On with Kara Swisher - Is AI Hollywood’s Newest Star? Runway AI’s Vision for the Future of Film

Episode Date: September 11, 2025

AI video generators are making storytelling more affordable — and one day, the tools may allow anyone with enough creativity to become a filmmaker. But the same technology can power convincing deepfakes that undermine our shared reality and destabilize our politics.  Cristobal Valenzuela, the co-founder and CEO of Runway AI, joins Kara to discuss how its tools are disrupting the advertising and film industries, and why Hollywood studios might not want to admit how much they are using them.  They also talk about how Runway is holding its own against (and warding off takeover advances from) much bigger tech players, and the AI video industry's responsibility in preventing deepfakes and political manipulation. Plus: Kara becomes a Top Gun pilot! Watch the video on the On With Kara Swisher YouTube page.  Questions? Comments? Email us at [e-mail] or find us on YouTube, Instagram, TikTok, and Bluesky @onwithkaraswisher. Learn more about your ad choices. Visit podcastchoices.com/adchoices

Hi, everyone from New York Magazine, the Vox Media Podcast Network. This is on with Kara Swisher, and I'm Kara Swisher. Really, it's me, the real Kara Swisher. You may think it's weird to stress that, or getting close to an AI-generated future where you might not know if the influencer you follow or the host of your favorite podcast is a human or an AI-generated bot. As First Lady Melania Trump said last week,

The Robots Are Here, Our Future is no longer science fiction. I think that's a really bad version of Melania Trump, but I don't know what she was talking about. That said, that's what I want to talk about with my guest today, Cristobal Valenzuela, co-founder and CEO of Runway AI, a media company that builds tools for creating AI-generated images and videos. It and its competitors are already disruptive, the advertising and film industry, and Runway is taking steps into gaming, and yes, Melania,

robotics. I want to talk to Valenzuela about how their products work, how they're competing with much bigger tech players in the space and warding off their takeover advances. I also want to talk to him about what responsibility the AI video industry has in preventing deep fakes, including ones that could have devastating political consequences. By the way, we tried out some of Runway's tools, and we're going to talk about that too. So if you want to see some weird AI videos of me. I think I look great. Check out our YouTube page On with Carous Swisher. Our expert question today comes from entertainment reporter Matt Bellany, founding partner of Puck News and the host of the Hollywood Insider podcast, The Town with Matt Bellany. This is

the kind of lively conversation that's hard to fake, so stick around. Support for On With Caraswisher comes from Sacks Fifth Avenue. makes easiest shop for your personal style. Falls here, so it's time to invest in new arrivals you want to wear again and again, like a relaxed prod of blazer and Gucci loafers, which you can take from work to the weekend. Sacks makes shopping feel customized to you from in-store stylist to their website Sacks.com, where they show you only what you like to shop. They'll even let you know when your arrivals from favorite designers are in or when something you've been eyeing is back in stock. So if you love a personalized easy shopping experience, head to Sacks Fifth Avenue for the best fall

all arrivals and style inspiration. Support for this show comes from Robin Hood. Wouldn't it be great to manage your portfolio on one platform? With Robin Hood, not only can you trade individual stocks and ETFs, you can also seamlessly buy and sell crypto at low costs. Trade all in one place. Get started now on Robin Hood. Trading crypto involves significant risk.

Crypto trading is offered through an account with Robin Hood Crypto LLC. Robin Hood Crypto is licensed to engage in virtual current. business activity by the New York State Department of Financial Services. Crypto Health through Robin Hood Crypto is not FDIC in short or CIPIC protected. Investing involves risk including loss of principle. Security's trading is offered through an account with Robin Hood

Financial LLC, member CIPIC, a registered broker dealer. When you're with Amex Platinum, you get access to exclusive dining experiences and an annual travel credit. So the best

Tappas in town might be in a new town altogether. That's the powerful backing of Amex. Terms and conditions apply. Learn more at amex.ca. slash Y.A.X. Chris, thanks for coming on on. Of course. Thank you for hosting me. Let's start with talking about the idea of generative AI video.

It's a hot field right now and runway is one of the top. top players in the industry. I want to ask you about two things you wrote in your blog, which has a lot of philosophical musings about art and AI. You're actually more educated than most people I interview in tech. In March, last year you posted, the medium is not the goal. Art is not finished because AI can generate images. We haven't solved cinema because we invented a system that creates videos, not all videos are cinema, not all images are art. And then last month you wrote, this is a new medium. The only way we'll uncover what this medium can do

is to stop judging it by what came before. Stop looking at the surface, start experimenting with the core. We're not watching films evolve or watching something being born. So it feels like there's an evolution between those two posts. But let me have you tell me

what you were meaning in both of those. I think it's part of what I'm trying to recognize and understand and make sure I can communicate is helping people understand the expectations around AI, specifically when it comes to art and media. And so the first pose, I think, it's mostly about trying to make sure we distinguish the idea of AI being able to create sequences of moving images,

with that being a replacement of art and videos and films and things we care deeply about, not because of how they were made, but because of what they transmit to us. And I think a common, I would say, approximation around AI when it comes to videos to think that because we can generate videos, because we can generate consistent, coherent sequences of things with AI, then therefore, like, AI is like, it's the doom of Hollywood, and we're suddenly, like, in a crisis mode. That would be the story. That's the narrative.

Right. And my point is, like, it's not. It's just a fancy way of making stuff. It's a camera. It's a medium. It's a tool. And in order for you to use it for storytelling, for art, for movie making, for whatever form of expression you want, you need to master the tool.

