Runway CEO Cristobal Valenzuela is simulating the world

This episode is brought to you by Accenture. When your advertising operations fall out of sync, everything else follows. Spotify and Accenture are working together to reinvent the rhythm of ad sales. Using automation, analytics, and smarter workflows to simplify campaign delivery and access better data across the business. The result? Less time spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture.com slash Spotify. This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome? That's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online make sense? There's no place like Chrome. Check responses, setup required, compatibility and availability varies 18+. There's a chance, a high chance that I'm pretty convinced, will happen sooner than most people realize is that you might be able to stream pretty much every pixel to your screen in real time. Any sort of like world that you want, interactive or not, into any screen that you want. And that might be a different thing altogether that might not necessarily like feel similar to a film. And I think that's the thing I would say people are undervaluing and underestimating the most. Welcome to Access. The tech industry's inside conversation and part of the Vox Media Podcast Network. It sounds so official when I say it this way. We should work on a more casual intro. I don't know. Maybe it's the day I'm having. But anyway, I'm Alex. Alex Heath here. Joined, as always, by Ellis Hamburger, my lovely co-host. Ellis, what's going on, buddy? Man, I disappointed some people this past weekend. I didn't dress up as the Hamburglar. Oh, come on. You don't do that every year? No. You know, part of me wants to just give. To people what they want. But I also just really like putting on this gigantic red medieval wig. And so I got to do something for me once in a while, you know? What was your costume? Well, I dub him Lord Dingleberry. It is a like shoulder length curly red wig with like a chest plate with a roaring lion on it, like a Lannister logo or something like that. Wow. Stole all my daughter's bananas. And a Laffy Taffy for myself. So I had a little bit of a haul. Wow. Because I think that's just a little bit too much for the kids' teeth these days, you know? Are you Lord Dingleberry every Halloween? When I feel like it. Oh. I was a Jedi Master last year because my daughter was Baby Yoda. So that was absolutely too cute to pass up. That's really cute. Yeah. Fun fact about me, I have never once dressed up for Halloween. Well, that's not very fun, is it, Alex? It's not. Especially with your stature. There's just so much you could do. You could be like a golem. Yeah. Though there's been a lot of opportunities, but I'll probably do it at some point. You could be Luka Doncic next year or something with your dad bod and seven foot stature. Yeah. Yeah. Our neighborhood during Halloween just gets absolutely ransacked by kids, man. Like where we had to do the emergency DoorDash order to get more candy because we ran out at, like, 6.30. You DoorDashed some candy? You DoubleDashed? Yeah. And then the DoorDasher came and was like, this is the very last bag at CVS. I've been doing this all night in every store within, like, a five mile radius. Oh, God. So. I hope you tipped. Oh, yeah. Of course. Always tip your drivers. Yeah. No, man. I am having a day. I will tell you. I'm surprisingly, you know, now we're recording. It's in the afternoon. But I'm in New York this week for several things. But. Primarily the Sources launch party last night, which was a rager. It was a great success. But it's been it's been the struggle bus today. I read on TMZ that Shaggy made an appearance. Is that true? Or is that just. Yeah. Shaggy was there. I don't know why Shaggy was there, which I think is the mark of a good party. But he was there. He went. I think briefly he left to go have dinner with Sting. And then he came back. As one does. As one does. We had a I don't know, tennis. We had a tennis player, I guess I probably shouldn't say, but we had a pretty famous tennis player there. And I also don't really understand why he was there. But for the most part, it was all people I knew. And it was great. Co-hosted in Casa Comos by the amazing Joe Marchese, a good friend from Human Ventures. Have you had Comos tequila, Ellis? They're not paying me, but they probably should, because I do love that tequila. No, I have not. You know what I should have done? I should have just like telepresence roboted in. You should have. I mean, do they even make those iPad robots anymore? Or is that already a thing of the past? I feel like I haven't seen one in a minute. Yeah, you don't see them as much, but no, they definitely do. And I feel like with AI, I mean, I feel like you'll start seeing them everywhere. But yeah, I didn't want to celebrate access without you. So it was just a sources thing. But people were very supportive of the pod and got a lot of good feedback. So that's good. All right. Next time there could be like a little spooky room. And when you walk through it, there's one of those Google Beam chambers. And Lord Dingleberry is on the other side. That would be cool. I mean, you're the only person I know who's tried Beam. I mean, how soon are we going to be able to buy one of those or have a mobile one of those? Yeah, I think mobile is going to be a while. Beam is amazing. One of legitimately the best demos I've ever done of any piece of technology. But it's mostly like replacing giant conference setups. You'll probably