you ask these companies, of course they will tell you yes, it works. We can tell you which films are going to be a success.

If you look at the sort of the scientific literature on it, it's a bit of a hazy picture. They will tell you that, yes, there are some predictions that can be made, but they're not always particularly insightful. They might be something very obvious, like if you put Tom Cruise in your movie, your movie is going to do pretty well at the box office. If it's a big action film in the summer, then it's also going to do pretty well. So there is a question, to what degree are these important or even (ph) useful insights, or are they just kind of confirming what movie studios already know?

CORNISH: And what does this mean for the whole issue of diversity in Hollywood, right? You're feeding in data that hasn't yielded very diverse products.

VINCENT: Yeah. I mean, that's a huge, huge problem. Machine learning is, in some ways, fundamentally conservative. It learns from past data, and so it's going to repeat the patterns seen in the past. In the case of Hollywood, then, it's going to repeat the pattern of un-diverse casting. It's a big problem in many areas of AI, and I'm not sure how well it's being considered within Hollywood.

CORNISH: What do you know about how Warner Brothers is planning to use this software?

VINCENT: So a source inside Warner Brothers told me that they're mainly going to be using it for marketing and distribution, not necessarily to involve it in the greenlighting process for films. That means they'll still be using AI, they'll still be using algorithms, but it'll be looking at stuff like, what's the best audience to target for a certain film? And, you know, how do we best target that audience? What sort of trailers work for them? What sort of images worked for them in posters?

CORNISH: Isn't there an aspect of creativity that machines can't capture, right? I mean, a screenwriter on Twitter pointed out that the Marvel franchise was basically built on the decision to cast Robert Downey Jr., who at the time, as this Twitter writer describes it, a middle-aged actor not long out of rehab and prison.

VINCENT: Yeah. I mean, it sounds like a total cliche, but really, it's very true. The data can only tell you what the data can capture. It can't tell you about how the film was executed. You know, it can't tell you about what the shots was like. It can't tell you about the creative choices made by the producer. And if you speak to these startups, they will be pretty upfront about that. They'll say that their tools are mainly supposed to be assistive. They're there to crunch the data and give you insights that humans might miss. But ultimately, the final call is with the studios.

CORNISH: So it can't necessarily prevent, say, a remake of "Cats?"

VINCENT: I don't think anything could have prevented a remake of "Cats."

CORNISH: (Laughter).

VINCENT: From what I hear, "Cats" was a force of nature that just sort of dropped into their collective consciousness and did its damage.

CORNISH: James Vincent is a reporter for The Verge.

Thanks so much.

VINCENT: Thank you very much.

(SOUNDBITE OF MUSIC)

Copyright © 2020 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.