And you need to master the tool in the same way that painters have mastered the canvas and filmmakers, a master of the camera. And for me, this tool is not working in the same way that we've used other tools. And so the second pose kind of like touches on that on the idea that in order for you to understand AI and what AI can do for you, you need to kind of like move away from how you've created videos before. Because if you thought about the way you've created films or videos before, you're bounded to how, you know, cameras work, to how traditional CGI works, this new medium requires you to kind of, like, lead behind those preconceptions and step into a brave new world.

And I think part of those observations, I think, are coming just from experience seeing people make things with runway, make things with AI, and really understanding that through the history of technology, every time we've seen a increasingly radical change in technology, it has affected art in ways that artists at the time couldn't predict it. You said, stop looking at the surface, start experimenting with the core. You're saying, stop belly aching. It's not necessarily a job loser, although it could be, that people are immediately jumping to conclusions, presumably, about it, which is very common.

What's it going to do to silent films? Well, you saw what happened there. What's it going to do to radio when television came? What's the Internet going to do to media? And a lot of damage in some cases, a lot of great things and others. So how are you thinking about it right now? Do you think that has shifted? Because it certainly hasn't shifted in the common narrative in Hollywood, for sure.

People are worried. No, I think it's a shift. It has changed a little bit, to be honest. I think people are recognizing have the time to experiment more with the tools. I think two years ago, people were coming from maybe from where you were mentioning this fear of like what AI could represent. But eventually, like, there's only one way of like moving away from fear, which is just, well, let's try the thing.

Let's see why it works. And then eventually you try to understand that there's a lot more into filmmaking than making moving images. And there's a lot more to art than just making pretty pictures. And I think the overall sentiment, at least from my experience within Hollywood, is that that vibe has changed for good reasons. And I think it will continue to change. What's the vibe now, would you say? And overall, what is the worst you're hearing and the best? I think the vibe has shifted towards, I think, acceptance and understanding of this as a creative.

medium as a tool, as a powerful tool that comes with a lot of challenges, but one that if you know how to use well, it can take you very far. So runway is being used in Hollywood and by big ad agencies on Madison Avenue. Obviously, it's not just spitting out TV series or a full ad with a couple of prompts. Explain how your tools are being used by professionals and give some examples. I know everyone likes to be secret about AI and film, but I'd really like some specifics. So there's a lot of that goes into making, let's say, a future film, which, which is actually very similar to how you make ads.

Just like the only difference may be the time, the budget, and the kind of like ambition of the project itself. But think about everything that goes before you actually shoot or make the final frames, which is pre-production, storyboarding, scripting, previs. Those are all things that are knitted for you to iterate on the story, the angle, the way you want to tell it, the art direction. Many parts of that process are now leveraging or using AI, including runway.

And so if you're a storyboarder or you're a screenwriter and you're working towards getting something done in the next couple of weeks, you can kind of like go from that idea to like a storyboard in a few minutes by using runway. And then after you've done with the pre-production stages, going into more of the post-production, the visual effects part of things, is where you can leverage these tools to make you go through the process of editing way faster. And so instead of spending too much time on a particular scene, you're going to have an AI system that can aid you on generating parts of it.

Will it make your film? We're going to make your ad? No. Will it help you? Yes, a lot. So give me an example of something recently you can talk about. On the film side, there are a couple of films that have leveraged runway in a particular set of scenes, mostly for either visual effects or, again, for pre-production.

So the thing is, like, for pre-production, you don't sit. Like, there's nothing you can actually, there's nothing tangible because, like, it's part of the process of making something, and that helps you iterate on that process. So, like, a storyboard you were talking about. Correct. Explain what that is for people. people who don't know. Sure.

So think about animation, for example. Animation is very expensive to make. And so you want to make sure you can iterate as much as possible on the early stages of the animation pipeline. So when you're ready to like render the frames and create the animations, you kind of know exactly what you need. The cameras, the angles, the characters, the positions of pretty much all of the objects and characters.

And so what animators have kind of figure out is that the fastest and easiest way to is on the storyboard. You basically create this kind of like quick drawings of exactly kind of what you want to tell the story and how the script kind of like translates to that that board. And then you kind of create a small sequence of videos that actually you... To see how it looks. Right, exactly. It's kind of how it looks.

And you're like, ah, you know what? Now I'm looking at it and like the joke or the line really doesn't work that way. So we should come back and redo it again. Most of the time you're spending on making something like that is on that cycle. Right. So you're seeing the story before it is so you don't make a mistake when you're going for the real thing. Correct. And the reason because it's like, again, it's costly.

It's very expensive. Of course, it's very expensive. It's very hard. And so what AI is kind of doing is compressing a little bit of that iteration cycle. So instead of waiting another week to see the storyboard, you might happen in a day or so. That's great. Now you can iterate faster. You can relax, see the scene again and be like, okay, let's change it again.

Right. So clearly we're not Hollywood professionals. We tested runway out. And I want to play our video and then a video that your team sent over and chat about the two. And by the way, for our listeners, we have these videos on our YouTube page so you can go see them in all their glory. My team used three pictures of me, one in which I'm wearing a Revolutionary War costume, and don't ask why, it's a project I'm working on. One was taken after my son's graduation with him in a cap and gown. And the third, I'm playing with my kids

in a ball pit. We prompted Romney to combine them into one image, and then we took the image and turned it into a short video. Let's look at it. Prompt. Generate a realistic image of Kara Swisher. Prompt, combine revolutionary figure in ball bath with son wearing graduation cap and gown. Prompt. The boy wearing the graduation cap and gown pushes his mother wearing the Revolutionary War outfit for fun into the ball bath and the two of them pretend to fight. Prompt.