experience it more like in an office than at home for the foreseeable future. It's expensive, but it's awesome. So you did another impressive demo without me, didn't you, last week? I do a lot of demos. What do you mean? With the ring? The ring thing? Yes. Sandbar. Mina, the founder, I've known for a while. We actually met when he was starting to raise for the idea several years ago. It's very cool. I guess after this episode airs, yeah, we can talk about it. But it's a ring that you can just talk to. And it's basically like your Chachapiti journal. I mean, they have their own app. Those who are watching on video, Alex does not look cool right now as he's showing what it looks like to talk to the ring. What I like about it is that it's a very clear social signal of what's happening, where a lot of these wearables are super, you know, you don't really know what's going on. You know, like the friend is like, what is that? The glasses, you don't know if someone's recording or what they're doing. It's very. Obvious when someone's using the ring the way they are. And it picks you up in a loud place. And it kind of creates this, not relationship, but modality of like using it that feels cool. It feels like very specific. And it's got a companion app. But the idea that he has is to be kind of the Switzerland of this and like let you bring your data into Chachapiti for the rest of your memory and all that. The creepiest thing about it is that it can also talk. Back to you and your voice. So you give it a few voice cues because he really wants it to be an extension of you as how he thinks about it. And like your inner dialogue. You can tell he's an introvert. He said he came up with this because he would do long walks and wanted a way to like reflect on his walks. And in the app, you can, you know, if you have your AirPods in, you can ask it things and it will respond in your voice. It's pretty freaky. I mean, I feel like. That just makes me believe even more strongly that Sam and Johnny are making some type of AirPods competitor because why have two devices when you could have one? I like the ring form factor, but, you know. The simplicity of the ring is nice because it's very clear what's happening. There's no hiding it. And I do like startups that are thinking about the actual kind of cultural societal implications of the hardware that they're building, which I mean, it definitely does. So, yeah. No, shout out to that. That project's really interesting. And, yeah. Otherwise, I've been getting a little bit of the New York tech scene this week. Tonight, we're recording this on election day here in New York. Tonight, I'm going to head to the Calche headquarters to check out. They're going to do like a little election wagering thing, have some like trade stations pulled up. And that's a super interesting. That's a super interesting company. Tarek, the founder, I think is pretty smart. I am sending a vibe vote over to Zoran purely because I read that he likes Mission Impossible Ghost Protocol, which I think is an incredible underrated movie. And he also likes The Strokes. So, that's all it takes to win me over. Wow, you're pretty easy. I didn't realize that. And also affordable housing. No, his ads are everywhere. I mean, it was in Brooklyn. I was in Brooklyn over the weekend, which, by the way, being in New York for the marathon weekend, it's the best weekend in New York. And I always loved that. Better than SantaCon? No. But, yeah, people were just canvassing like crazy for him. I should check Calche. I bet he's like way up on projected election. He was like over 80%, wasn't he? Yeah. Yeah. Yeah. You've been thinking about this whole kind of prediction markets gambling thing, right? Didn't you have something you wanted to talk about? - Yeah, I mean, I feel like I was listening to Derek Thompson's podcast called Plain English, which is funny. I find it not to be necessarily the most plain English when it is so statistics driven, but I love that too. But anyway, he was talking to a guy about just kind of the ascendance of instant sports gambling here in the United States. And it just kind of hit me that like every conversation we have is kind of the same conversation, whether it is about short form video slop or algorithmic optimization or in shitification. I feel like there's this really big debate that tech in some ways started with its avant-garde approach to behavior maximization and money extraction that we are just being preyed upon by all these services using all of the great technologies that we have. and tricks and tools of like the last 15 years of technology to really certainly give us more of what we click and what we want. And I don't know, it's just pretty scary to me. But yeah, I texted my good friend Nathan, who was the sociologist at Snapchat for a long time with me and has informed a lot of my views. And he was saying that people treat the outcomes of algorithmic systems as like the weather or the whips. And I think this is something that needs to be talked about more is that it's not just the free will of these platforms and algorithms to give people more of what they want. It's an actual choice. And it also collided with this piece by Max Reed, hashtag ReedMax, where he was citing this piece about how Section 230, and I'm curious if you have any thoughts on this, so often talked about as this like blanket protection for for tech platforms, for anything that's on their platform, shouldn't really correspond to when they actively promote things. It's one thing to be substack