Democratizing Data in Hollywood: Jumpcut’s Kartik Hosanagar
28:41 Share

Today, we're airing an episode produced by our friends at the Modern CTO Podcast, who were kind enough to have me on recently as a guest. We talked about the rise of generative AI, what it means to be successful with technology, and some considerations for leaders to think about as they shepherd technology implementation efforts. Find the Modern CTO Podcast on Apple Podcast, Spotify, or wherever you get your podcast.

How can AI help bring new ideas and products to market in industries where risk aversion is rampant? Find out today when we talk with Karthik Vasanagar, professor at the Wharton School and founder of JumpCut, a startup helping previously undiscovered talent produce movies and TV. Welcome to Me, Myself, and AI, a podcast on artificial intelligence in business. Each episode, we introduce you to someone innovating with AI. I'm Sam Ransbotham, professor of information systems at Boston College.

I'm also the guest editor for the AI and Business Strategy Big Idea program at MIT Sloan Management Review. And I'm Shervin Khodabande, senior partner with BCG, and I co-lead BCG's AI practice in North America.

And together, MIT SMR and BCG have been researching AI for five years, interviewing hundreds of practitioners and surveying thousands of companies on what it takes to build and to deploy and scale AI capabilities across the organization and really transform the way organizations operate.

Today, we're talking with Karthik Hossanagar. He's a professor at the Wharton School and founder and CEO of JumpCut. Karthik, thanks for joining us. Welcome. Well, thanks for having me, Sam and Shervin. So this is a bit of a different interview for me because I've known Karthik for years in academic circles. But Karthik, your latest venture is JumpCut. I think the idea basically is to help surface new and fresh stories for Hollywood. Can you tell us a little bit about that?

My new startup, it's called Jump Cut, as you mentioned. What we are doing is essentially trying to create a new data-driven studio that's reimagining the way films and TV shows are developed with the specific goal of elevating fresh new voices. And the context here that got me interested in this is Hollywood has historically been an old boys club.

A few execs making decisions on what movies get made, who's in those movies, at what budgets. All of these just pretty much based on gut and relationships. You know, who knows who.

And there are costs of this kind of decision-making. There's the economic cost. You know, Hollywood has had historically a very poor batting average. There's the social cost. By just about any measure, Hollywood has not been a particularly inclusive industry. And then there's the cost to audiences. And so what we're trying to do is break that mold and use data to make more objective, better decisions. But ultimately...

By doing that, we can assess storytellers and stories on their merit as opposed to, you know, who is connected to whom or just gut feelings of a few people. And so that's how we're trying to democratize Hollywood using data. I mean, maybe your world is slightly different than mine, but, you know, no movie execs are my classes. So how did you end up with this idea and how did you end up with Jump Cut?

Yeah, so it's interesting. No movie execs in my classes either, Sam. Do you know that I went to high school with Julia Roberts, though? So, I mean, there's a connection there. But anyway, sorry, keep going. Well, I hope you're still connected to your high school. Unfortunately, this is pre-social media, and I made lots of tragic mistakes in that era. But anyway, sad tale for another time. Right, right. Yeah.

Yeah, look forward to catching up on that story sometime. But yeah, coming back to your question, what got me interested in this? I've always had an interest in content and storytelling and films. I'm actually an amateur filmmaker. So back when I was...

a newly tenured professor, and also I didn't have kids. You know, it was an interesting combination where I had my weekends to myself. So I would go make short films and, you know, just put them out on YouTube. It was really just a fun hobby of mine. In fact, during my first sabbatical from Wharton, I wrote a screenplay. I flew into Mumbai, met with a bunch of film producers there and pitched my screenplay. A number of them liked it.

But, you know, the common response from many of them was, we really like this, but how do we take a bet on a completely new writer director? One of them said, I'll buy your script, but I can't have you directed. You may be good, but I just can't have you directed because I can't go to financiers and to actors and ask them to take a bet on a new voice director.

And so I, you know, my response back then was fair enough, makes sense. I realize why you can't do it. But at the same time, I didn't want to give my script away. So I came back.

Then my sabbatical was over. So back to Wharton, back to teaching and all of that. But over the years, I've met with so many writers and directors who have shared similar perspectives. There's a friend of mine who told me it took him 15 years to break in to the industry. And he's a successful writer director, 15 years for his first opportunity. And I hear this so many times. And in recent years, we see

Really successful movies or shows like Get Out or Stranger Things come from very unexpected places. So I've always been fascinated by this space. Initially, it was a hobby. But one fine day, I kind of felt like the problem I'm seeing is something my skills and data might have some relevance to use to solve.