The boy wearing the graduation cap and gown pushes his mother wearing the revolutionary war outfit for fun into the ball pit, not bath. Runway did not know what I look like, obviously, my son and I are falling back into the ball pit, but that looked kind of cool. I kind of look like a man and he looks almost feminine. Explain what it did well there and what it struggled with and why. Yeah, so I guess it seems like the... Besides, we suck. But anyway, go ahead. No, look, I think this is a great example of what we're articulating before, which is it's a

tool. And like using a tool requires you to understand how it works. and what it does and you can do. And if you're new to it, you're probably going to not be good for the first times. It's going to take time for you to understand how it works. And I think part of what we're trying to showcase and hopefully we can see the video that we made is if you spend the right amount of time with the tools themselves,

you can make great content. If you just come and type your name and it doesn't make the thing that you want, it's probably because you need a bit more time understanding how to get to where you need to go. So what went wrong there? What did we not do, right? Why does he look like a lady and I look like a man suddenly?

Well, I think I'll have to look at the inputs more particularly, more specifically. But in the first example, for example, just with your name, the model might not know exactly what you're talking about. Like, who is this person? Like, whose car? Do you need more pictures? Yes, I need more references, right? And I need to be specific about exactly what I want. I think a common misconception with AI videos that it's going to work in the same way that chat boards work.

Right? So I think it's a realization these days that most people's experience with the AI. via like, I don't know, chat GPT or a clot. So it'll know Mount Everest. Right. So it's a factual thing. You ask a question and like you get an answer back. But the thing is in art, in video, there's no concrete answer.

Like when you tell me, I want to see a picture or I want to see a scene of me jumping through a building, like what you have in your head, it's very different from what I imagine in my head. And so we're never going to reach like consensus unless you show me which the most amount of references and details exactly what you want. Right. And so that's a good reflection overall of the video itself, that it's not a bad start. It's definitely like a good start, but it requires you to now understand that you might need to go deeper.

Maybe a close-up of your face might be better than like a medium shot, because there's a lot of things in the image that are going to get lost. Right, right. Maybe you want something that might require you to, like, split the images into four parts first and then combine them together. So it's a little more complex than, say, a chat, but people are used to the chatbot having the relatively correct answer. It's getting more correct than ever, I guess. All right. Your team, the video your team sent over is obviously more professional, and I'm a top gun pilot. Thank you very much.

Let's watch. I want you to talk about how this was made. So a couple of things. I'm a lot of challenges, getting the movement right. My glasses, you'm completely wrong, but that's me looking at it. Talk about the making of this and how long it. took and the biggest challenges. It's also highly entertaining and well done. It's not that. It's pretty good. It's actually very good. So talk a little bit about that. It took around two hours

of one person's time to make that video. The biggest, I would say, challenge was the comments you're giving us, which is like, oh, the glasses on the right are not exactly the same. And my face looks slightly different from this shot to this shot. It basically is the consistency and the controllability aspects of it are, it takes time for you to understand how to manage and iterate through it, to get exactly where you need to go. But the fact that you can make a video like that in like two hours

for me is... So you have templates of a... My sea captain is not very attractive, but you have templates of an astronaut. I assume that then you lay a real person. No, it's all generated. No, it's all generated. Okay. Yeah, it's all generated from scratch.

Those videos don't exist. None of those sequences have ever been created before. We're just customizing, creating them for you. And so you can think about it as it's an infinite amount of like things, combinations and videos that you can make and with the right inputs and with the right like experience you can make, again, things like that in two hours or so. And of course, if we spend way more time, we can make it even better. But I think to your point, it's actually really fun. Like making stuff like this, it's just so entertaining. It's fun to be able to like take an idea

that you have in your head and do it for the sake of like, you know, entertaining. You're What did the challenges mean for Hollywood then? Obviously, it took someone who knew how to do it two hours, and it's still not perfect. But what did the challenges mean for Hollywood then? Training people to use this, you know? Again, setting the expectations in the right way. I always mentioned this to people that think about this as a new camera, right? And what was the last film you watched?

Oh, Superman. Superman. So if they gave you the camera that they used to record Superman, would you expect to make something like Superman? No, no. Of course not. This is kind of similar. If I'm giving you the tools and I'm showing you a video, where you make the same video just because they give you the tool in like a minute,

you're probably like, no. You need to spend more time. You need to understand it. What about people who are experienced in CGI? I mean, like a lot of people can. Original computer programmers didn't necessarily shift to the Internet as quickly, but some did, right? Or new people came along and had a different.

100%. Yeah, that's kind of a point. For a lot of like CGI and visual effects artists, this is amazing. Like, you're going to be able to do things that used to be very hard and expensive. And the software that people use that it's like 20 years old.

And so I always joke with people in the BFX and CGI community that this is a way for them to get a weekend off because if you're working a production cycle, there's probably getting a lot of notes on the weekend and you're iterating or cranking through every aspect of everything you got. And imagine a system that allows you to do those iterations faster. Like, great, you can go on a weekend now.

Right. You're going to have time off. I think they're worried about total time off, but go ahead. Well, that's my point. I think there's parts of it that will definitely task that will go away. Like there are things that you just might not need people full-time dedicated for that. But to your point, I think it also opens the door for many other people to discover an entire new task and entire new jobs. I think it's a consistent pattern in technology and art and filmmaking that technology announced you to do just a bunch of new things you were never thought of doing before.