and host a blog that some people may disagree with. It's another to actively promote it in the feed. And I thought that was just like such a simple, fresh distinction that I feel like needs to be talked more about. So yeah, that's the jumble of things going on in my mind right now. Did that make any sense? Or are we just going to cut it off? No, no, no, no. I think I get where you're picking up or laying down, whatever the phrase is. God, I'm so tired. You're picking up what I'm laying down. Yeah, yeah, yeah. But are you saying that Section 230 should apply to like prediction markets? Or are you basically saying that you think prediction markets are a justification? I'm saying that extraction is at an all-time high, and it may cause the next recession. Well, yeah. It is making people a bit more lonely. And less happy as a result, which means that they're on their couch more, which means that they're betting on more sports because they got a $500 free credit. And then they open Instagram, and they have gotten hit by the same ad 17 times, which is the exact number of times it takes to get you to insta-buy something on Instagram. Oh, not me. I'm a one-shotter. You're a one-shotter? No, I totally buy that. I mean, on the prediction markets piece, I think Tarek's going to call on the show. I want to talk to him about this. But I do think it's smart to let people see the wisdom of the crowd, so to speak, if that is actually what it is. I think that's a good idea. It feels very, like, democratic. But the gambling piece of it is obviously, you know, something to be concerned about. I'm okay with it. But I do, yeah, I do hear you, though. Like, we are getting extracted from a lot right now as a society. Yeah. I know you're a gambling man who appears to have control over their impulses. Yeah, it depends how much tequila I had. I don't know if last night was a good example of that. Shout out Joe again. And Comos. Well, we've been talking this whole time, and we haven't intro'd our guest this week, who's a great one. Cristobal Valenzuela, the CEO of Runway, a very popular, fast-growing AI production tool. It is freaking out and changing Hollywood on an hourly basis. It comes up quite a bit in my discussions just with Hollywood folks. I have known Cristobal for a while. I think he's always really sharp on a lot of these issues and is not, you know, afraid to speak his mind, which he certainly did in our convo. I don't know. What did you make of Cristobal? That was your first time meeting him, right? It was. Game recognized game. He's a great writer. This is. And has been articulating a lot of. Very complex ideas in simple, effective ways for a long time. I think many founders, as you'll hear, look to him kind of as a luminary for self-expression, I would say, within tech, whether it is, you know, being one of the first to kind of establish themselves as a research lab, which is always a good vibe for recruiting and your societal benefit goals, if you want to imply those. And he's a big thinker. Yeah. And he also comes from an arts background, which I like. So he has an interesting take on, you know, the whole tools versus cost-cutting debate when it comes to our new AI services. Yeah. And is very big into world models, which we get into, but is a topic I think you and I will be talking about a lot more in the coming months. And people broadly will start hearing about a lot more as everyone from Google to new startups to middle stage growth startups like Runway are really diving. And I liked Chris's perspective on, like, this may just be an entirely new medium that's happening where you're going to be able to generate pixels on the fly in 3D with a sense of space and depth and awareness of physics and all that stuff. And what that does for entertainment. Is it a new thing? Is it the next phase of video games, movies, both? Probably all the above. Pretty wild times we live in, man. Hashtag we'll just have to wait and see. With that. With that, if you're Alex, go grab a Gatorade. If you're a listener, stay tuned for Chris Valenzuela. Spent on operations, more time connecting brands with the moments and fandoms that matter most. Learn more at Accenture dot com slash Spotify. You think, you know, a browser, but Gemini and Chrome, that's new. It can help you with practically anything on the Web, like restoring a vintage motorcycle from a 50 page restoration block or finally break down that long article you've had open for weeks. Check responses set up required compatibility and availability. Various 18 plus. Chris, great to have you on the show. You and I last spoke for God. What was it called? The worlds? Is that what it is? Yes. Runway worlds. Game worlds, game worlds, which I want to get into. But first, can you share with the listeners a little bit about you and your background and how you came to be doing doing what you're doing now? Well, it all started, I would say, like almost ten years ago. I fell into a rabbit hole of deep learning at the time, came across some interesting research papers that I think were kind of the beginning of everything we've seen today, and I've kind of became obsessed with this idea that you can take these algorithms and these approaches to AI and use them to create not only language, but also pixels and videos and images. And so I've been spending the last, I would say, eight years just on this. Deep on that rabbit hole. Came to NYU to study this, met my co-founders, and we started the company like seven years ago. So it's been perhaps the same journey and obsession