This is actually quite interesting because there's several angles to the story here. One is the equity angle and the social angle and giving those who deserve and have the merit

the voice that they might otherwise not get and then there is the more business data driven angle which takes my mind to like money ball in baseball or underwriting you know giving credit to folks who don't have credit history because like they emigrate to this country and

But for the fact that they've just arrived here, they would be perfect in terms of the profile and the background and the education and being able to actually assess talent based on a bunch of attributes and features and sort of de-risk the very executives that say, how do I take a bet on you because I don't know anything about you?

Yeah, absolutely. And you used a very interesting word here. You talked about de-risking. You talked about credit applications and there are people who are applying and how do we de-risk them? And a lot of this is really about risk. If you look at Hollywood, one of the big things I've realized is Hollywood would like to take chances. They understand the problems of inequity. They understand the industry has not been inclusive.

But everyone's worried about the risk issue. What was interesting for me was a lot of execs in Hollywood told me, well, you know, the reason everyone's so risk averse is if you do a movie with Brad Pitt in it and it doesn't work out, you won't lose your job for making that bet. But you take a bet on something new, unexpected, and that doesn't work out, you'll have some explaining to do. Why did you make this bet?

You know, this is not the kind of bets we've made in the past. And I remember at some point I even read a statistic which was on movies and it said like 74% of the highest grossing movies are all sequels or adaptations of existing IP. And so there's no one willing to take a chance on something new. It's everything is a sequel as an outsider. I was wondering why is it that it has to be a bestselling book or something

comic book for something to get made? Why can't it be an original screenplay or an original idea? And it came back to, oh, yes, because IP has been de-risked. But the insight that I had was that IP isn't the only way of de-risking. There are other ways to de-risk as well. And so that's essentially how we approached it. And there's really three pieces, and I'll mention each at a high level, and then you'll tell me if there's any of them that's worth going deeper into.

One is, first of all, how do we discover stories and storytellers? The classic Hollywood approach is that studios get submissions of scripts from agents, and agents are representing talent, and people who are outside of the system, they don't have a godfather who'd connect them or don't already have the connections. And so the first thing we had to solve for is how do we find stories and storytellers that are outside of the system?

And so for that, one of the things we do is our algorithms are assessing content as well as creators on various platforms like YouTube or Reddit or even storytelling platforms like Wattpad and others. And what we're doing in those platforms, I'll use YouTube as an example.

If a screenwriter or if a director has created short films and has posted those short films and that are already resonating with audiences, then we try and discover them. So the idea is to find people where they are, that's whatever country, whatever platform you're in, and go discover them there. And essentially, the algorithms and analytics are trying to analyze the content to look for

High production value, which you can infer from the frames and images in the videos, for example. You're looking for strong storytelling and strong audience response, which you can infer from the kinds of comments that are elicited in response to these videos or even text stories. And so you can shortlist a set of stories or storytellers for our creative team to look at.

So it's not AI making the decision, but it's AI making humans efficient. Because if I had a creative team that had to scour through YouTube to find people

the needle in the haystack, I would need a massive team and several years to comb through 250,000 short films on YouTube. Or just my two kids. Well, that's true. That works too. Except I think they will just send me Minecraft videos or something like that. Given how many hours of YouTube they watch. But I digress. Sorry, go ahead.

Let's interrupt there. I mean, that's a little bit of a, that's the opposite of the Moneyball problem that Sherwin mentioned. Because Moneyball, you had the people and it was about figuring out which one was good. You're more about expanding that search space. Before we can say among the stories, which is the one to bet on, before we can do that, we want to know is the pool of people and stories we're looking at, is that the right pool or is that the complete pool?

And if we don't expand the pool, we are not solving the inclusion problem. We work with agents. We work with established writers as well. But in addition, you know, we're actively seeking out new people and not waiting for them to get discovered by an agent. And the agent then forwards them to us, but instead finding them where they are.

Okay. So that was one. Yeah. I'm sorry. You're going on with two and three here. Yeah. Yeah. And the second is, again, how do we de-risk them, right? So once we find somebody, it turns out they have, we ask them, hey, do you have a story for us? And they don't have one, they have 15. And then we hear them or we read them and we get excited with a few. We want to figure out how do we bring in some objectivity to this? And that's where data comes in. And

Some of it is backward looking, which is the classic machine learning kind of approach, which is let's look at data on what's been doing well. And that could be what movies are doing well, but also what kinds of stories are trending. And it could even be search queries and looking at where is the cultural zeitgeist? Where are people going? And understand which of the stories that we have

are stories that we think audiences will respond to. But that's not everything, because I think as long as we are backward looking and we're looking at what's worked, it's a fundamentally conservative approach because we'll do more and more of what's worked in the past. It's the sequel problem. The sequel problem. We'll be stuck there. And so the other question that we have to solve for is how do we go take a bet on something that's never been done before?