Yeah, that's the technologist's argument. We'll be back in a minute. Did you lock the front door? Check. Close the garage door? Yep. Installed window sensors, smoke sensors, and HD cameras with night vision? No.

I... And you set up credit card transaction alerts, a secure VPN for a private connection, and continuous monitoring for our... personal info on the dark web? Uh, I'm looking into it. Stress less about security. Choose security solutions from TELUS for peace of mind at home and online. Visit tellus.com slash total security to learn more.

Conditions apply. Support for On with Keroswisher comes from LinkedIn. As a small business owner, you don't have the luxury of clocking out early. Your business is on your mind 24-7, so when you're hiring, you need a partner that works just as hard as you do. That hiring partner is LinkedIn jobs. When you clock out, LinkedIn clocks in. LinkedIn makes it easy to post your job for free, share it with your network, and get qualified candidates that you can manage all in one place.

LinkedIn's new feature allows you to write job description and quickly get your job in front of the right people with deep candidate insights. You can either post your job for free or pay to promote in order to receive three times more qualified applicants. Let's face it, at the end of the day, the most important thing for your small business is the quality of candidates, and with LinkedIn, you can feel confident that you're getting the best. That's why LinkedIn claims that 72% of small business owners who use LinkedIn find high-quality candidates.

So find out why more than 2.5 million small businesses use LinkedIn for hiring today. Find your next great hire on LinkedIn. Post your job for free at LinkedIn.com slash Kara. That's LinkedIn.com slash Kara to post your job for free. Terms and conditions apply. Hey, box media listeners. It's Mike Murphy. What happens when you get two political hacks who've been running campaigns against each other for forever and had a world-class journalist? You get a big bar tab. That's what you get. This David Axelrod telling you you also get a great

podcast called Hacks on Tap. This is John Heilman. I'll tell you what we'll give you a weekly podcast that covers news, the headlines, and also the longer-term friends driving our politics from the perspective of the three guys who've seen it all from the campaign trail to the forward cabin of Air Force One. Join us every week on Hacks on Tapp. On the Vox Media Podcast Network. In every episode, we get a question from an outside expert. Let's listen to yours.

Hi, it's Matt Bellany, host of the town podcast and founding partner at Puck. I have a question for Crystal Ball. Runway has announced a couple of deals with Lionsgate and with AMC networks to integrate runway models into their workflows internally. But there haven't been a lot of other announcements with big studios in Hollywood. And I'm curious why. Lionsgate and AMC are two of the more struggling studio outlets at this point. Lionsgate's looking for a sale. AMC has had its value drop significantly. It seems like they're a little bit more willing to do these deals because they are struggling. Is that the case? Will we see others jump on board with these types of deals to let

runway into their internal workflows? Or do the others just have a little bit more power and are kind of resisting at this point? Thanks a lot. Now, just before you answer, I know that Netflix had been using your tools. Disney is said to have been looking at the technology. But answer Matt's question, what's the difference between Hollywood using your products versus integrating them into the workflows themselves? I think that for me the question was more about who's announcing it publicly and who's actually using it. I think it's fair to assume that you should expect a lot of, if not most of studios and people

experimenting with AI in Hollywood. I think people are getting more comfortable also with announcing and speaking more publicly about it. But like the landscape partnership for us was years in the making. And we announced it like recently, but it's been happening for quite some time. And so we've never done relationships with studios or with other partners for the sake of announcing it. It's more of like they feel comfortable with announcing than like we should. But there's a lot of work behind the scenes with other studios as well.

What about integrating into their workflows? Well, that's the goal. Like we want to integrate. We want to make sure if you're making something, we can help you in ADU at, you know, getting that stuff done, faster, cheaper, better. Everyone works differently. Everyone does things differently. and so there's no one side fits all, to be honest. I think we want to work with a lot of them.

We actually have a specific studio team inside the company. So Runway is a research technology company that has a, you think about it as like small... You're embedded. Exactly, a small studio. And the goal for that studio has two kind of like main goals. One is we're making original content ourselves. And so we kind of use the tools and make animations and make short films.

And that helps inform the research on the product itself. And then we also go into studios and kind of have that people work alongside the other, our clients, our customers. And that's, again, we're creating this new kind of camera. And so we need to show you how the camera works. And when they're using them, usually all the studios are trying these things out or trying these tools out or looking. And then possibly they don't want to talk about them because of the fears around from the unions and jobs and things like that would be my assumption. I think partially it's like competition. Like if you figure out how something works and it's given an edge and a very competitive landscape like the media market,

it is. You want to make sure you can just keep that for you. And I think there's a lot of internal, unique pipelines and workflows people have been working with AI. They're free, I think are very unique and it's fair for them to try to keep them like as private as possible for as long as it can. Renway has been called the number one all in one powerhouse for AI video generation, but you have huge competition. There's Google VEO, Meta's movie gen, Midgerne and of course SORA from Open AI, which just announced it's working on a feature. length animated film called Critters, with a Z. Runway sponsored an AI short film festival that ran in IMAX theaters this summer, but

Open AIA will debut at Cannes. Talk a little bit about this. The film, this film is supposed to have a much shorter timeline, nine months, instead of three years, and a much smaller budget than your typical animated feature. Yeah, so I've been working on this for seven, almost eight years. And I think we've created an industry and the idea that you can use AI for like film and art. And my realization is from a market perspective, if the market is interesting enough, you're going to attract some of the biggest competitors to try to build the stuff that you're trying to build. And I think that's a great validation for us. It's like we were a small team trying to do this very hard thing.