we've had back then, but now with a bit more scale and attention. Yeah. Give me a sense of Runway's scale right now. You're a private company, but, you know, walk us through everything. Share everything, please. Valuation, money raised, revenue, customers. Yeah. Okay, let's do it. So I guess scale of the business, I think we have like millions of users. We're working with everything from the largest and biggest and best studios in the world to a lot of agencies and brands. We have a lot of game studios using Runway. We have architects using Runway, designers, filmmakers, and a lot of just consumers and creators. It's a platform. It's a tool. It's a model. It allows you to do everything from like simple, you know, images. It allows you to do anything in videos to way more complex, cinematic, like 4K footage. And so the business is like growing fast. It's been growing fast for the last couple of, I would say, two years. We're around 120 people. Most of that, the team is research, engineering. So we have an office in both New York and SF and a small office, also a research lab in London. And so, yeah, I think, I mean, the models are getting to. Up on where like, they're pretty much usable in a bunch of different domains. And we're like, I would say venturing into, into many of those. Yeah. And valuation, you're well past unicorn. Are you technically decacorn? Where are we at there? The last public valuation was around like $3 billion. That was like a year ago. We've raised, since we started the company, on average every like 12 months or so. So hopefully we'll get some news very soon. And how much have you raised so far? Alex, take it easy, brother. We've been here for less than four minutes. The reason I'm asking is I want to also compare this to the state of Hollywood. And when we talk about Hollywood, and I just think it's really interesting how you all as a company are growing and people probably don't actually understand how big you are. And that's why I'm asking. Yeah, no, totally. Happy to answer as much as I can. We've raised like half a billion by now considering our last round. Again, we started in like 2018. So the company is called Runway ML technically. So we used to call it ML back then. And the rounds and the sizes and the valuations were very different if you were building ML products back then. And I think we've been consistently just, I would say, obsessed with making sure that we build actual things that can translate into customer and usage and value. And yeah, I mean, in Hollywood, I think we have really good adoption. And so it's kind of the first, I would say, stepping stone for us is making sure that media and creators can actually use this in consistent ways. Okay, I'm turning the grill temperature down. Alice, I think you don't go for it. Yeah, so speaking of the ML to AI transition, I mean, I feel like we're all looking for labels for what we're making. And I know you've written before that this kind of represents a new medium or even a new kind of camera. Is it video that we're even still doing? How would you describe it with what you're building and where it's going? Yeah, so I guess for context, maybe for listeners who are not kind of like fully aware of that idea, I've always thought of, we always have thought about AI and specifically what we're doing at Runway as literally building a new medium. And so when I mean building a new medium, I think it's important to refer to things we've seen in the past that have had that similar impact. And the most recent one is perhaps, like as you were saying, the camera. The camera created entire industries. It's created an entire new art form, which is filmmaking and cinema and the birth of photography and created a whole array of social and cultural changes. I do think AI is moving towards a space where we're understanding what you can do with these models in ways that are not equally or similar necessarily to film. Like you can borrow like the language and the grammar of film in the same way that cameras borrow the language and the grammar of like painting, but they're ultimately different things. I think AI, the ability for AI to basically simulate, because these are ultimately these video models are simulation systems. They can simulate anything technically, in the world and how things should behave in the world. You can have agents deploy and simulate how agents interact with things in the world. That on its own, if you just think about it as a creative tool, it's a new medium. It doesn't necessarily operate with the same like principles and affordances of like film and cinema. It has a set of different like ways of working. And you can see this with like people who are already very deep into AI. The things they've created, the way they use it, how they feel is just different altogether. And I think like one of the things I think investors and I would say the public is like always trying to find is like, what's the label for this? Like, how do you call it? My belief is like, we're not sure how to call it yet because we're living through the thing right now. So it's hard to find a name for it. Like photography didn't have a name until like 20 years of experimentation happened. And like, we come up with this idea of photography. I think we'll eventually we'll collectively figure out a name. We're calling it like world models. These are like systems that can, that's like, they're from a research perspective, but I would say from a medium perspective, we'll collectively have to figure out how to label this. - Chris, I