And there is no historical success there. And that's where we bring in ideas from A-B testing and experimentation. So we interact with lots of audiences online where we pitch stories and we're having multiple stories compete with each other and see which ones people are gravitating towards. We pair the classic machine learning based on past data with digital experimentation.

And we're running these experiments and then figuring out, you know, not just which stories we like, but also sometimes questions like, and this is all very hypothesis driven. And it actually also creates very interesting breakthroughs. Like one of the shows we are working on, it's with a very senior writer in this instance. He had a...

interesting sci-fi story, and it's a very high budget show. And he came to us saying, if I have to sell this, it has to be the case that this is a four quadrant show because of the budget, it can't be that this has a niche audience. Can your data help prove that this is a four quadrant show?

And we said, okay, let's test it out. And we tested it and it tested off the charts. And I went back to him and I said, it's almost a four quadrant show. It's testing off the charts. Gen Z, Gen X, millennial, doesn't matter what the age group, they're responding. Sci-fi, drama, all these different audiences, people who are into sci-fi are responding, but also people who are not into sci-fi or into drama, they're responding.

People in the US are responding. People outside in international markets are responding. So all that's great, but women aren't responding. And I said, well, this is what's going on. And his first reaction was, oh, we can't go with that kind of data to the buyers because if that hurts the show. And that was our initial reaction, which was to be defensive about it. Can we hide that data? Can we not show it and so on? And then as we were talking, it's like, why are women not responding? What's going wrong? We have this

show, you know, there's three main characters in it. There's a woman in there. There's two male characters. To simplify it, I'll just say there's the good guy, there's the bad guy, and then there's the woman in the show. And as we talked about it, you know, we realized, look, the woman in the show doesn't seem to have enough agency. She has no motivation. The female character is serving the motivations of the male characters.

And he realized that he's approaching it with this mindset and he's not thought hard about, you know, what's in it, what is driving her. And then he reimagined the female character. And now this, you know, I am getting into the weeds, but it's now the female character is a genetically crisper kind of female, you know, modified and she's got superpowers or special powers and so on. And suddenly when we tested the new version, women were responding to it.

And now the idea improved. And this isn't like soulless data telling you that you need to insert a chase sequence on page 13 to increase the audience. This is hypotheses and asking, what is my story missing? And really improving creative decisions, very much human-led, but data-informed. This is quite exciting. And what I particularly like about it, Kartik, is

you're breaking into the last standing bastion of judgment-driven decision-making, right? I mean, if you think about like 20 years ago, the loan officer making a judgment on who's going to get the loan and then fast forward to like 10 years ago where it's the retailer decides what price he or she should charge or how he should...

stock up the shelves. And all of those industries have been completely revolutionized with data and analytics, and they're making actually data-driven decisions, and they're doing test and learn, and they course correct. And we're seeing a lot of that in entertainment, but still a lot of the studios are very much judgment-driven. And I think this is very interesting because it's the beginning of the beginning for this industry. So hats off to you for doing that.

Well, yeah, thanks for saying that. I mean, I'm super excited to apply data to a setting where people are the most skeptical of, you know, should data be used here? And while, you know, we are finding value in this, I will also say that Hollywood has enough and more skeptics with regard to whether data should play a role. And sometimes there is this even misconception of what data is.

And sometimes it is, oh, there's an AI system making all decisions and I'm just, I no longer have a role to play. I think there's a lot of room for subjectivity, but how can data be a useful tool is how we're looking at it. And hopefully in five, 10 years, we can establish a track record of what data can do. And that might change the mindset here in the whole industry.

I did see a lot of that reaction saying, oh, data in creative has no role. But I saw another kind of reaction. And that was interesting. I had some people in the industry saying, this is inevitable. And Netflix is already starting some of this stuff. And we are forced by Netflix to get into this. And so we have to start doing it. And we may as well partner with people like you to understand this game.

So in fact, one of them told me that, you know, and this is quite simple, intuitive. This is a producer who told me that when his show came on Netflix within days, first of all, within a day or two, Netflix was able to give him feedback, of course, on how they thought his entire first season would play out.