And now, like, all of these people are now trying to build similar things. I think it makes us like realize that there's something special here. and it also makes us realize that we need to continue being very, you know, independent in how we think about the roadmap of the product. I give you an example. We posted our film festival for almost three years now. Before anyone else thought about AI and film, we celebrated our third version of a film festival a couple months ago.

We sold out the Lincoln Center in New York City. We had thousands of people come and watch the films. We partnered with industry standards, including like the Tribeca Film Festival. and the Gotham's in New York. And I think part of it is we've always thought about making sure that we can work with industry really closely. And we realize that also will attract others to try to do it. To try to do it.

The worry, of course, is you're either Netscape or Google, right? You're either the one that pioneered it in one or pioneered it and got run over, essentially. Well, I think you have to... Same thing with Open AI, possibly, by the way. I think, I mean, look, the market is competitive. There's no question around it. I think AI itself is very competitive and it's very intense. But I think we're very confident in like, well, if we pioneer it, we need to make sure we can keep on doing it.

And there's no, it doesn't mean that we've reached, like, now is the final form. There's so much research and so much interesting ideas of products that you can continue building. And I think what I've realized over time is you deeply, you have to deeply care about what you're trying to solve and work for. Like we, I work on this, not because I thought it was a good business. It's my obsession. It's what I deeply care about. It's my, you know, combination of interest from like software to art to research. church. And I think our team deeply cares about that more than anything else. I think that overall

eventually becomes a much more powerful way of winning the market than just like throwing money of the problem. Well, meta was trying to beat substack and it just didn't, for example. But let's be clear, what usually happens is that one product becomes the industry standard, the Kleenex, so to speak. And on your blog, you recently wrote a big mistake is to think AI will play out similarly to how the internet story develop. The truth is the internet needs thousands of company to build infrastructure, AI needs perhaps a dozen. Talk a little bit about this. Yeah, so I think I'm realizing that a lot of value in AI goes to the extremes. So the cheap manufacturers, the silicon, and the other side, the research lab, the model builders and the

tool builders. I think what's different in AI with previous tech transformation was that the fiction shovels, everything in between there, I think it's eventually going to get capture or like devalued by those building the models themselves. And this is coming from my experience where I guess what I'm trying to say is we get a lot of, I don't know, products and services and companies that want to offer us things to build our models. But we've developed most of those things kind of like in-house. There's no need for us to like outsource that because it's critical to our business. It's critical to the research that we do. And so it's more of a, I would tell you, like the Apple Playbook where you have to build the entirety of the stack or like SpaceX.

We have to build like almost to the nuts and the bolts. The value there is that allows you to have a lot of control. You know, if I need to change something on my model or in my training infrastructure, I can just do it. It's difficult. It's way difficult because you have to build pretty much every part of it. But obviously, you have competitors with a lot of deep pockets. Google has a market cap of $2.8 trillion. Medes is $1.9 trillion, give or take.

Open A's latest valuation was around $500 billion. And after your last wedding round, it's nothing to sneeze at runway was valued at $3 billion, but that is kind of small. Again, I don't want to play that down, but you're a comparably small company, and the CEOs of those companies all had dinner with President Trump at the White House last week. I noticed you weren't there, but don't worry about it. I feel better about you because they were toadies upon toadies. How do you or other small AI companies compete with those deep-pocketed tech giants who are trying to get any angle they can in this sector? It's just relentless execution, the best product experience. It's David versus Goliath. Like, you've, these companies need to be challenged. Like, we can't believe in a world where those, like, one or two companies deciding every single product and experience we have. I think companies like Runway are trying to question that to the core. And I think part of it is we need to recognize if we succeed, we're going to attract them to try to even compete more with us.

I think it's fine. We've done this for quite some time. We've had a lot of competition come and go. But we're still here. So what is the slingshot then? What is the weakness they have from your perspective? And what's the strength that you worry about? I think, look, I'll tell you a story I've heard recently from one of our, like, clients and customers.

He's a head of a studio. He has made movies that you probably have watched multiple times. And many folks in the audience might have watched multiple times. He told me that the reason that they can go with any of the large companies is that they fundamentally don't understand what they're trying to do. And so there's a little bit of that argument. that it comes across just like the runway when you show some filmmakers like an AI tool and expect them to use it.

And it feels a little bit, I would say, rub them the wrong way. We're different. We come more from, like, I would say a similar background to them, to the type of problems you want to solve. And we sit with them understanding the challenges of making something

like a future film or a commercial or an ad or a short film. And the tools that we build are with that in mind. And so from a user perspective, I would say there's something around knowing who you're solving for and understanding their pain that I think it's different for the people working in the arts and in media. And I think that for us has been a huge advantage for a time. I mean, they wanted to work with us and not with everyone else.

Right. What's the thing you worry about is just the money and their aggression? Or what is there? What do you mean from them? From them, yeah, from the bigger competition. I get it. Netflix ran circles around all of Hollywood, right? You know, there's example after example.

There's also example after example of getting run over, right? Sure. Yeah, I mean, you have to just, like, keep that in mind. I do feel like if you just look like practically what has happened over the last couple of years, all you're saying is true. All of these companies have a thousand times the budget we have, a thousand times the researchers and the engineers that we have, yet we're still winning on the best at what we do.