believe it's called slop. Come on now. - Slop, I think, well, I think it's not. Look, I think like film and photography were slop like 150 years ago. People made fun of it because it wasn't, I think the mistake was to make painters judge photography and painters were like, look, like this daguerreotype that you just showed me, it's like really bad. Like I can paint this monster. I can paint this monster piece, which is, which is fair, but those are not equal like comparisons of sorts. And I think we're now doing kind of this thing where like someone is like, look, I'm a filmmaker. I've done all this thing and you're making this. And I think it's, you're right. But also you're missing the point, which is like, these are not equal things. So kind of like takes slop as a, actually as a great compliment. It means that we're heading to something totally different. - The labeling conversation is so interesting to me. And I am honestly a little triggered by the camera reframe. Cause that's what we tried to do. At Snapchat for many years. And I think at the end of the day, even when I talk to clients today, I'm like, well, the simplest answer is just to use the language of what your customer wants. And how are you talking about it with your customers these days, especially the studios? Do they think of this as a measure to kind of cut costs and make things more efficient? Or are they thinking about the possibilities that maybe only AI could afford maybe in terms of like, you know, like the marketing results that maybe the artist wouldn't have even expected. How do you talk about it with them? What do they say that they want? - I think sometimes they don't really know what they want, to be honest. And I think part of it is there's, there's a gap in AI literacy that needs to be closed. I would argue that one of the biggest bottlenecks in AI these days is how people fully understand how these things work. And you can see this like in like enterprise adoption and like people using AI, like the models can do very sufficient, really good economic work. The bottleneck is not the models. It's like the people within specific companies or within teams that know how to do it and how to use them and how to leverage them. And understanding that the way you've done things in the past might not necessarily adjust well. And so a lot of what we do is we, for companies that we work like really closely with, we have teams of people that go and work with you and they see it and like explain to you kind of the different things that you, these models do and how they work and how they operate that are not based on how you've done things in the past. And that's kind of the beauty of it. It's kind of the mental models, the human mental models are the ones that we need to update these days and not the models themselves. Like we get so many people who ask us about things they want to see in Runway and Runway can already do those things for like the last like six to, you know? And so I think like that deployment of people within enterprises and companies is going to help just clients and folks figure out much better how to get the most value out of it. - Isn't that called a forward deployed engineer? Isn't that the new like buzz word in the labs? - Yeah. Alex has been spending too much time in San Francisco. - Yeah, I've been spending way too much time. - It's funny, we started, so we hire our first, we call them forward deployed artists like three years ago and it became like a, we didn't have a name for it. So we call them, we call them workflow architects and it was basically like come up and like help us like build this inside, specific partners and I think yeah now it's becoming some sort of like a job and a role of its own which is which is I think knitted I think again we sometimes forget that even for like the internet people have to be taught how to use it you know it's like national campaigns around how to send an email and and I think it's it's new you have to like figure out how to like how it all works and I think we're transitioning through some of that same thing kind of right now something you all did really recently that I thought was smart on that note was the apps concept is that what it's called runway yeah correct so basically because it is a when you look at runway it's I mean it's it's a complex tool like I would not know how to use it I don't edit video but I would imagine even if even if you've come up with premiere or final cover or whatever you look at runway it's probably pretty foreign a lot of the concepts talk about that apps launch and why you did that is it for the reason you're saying which is just people have a hard time understanding what they can even do with us yeah so you're absolutely right apps is so you can think about like raw models it's like an intermediate layer like you're you're obstructing and you're you're building some sort of interface to tap into this system that can join anything and the most convenient ways you use prompts or images or different things but then for a lot of newcomers for brands for designers for agencies there's specific things they have in mind they want to make and sometimes learning the structure of the prompts and how to best get the outputs is complex and so we build an obstruction system on top which is mostly focused on what do you want right what is it that you want it's a customer as a user if I'm a brand designer if I'm working on I don't know a concept video or I'm making some visual effects shots I know exactly the thing I want