And by the end of month one, they were already in conversations about season two because Netflix was able to project out what things would look like. Apparently, he was also told, given information about which characters audiences are most interested in. They wanted to kill one of the characters early in season two, and they were told, no, you cannot do this. That's your main character.

And so this producer was telling me, yes, I mean, Netflix is already pushing us to do some of this stuff. The difference is Netflix's data is coming in after the show is released and JumpCut is trying to bring in data before the show is created. But I think some of them kind of see this as being inevitable because of companies like Netflix and are happy to embrace our approach.

Yeah, I was a bit surprised because obviously your book from a couple of years ago was about how algorithms, I think in your words, are algorithms shaping our lives. But this seems to be one area that algorithms are really not shaping our lives. So how are you getting this constituency of people to pay attention to data and algorithms that if historically not? That's got to be not just selling a story. You're also having to sell an approach.

So in the book, which is called A Human's Guide to Machine Intelligence, I documented several ways in which AI is shaping our lives, as Sam was just mentioning, and including some examples in the movie world. So for example, I talk about how

On Netflix, algorithms are driving a lot of our consumption. There was a paper by data scientists at Netflix that said something like 80% of viewing hours on Netflix originate from algorithmic recommendations. So it's pretty clear it is shaping our lives, even in a setting like the movies. But I think while it's shaping our lives, certainly in terms of how we consume content,

How content is created, that side of the business hasn't really evolved. So the supply side still looks the same, even though the demand has completely been reshaped by the market. To me, there is a lot of analogs in what you're doing here to the evolution of data-driven decision-making.

Not replacing human, but basically creating a smarter, more effective, more efficient human. And there is exactly the same thing in retail and in personalization and in marketing. So like I said, it feels to me that entertainment and media has been sort of the last bastion of it. I think, Kartik, as you said, a lot of it is risk rather than anything else.

And I think in another book of yours, you alluded to the chicken and egg problem. If you don't have data on the talent, how do you make those decisions without some level of experimentation and some level of data? And then, of course, there's disruptors like Netflix who are forcing everybody to sort of become more data-driven as a way of surviving. And I wonder whether studios could survive 10 years from now without making a major step that way.

Right. I would think that, you know, the time has come. In fact, maybe the time was yesterday for something like this, which is, of course, why, you know, I went and started JumpCart. But I think it's inevitable because we know that human judgment has its big share of problems. Obviously, human judgment is also great in many ways, but we have our biases. We're colored by them in ways we don't realize ourselves.

And having a tool that can free us from those, I think it's a no-brainer that we should be embracing them. And yeah, the time has come to do this in fields that were unexpected. I think sports is not, again, a place where I would have guessed there would be early adopters, but clearly they have been. And they've shown that the Moneyball kind of approach works. What you really need is...

To integrate data into decision-making along the way, you have a deep understanding of data to know when to lean on it, but also when to question it. And you have a strong, creative point of view. And you bring the two together. So you've got to almost create it ground up, which is why we said we're not a data insights vendor. We're a company that's creating really a new kind of business that brings data and creative together.

You have to really close the loop because otherwise you have the problem that I think a lot of early adopters of the data-driven decision-making fell into, which is they said, well, data is our future competitive advantage. Let's acquire as much of it as possible and put it in

some Hadoop cluster somewhere only to know we can't do anything with it because we haven't thought through. But they've got it. They've got it. They have the data, right? And they're like, you know, the number of companies that have done just that, you know, 10 years ago, 15 years ago, even doing it now, they're like, we'll get all the data. Once we have it all, we'll figure out what to do with it. But as you're saying, sometimes it's about doing...

so much more with the data you already have by connecting it to the business process or the creative process and connecting the endpoints. Karthik, really enjoyed both your background and how you're applying this into your perspective on algorithms and the difference that algorithms can make in terms of exploring our search space, I think is fascinating. But also those parts two and three were pretty fascinating too about how to integrate that with creative. And we appreciate you taking the time to talk with us. My pleasure. Thank you for having me.

In our next episode, we speak with Sarah Karthigan about how she's helping ExxonMobil use AI for self-healing process improvement across business units. Please join us. Thanks for listening to Me, Myself, and AI. We believe, like you, that the conversation about AI implementation doesn't start and stop with this podcast. That's why we've created a group on LinkedIn specifically for listeners like you. It's called AI for Leaders. And if you join us, you can chat with show creators and hosts, ask your own questions, share your insights,

and gain access to valuable resources about AI implementation from MIT SMR and BCG, you can access it by visiting mitsmr.com forward slash AI for Leaders. We'll put that link in the show notes, and we hope to see you there.