So there's an argument to be said that, like, that shouldn't be the case because, like, all these companies who have just invested way more. And they have, they have invested way more, yet they still can do it. And I think part of it, again, is like, it takes a different set of cultural and intuitions and priorities to try to lead the way and be the one, like, helping versus, like, trying to catch up. And I think a lot of these large companies are playing more of a catch-up game, which I think it's something we're thinking about, of course, like, there's still competition. Right. But I think catching up is just different if you're trying to, like, make sure you're in the edge and you're helping customers, like, understand. understand what's next, versus just selling whatever the market has already, like, stab it issues the thing.

So, but Open AI Meta and Google are offering these huge salaries, tens of millions, hundreds of millions to recruit AI researchers and engineers. Is that a difficult thing to deal with these salaries? Because that's something that's affecting the whole market, I think. I don't know. I don't know how much of that narrative is actually true, to be honest. I think it's definitely like a good story and a good headline, but I don't think it's that true among the, like, larger research industry. and I've compared notes with other companies and our researchers. And it's not that everyone is getting that.

No. I don't think, yeah, I don't think it has affected like most companies, to be honest. In the way, it's just a big story. That's a very good point. Yeah. Speaking of Mitch, there was reporting earlier this year that META was looking at runway as an acquisition target. Before it bought a huge share in scale AI and aquired Alexander Wang. You've said you're hitting an inflection point is getting too exciting for you not to be independent.

As someone who has stayed independent myself, I get the attraction. and everyone was going to kill me, and I'm still standing, and they are not killing me. I imagine it's still very hard to stand dependent when the checks keep getting bigger and bigger. Depends on what you're trying to, like, obsess around and, like, work around. Again, I don't see myself doing anything else and, like, building runway. And so the team, I feel very similarly to that. And I think for us, like, thinking about selling the company around, it will be a mistake. Like, we're getting to a point, as I was saying before, it's an inflection point.

The models are getting good. Adoption is growing. It's the most fun. It's probably like similar to when, like, you went independent. It's like, now it's the most exciting time to work on this. I think that's my north of right now, just making sure we can keep doing that. Meaning that you're doing well, so why should you? So I just was actually talking to someone about a sale of a media company.

I said, you don't sell if you're doing great. Yeah, of course. You sell because of another reason. Do you feel like you have enough control over the company to be able to resist that? Yeah. You feel like you have that if there's some crazy number came in from somebody. One of the dark flip side is we talked about the huge salaries for AI engineers, a lot of workers, as they said in the film and advertising industry, could lose their jobs, everything from actors to animators. And I know tech people always said there will be new jobs. How do you calm those concerns? What is your patter about that when you're trying to make deals with the film industry, including unions, this has become a rallying cry for them? So what is your best argument for them when this is the concern?

And especially because no one quite knows, as you said, what it's going to be yet. Well, first of all, engage with them. I've met with the unions themselves with the guilds. And I think part of it is, again, helping understand what this technology actually does and how you can benefit from it. I think my overall summary has been focused on people rather than jobs. like jobs change. Technology will change jobs all the time. Look at the history of film. It has changed many times before, through many of the decades

we've seen. And it wasn't really about the jobs as much as like the people doing the jobs. And so if you're hiring people, if you're a guild member, if you're in a union, like, well, help your people understand how to use these tools, like train them, get them on board with the latest, understand how they can upskill what they already know. I think it changed. changes a little bit the perspective, and I think that's been my position so far. I think it's allows us to work alongside them in much more productive ways. We'll be back in a minute. Hi, this is Bella Freud.

Each week on fashion neurosis, I interview highly creative people in the world of fashion. film, art, sport, literature, and music. The format of the show is that my guest lies on the couch and I sit in a chair like an analyst. I know. I ask questions, quite simple things that revolve around how we reveal or hide things through what we wear, how we do or don't draw attention to ourselves.

This week I welcome Amelia de Moldenberg, the host of the YouTube series Chicken Shop Date. One of my friends came up to me, I remember, like, on the early days and said, do you mean to dress frumpy on the date? I said, frumpy, this is my wardrobe. Find fashion neurosis on YouTube or wherever you get your podcasts. According to the richest man in the world, my guest this week on Solutions with Henry Blodgett has a heart that is, quote,

filled with seething hate. I have a different view of Kara Swisher, and I was very excited to ask her all about how she became such an incredible and incredibly outspoken journalist. When I am talking to students, they're like, what's your best piece of advice? I said, you're going to be dead in a hundred years. That's really my best piece of advice.

Hear more of Kara's secrets to success and world domination. Follow Solutions with Henry Blodgett wherever you get your podcast. Hello, Daisy speaking. Hello, Daisy, this is Phoebe Judge from the IRS. Oh, bless, that does sound serious. I wouldn't want to end up in any sort of trouble. This September on Criminal, we've been thinking a lot about scams. Over the next couple of weeks, we're releasing episodes about a surprising way to stop scammers.

The people you didn't know were on the other end of the line, and we have a special bonus episode on Criminal Plus, with tips to protect yourself. Listen to Criminal wherever you get your podcasts and sign up for Criminal Plus at this iscriminal.com slash plus.

One of the things I think many people in how at least they're worried about is the more realistic gen AI video gets the bigger concern for privacy and for security. Con artists are using

deep fake videos of celebrities and pro athletes but also everyday professionals like doctors to scam people out of money. It's a huge issue also for lawmakers because it's gotten their attention. There were deep fakes of Congresswoman Alexandria Ocasia-Cortez and Senator Amy Klobuchar circulating early the summer. President Trump, of course, has reposted doctored videos, including one that portends to show President Obama being arrested in the White House.