and so you have an app for that and you click on the app and the model the app basically run it tells you like what you need to input like references or images and they're kind of like it all cooks behind the scenes and it shows you like the output and it's one of our best and like most fastest growing products that we've released recently you you tweeted one where where the app changed the time of day like it changed the lighting and that was wild to me you called it like a time machine it's yeah it's in it's um like those things are have been there but now figuring out how to best present them to people it's like a ui an obstruction challenge of like again like think about that like you're you're translating uh like a specific like shot a real scene of like day into night like out of the box you're like oh my god I don't know how to do this and you're like oh my god I don't know how to do this and you're like oh my god I don't know how to do this Outside of AI, we're talking about, if you want to do that, we're talking about and thousands of dollars from like a professional studio to just get to somehow similar results. Now you can do it in like 10, literally 12. And usually day to night looks like absolute shit. Yeah, it looks, I mean, when studios try to do it traditionally. Yeah, I mean, probably. Because it's hard to do. It's a very hard task. And for the model to do it, for Runway to do it in like 12 seconds feels like incredibly powerful. But I don't think enough people realize that there's these powerful systems and these powerful machines that you can do this stuff. And so again, I have to just go and show you. I have to go and like explain to you how it works. I think part of what I would say has prevented some of that like more mainstream like adoption is the curse of the prompt. It's what I call it, where prompts have taken over AI. You know, if you think about AI as an industry, it's like a whole economy of certs. There's all different sort of obligations. But I would say still. The biggest narrative I see around is people think of AI as just a prompt. It's a prompt. You type in something and you get something out. It's way more than that. You can do way more sophisticated stuff. And so you need to like, again, hopefully explain and tell people how it all works. I'm curious. Were you excited about this app's idea before it happened? Because I feel like a lot of founders, especially when they're making a tool that could ostensibly use for anything, they see putting those frames or use cases or templates in the UI as like a tool that could be used for anything. And they say, oh, well, if we only say we could do a handful of things, then how are people going to know that they can use it for anything? And it's a debate you often see. Yeah, I think you're right. It depends on like exactly what you're focusing on. So we're not focusing on apps alone. Like we have different levels of like, you know, complexity. We have a thing called workflows. So workflows is like for much more advanced users. I tweeted like a recent kind of short film someone made just with this workflow interface. It's a node-based graph. It's kind of like a system for creating really complex things. And so really the models and runways and idea is that you can generalize the models to basically tackle any single hopefully problem you want to solve. Some of them are easier. Some of them are a bit more complex. But you can lean and probably solve for like all of them if you want. So apps is like I would say one offering. It's not the offering that we're like putting out there. So I didn't get to go to the most recent one. But last year I went to the festival you guys had. And it was a really interesting experience seeing all the films. I think I went to one the year before or a couple of years before. How many years have you been doing those now actually? So we've been hosting our film festival for three years now. Three years. Okay. So I went to the first one and the second one and I missed this last one. So you went to the first one. That's like OG. I went to the first one. Yeah. OG. Well, you and I got coffee in LA I think around that time. And it was you all were just a baby, not even a unicorn at that point. Right. But it was. It was really amazing to see the progress from even just, you know, year one to year two. Right. It was pretty dramatic. I assume it was even more dramatic this year. But I got to be honest, like when I was there, I was marveling at that. But I was also kind of I left feeling like a little empty, you know, like when you leave one battle after another, which I've now seen in theaters three times. Like you're very alive. Right. It's an amazing film. It's very human. And I missed that in the shorts, even though they were very creative. They looked awesome. The stories were, you know, whatever. Shorts can be rough on that. But there were good stories. But I, man, I just I left a little empty. Do you hear that from people still when they see a bunch of AI content like that? I think at the end, it's all about the story. Like, I don't think people watch things because of how they were made. Like one battle after the other is just great storytelling. It's one of the greatest filmmakers like we've ever seen. Alive, like makes one of the biggest films he's done. The reason you watch it just because it's a good story, it's good acting, it's just group script. I think that doesn't really change the nature of like whatever you use to make it. If you use AI, if you use cameras, if you use