Talk about this issue, because one of my big things is many of the big tech companies have no interest in safety whatsoever. But what are you doing or willing to as a company to prevent your tools from being used for, scams or even political manipulation. It may not be your full responsibility, but it certainly is partially your responsibility. Yeah. Well, I can really speak about other companies on their approaches, but I can tell you about ours. We take it very seriously, and I think that's why most of these studios want to work with us and are approaching us and have adopted runway. We have a trust and safety team. We have a terms of service. There's stuff that we won't allow you to generate. I think there's a particularly consistent approach towards

moderation and safety. And again, part of it is us doing things, but also learning from who we're working with. I think we've erred sometimes on the side of being too safe. We've moderated people from users, their own likeness at runway. And so if you come and you try to unit with your picture, we might block you. And you're like, but it's me. Like, I own my, I can do it. And so it's a hard problem because, like, you want to make sure people understand the right way if you're seeing it. And I agree with your point. It's, we don't have, it's not that runway will solve and we'll have all the answers to a question like that. I think a lot of it has to do with just the social and cultural implications of how you

help people understand whether you can do and come to a world where it's going to be perhaps different from what you've seen media before. I do feel that a lot of the questions around, like, likeness are still somehow the same that we've dealt with it before without AI. Like, you can still get in trouble with non-AI by recreating someone's likeness. That doesn't change. you still should be responsible how we use the technology itself. And it's in the user's, like, best interest to make sure they're using it the right way.

Would you support Congress codifying guardrails that were specifically related to AI? I think there needs to be norms and changes on regulation to adapt to AI. It's a reality. It's like loss change to adapt to the Internet. I'm not sure how and when, to be honest, because it feels very early. still, and you need to make sure you understand both innovation as an incentive for companies to keep, you know, making progress, and the challenges that, like, they will come around. I think that we will, and we need. I'm not sure just exactly like when and how.

I'm speaking which the EU's AI Act now mandates that AI generated content be clearly labeled and have watermarks right now consumers in the U.S. pay a premium to get non-watermark videos from runway and other AI companies. It's part of your business model. Would you support regulations mandating watermarks or labels that would ensure that it's identifiable in some way? It doesn't necessarily have to be a watermark. There might be another system. There are different systems.

We've experimented with a few. I think it's also a hard problem because there's ways of removing the watermarks. And so I think part of it is there's a technical challenge around how you make sure you can keep watermarking and, yeah, identifying things. And at the same time, I think there's more

of a cultural and social like watermarking of sorts. And I think we've done this before with like Photoshop. People were freaking out with the idea that you can like modify. images back in the 90s. And I think eventually Photoshop became a verb. You know, it's like, oh, you can just Photoshop things. I think we need somehow something similar here. Where, like, you're becoming a bit more skeptical about what you see because you've seen how

things can be modified. And that's just a social, social kind of like update. Sure, but is there a technological? I mean, everyone can get around everything. It's not, I don't, I'm not naive, but is there, should there be some sort of new providence system for these materials? Yeah, there's an argument. You know, I heard the other day, an interesting argument we're like, we should do the other way around, which is you should validate and watermark the real content. And I think maybe that already is happening in some way. We're like, if the amount of content given AI will expand exponentially

and like you have just way more things, then we should kind of reverse maybe the question and be like, well, what actually like is editorial and it's real? And then protect or like make sure that those are verified. And look, in a way, I think that happens already. You have like community notes and like social media. You have companies or brands or institutions that you know you're going to trust when it comes to knowing something is real or not. So in a way, I think the system is sort of heading towards that space.

Although you could say that that was when YouTube was telling all the companies, well, find it and we'll take it down. But that was all the work came from them versus YouTube, which was benefiting. You can also automate that with AI. You can use AI to help on that curation, a sort of process as well. Eventually they figured out and they're doing rather well with it. So last question in this area, last week, President Trump was asked about a video of what appeared to be a black garbage bag being thrown out the window at the White House. It really was one. It was real. But Trump said it was probably AI generated. Then he told reporters, quote, if something happens, it's really bad. Maybe I'll just blame AI. Thoughts on AI video being used as a get out of jail free card from a particularly specious politician in a way to undermine public trust in reality. Are you worried about that? That's an interesting question. I think what happens actually with those kind of like, questions is that we socially get and culturally get much more attuned to like identifying

what's real and is not. I'm probably him saying that made a bunch of people go on like debunk the video. And so it almost feels like if there's video evidence, it's easier to like debunk something versus if it's not. Like there's a lot of fake stories out there that are just words, you know, they're just like, oh, I saw something. Someone told me about it, but it's hard to debunk because you can't see it. I don't think it's going to go that far. to be honest, to rely on videos because it's very easy to see if it's real or not. I think that happened with that video. People were like, yeah, it's real. So in a way, I think it casually creates...

He wants to cause chaos everywhere he goes. But it doesn't last too long. And my feeling is that it's not going to be there for a long time, to be honest. Before we go, I want to talk briefly about the areas you're looking in the future. You've taken some first steps into gaming, which includes non-linear storytelling and robotics. I didn't even understand. I kept trying to figure out what the hell you're doing there. Talk about the potential for gen AI video in these areas

and what challenges do you need to overcome. Yeah. We've ventured to a couple of new avenues for our research and products. As you were saying, nonlinear stories. And so you can think about all the work with that until this point publicly. It has been around films and ads and videos. And that, for me, is linear media. You watch something and it looks the same every time you watch it.