whatever like technique. I think three years ago when we started the festival, we wanted to basically provide a venue and a space for people that were a different tier of new technologies to expose themselves to like this form and like sharing what they're doing. I think that doesn't really change the nature of what they've made. I would say the transition from the last three years has been from the technical like obsession of how things can work and how you can experiment and get to new aesthetics and new like to storytelling. Like the winners of this year's festival, they won because of the story. Like the point was at a point you kind of forget that you're seeing is generated. That is like the last frontier is you forget you're seeing something that was made with AI. So the winner of this year's festival is now producing like a big like film. And he wasn't even a filmmaker. He was a musician who had this idea around like a story he wanted to tell. He worked on it for like a year. He won. It was the like favorite of everyone who was in the festival. And now he wrote a project. He's actually making like new stuff like with actual production budget behind it. I think for to grab that attention and to make sure like it that that works, you have to have a story behind. And I think that that doesn't change regardless of like what you're using, you know, is it fair for Alex to compare AI video content to Paul Thomas Anderson? Uh, well, I think that's like, again, if you think about it as a medium, it's that you're comparing like the Lumiere brothers to like Rembrandt, you know, like they're they're different things, like, of course, they're great painters. But this is different, of course, like PTA is like the greatest one of the greatest in the world, but this is like a new thing that's just it's like a year and a half old. So it's it's I think it's it's not probably fully fair because they're different things. But I mean, at the end of the day, for viewers, it might be just like the same. So but I'm convinced that models will get better. Quality will get better. The stuff that's coming out right now, it's moving away from the technical novelty and into like storytory real territory, which is the most interesting one for me. I'm in the mood to make fun of Alex today, so I'm going to continue with this line of thought. And an impression. So I just saw one battle after the other, like three times and like everything online is so slop. I've got a mustache now. I saw it at the Vista Theater. OK, I saw it in VistaVision, Flex. I think Chris is right, though. I mean, there's a pretty big difference between using a new tool to like create an impressive explosion, if you will, or to execute your authorial intent with some potentially unexpected results that. Might delight you. Like, I think people can can see that. And I mean, this is how I felt at like the OpenAI Sora. Event I went to a year or so ago, Braindead, like everything felt very auteur driven in terms of like the experimental filmmaking and storytelling. And it is definitely a different aesthetic. I feel like that you feel with AI. I think the only time I really worry is when the AI is just meant to purely dazzle and pacify us as we scroll where there isn't much authorial intent versus using it as a tool, if you're an artist who just may not have the skills. Like, I think there is a distinction there. I agree. And I think like that's mostly like perhaps different product experiences. Like these models are generally not from like big enough that they can be using all sorts of different ways. Like if you want to build an infinite slot machine, you can like we're not building that like it's a different choice, like. Some companies want to go there. We think it's this is much more of a creative system of sorts and we want to empower it. But that probably doesn't mean that others can do it. And I think you start to see this right now happening in the market where there's like slot machines on the one hand, like just grabbing attention and trying to like optimize engagement. I think what we're doing is a bit more thoughtful because I think we believe of these models and these tools as eventually like creative tools for creative people, for really anyone. And that informs our ethos and our vision and our product experience altogether. I'm curious if you think movies and TVs will go the way of music in the sense that I saw I think it was a billboard headline recently that kind of blew me away where they said that at least one purely AI generated artist is now charting every week. Maybe you share that. I don't remember. But I saw and I was like, whoa, that's crazy. And then I also thought ahead of this convo, is that going to happen to movies and TV? Like, are we going to have like fully generated stuff that is like performing at the box office? Or do you think it's always going to be hybrid? So I think the way we make films today, if you just think about like media, traditional media might be there still in the next like 10 years, you know, in the same way that we have like theater plays in Broadway, like those things are still there after like movies and television was invented. So you might still do things the way PTA does that does it. And I think that's great. I think ultimately there's a chance, a high chance that I'm pretty convinced will happen sooner than most people realize is that you might be able to stream pretty much every pixel to your screen in real time. And so you're going to eventually have this system that renders any sort of