Right. And we think about games, those are non-linear, right? You can move around the world in open ways. there's a lot of use cases around AI, specifically when it comes to real-time video AI that allows you to create game-like experiences. And so we are now releasing a couple of products on that front. And now, if you think about how you use these models

to create those kind of experiences, the outputs are just pixels, you're just moving images, which is not that different from how you train robots to understand the world. And so let's say you're training a robot to understand how to pick a box. What you do is you show the model or the AI

a bunch of different videos of people peeking that box. So the model understands, here's how I use my robotic arm to pick that box. Now, that's a very expensive thing because you have to create that actual peeking of the box. Cori, you can think about that extrapolated to a thousand examples and it gets really expensive really fast. And so there's use cases of hard technology

where you take the same insights of how you generate video for film, but you use it to train robots. And the robots can learn from that, like, synthetic or generated set of videos. Right. It's something kind of new, and it only was possible because the models are getting good at it. Actually, that's how I was there when ImageNet started, and that's how they did that.

Like, this is the Eiffel Tower. This is a man. This is a woman, you know, and then it iterated from there, which is the original one. Yeah, it's close to that. What are the challenges you need to overcome it? What is the biggest challenge in this area with robotics, for example? For robotics, I mean, it's just getting good high-quality data that, like, fits the specific needs of different robotics like teams. Right.

The problem, their company is solving different things. So driving cars on the one end, I don't know, there's like the robotic gaunt on other ones. There's like physical interactions in homes. I think robotics is a field is moving very fast, and part of the things that it needs to keep on moving is data, which is very hard to gather that data to train the models. Sure. Yeah, good data. And gaming is a huge industry. That's an obvious one to do it.

And in that way, what's the challenge there? I think the challenge is, there's both. There's a real-time component to it. It's like, how do you make sure these games are rendered fast at the expectations? that users are expecting for, like, a game-like. And also, I think the mechanics. So what I mean by this is, if you think about, like, a game, you're playing something that someone already created.

And so the rules are very defined, the cards are defined. In this idea of AI video, you can create worlds that no one has ever defined. And so the mechanics are open. And so the challenges there are about, well, how do you make sure people have a story to follow? And what are the rules of that world? You might not be designing every single interaction. you might be designing the way you want things to behave a role

and then have the system kind of like play it out. And so there's a lot of mechanics around the experience that are still need to figure out. So last question, you co-founded runway in 2018, which seems like a million years ago. It's hard to state how much has changed in this field since then. And if you look down the road to three, five years, what does runway look like?

Are you providing AI service to films and studios and agencies or you've expanded to become an original creative studio yourself? if that's certainly possible. What would you imagine the next step in the evolution that will allow it to come into the companies, a dozen AI companies that survives in the long run? You know, we've always thought about runway as a company focused on storytelling more than anything else.

And our goal is to make sure that there's many stories out there that we haven't heard from before because people haven't had the means to tell those stories. And we want to be the company who's creating that new kind of camera, the new device that allows you to tell those. stories. And those stories can be for like AAA filmmakers making Hollywood movies, but also could be for you in your home and for your friends. And there's many opportunities around

storytelling that I think we haven't kind of like tapping before because technology hasn't allowed us to get there. I think five years, ten years in from now, I think we want to be at the forefront of that new way of storytelling. But does that mean you become an original, I mean, you're calling yourself a camera a lot, but are you the director or the writer or the what? We're experimenting with making things in our own studio, but I think the broader opportunity is to make sure that the billions of people out there can make things. I think from a personal note, I always wanted to be a filmmaker. I grew up in Chile. I never had the means to work around film.

And I think there's many people like me back then that if they had just the right resource from a technological standpoint, maybe they're the great storytellers, the great filmmakers. I think I want to empower them more than anything else. That's what drives us. What would your movie be? I don't know. I have, look, I'm making, like, I was on the weekend just now, making a bunch of, like, short films with runway and, like, short ideas. And they're not, like, blockbuster movies. I don't want to, like, become a filmmaker and devote myself to it. But it's great to, like, I had this idea, experimented through it. I make it and I made, like, four or five minutes of that scene. And I feel great. It's amazing. It's, it's, like, painting. It's, like, writing. There are many people out there who were obsessed with AI because of that. It's a avenue. It's a way of cultivating your creative mind. That, for me, is extremely rather, more than anything else. Oh, my God. Now everyone does want to become a director and will, you know, that old thing.

Why not? I mean, if they come, they should. Some people shouldn't become a director, but... Well, we'll see. We'll see. I really appreciate it. What an interesting company you have, I really like to talk to smaller companies. I'm so tired of the large companies.

I do hope they don't run right over you. I do, because I think it's... No, we'll continue. I don't necessarily think they will because I think they're the least creative people on the planet. So in many ways, that's in your favor. Anyway, thank you so much. I really appreciate it.

Thank you. It's been great. Great chat. Today's show was produced by Christian Castor Rassell, Kateri Yocum, Michelle Eloy, Megan Bernie, and Kaelin. Special thanks to Catherine Barner and Eamon. Our engineers are Fernando Aruta and Rick Kwan, and our theme music is by Tracademics. If you're already following the show, you are David, not Goliath. If not, find a slingshot stat.

Go wherever you listen to podcasts, search for On with Caroushwisher and hit follow. Thanks for listening to On with Caraswisher from Podium Media, New York Magazine, the Vox Media Podcast Network, and us. We'll be back on Monday with more.