like world that you want, interactive or not, into any screen that you want. Maybe hybrids where you can combine traditional filmmaking with this new thing but the fact that you can render pixels on a real-time basis kind of changes the dynamic the economics of like streaming services of platforms of ip um and i think that's the that's the thing i would say people are undervaluing underestimating the most like the moment you can stream pixels on a real-time basis of anything you want it's like different game altogether for for media but for also a bunch of different industries as well how soon are we out from that it's it's here i don't think we're like that far i mean it's probably there on the tech side like i'm sure you guys have that internally but i'm just saying it's not productized like at scale i'm saying when is it a productized at scale thing it's not i think the product part of it might come just from like optimization parts like it's very expensive still but if you want to figure out like what's the sweet spot of like quality and like speed and if you want to maybe constrain some of the generations to specific applications like i've tested it it just feels dramatically different you know from anything you experienced before it feels it breaks your brain a little bit you're not sure that what you're watching and the fact that you can interact with it kind of changes a little bit of the paradigm of how we've used ai before and even if we're like a few months away like probably might be the one of the biggest and i'm of course i'm biased but of course i would say it's one of perhaps the biggest like next frontiers of ai is that you can simulate and generate pixels on a real-time basis in every screen out there i am obsessed with world models i do agree with you i think it's a thing everyone's going to be talking about especially next year um but before we go to that uh ellis and i both live in la um the mood is i would say i don't know how you would describe it ellis it's like not great uh you know it's kind of grim it's like a mix of not great and like highly buzzed on matcha i would say which is not a good combo there's a lot of like out of work writers at all time uh in ellis's neighborhood like on a tuesday afternoon you know like i bring the tissues and they're all gone by the time i leave but do you think that has something to do with ai or is something else structurally no i don't i think it has a lot to do with the government and the tax incentives and a lot of practical stuff that people don't talk about but yes there's already this mood and then there is you know i was just having lunch with like a talent agent last week and he's like this is all anyone can talk about right is ai it's going to take all our jobs everyone's already feels like their jobs are being taken i don't really see it as that two-dimensional i'm assuming you don't either but yeah i mean address this like topic of like is hollywood fucked or not um well i think you actually said something interesting which is like a lot of the causes for why hollywood is like in this spot right now has to do with something people don't want to speak about uh which is like all the things people know are structurally like not really working but they're hard to address which is like tax incentives the strikes there's like projects being done overseas there's like all the things that people know are structurally like not really working but they're all the things that are just like outside of ai it's like the you know it's like similar to the water problem with with ai you know ai i don't know if you guys have kind of tracked some of the narrative it's like people who are obsessed with like consumption of water uh of ai and like if you plot that against like every other like water like consumption of water in other industries like golf courses consume like six percent of the water in the u.s and data center is like less than one percent but like fighting golf courses feels like it's like it's like it's like it's like it's less interesting than like being against the eye yeah um and i think in in media i felt a little bit the same or like there's all these structural things about how the industry has evolved over the last 100 years there was transition from needle to analog to analog to needle sorry then like streaming then like then the cost of things just go to too too high up then like tax things there's all these things are structural to it and then maybe two percent is ai but like fighting the structural things it's like too complex just figure out the you know the easiest like more interesting one to say i yeah that doesn't go to say that like it doesn't have an impact i think it does have an impact and it will change it but i would say what's happening with i was in la like last week uh what's happening there uh has way more to do with other factors than ai itself ai is becoming just like a scapegoat of sorts um but but maybe going back to i guess the point you're trying to make uh yeah i do think ai will radically change a lot of jobs and that's that's that's an obvious thing i mean you're raised within media there's a lot of things that are very rote and repetitive that you can totally automate and i think that focus shouldn't be on the like on the jobs themselves it should be on how do you upgrade or upskill the people doing those jobs um rather than like oh we need to protect you know that particular thing because the thing might change and it's okay it's part of the process this episode is brought to you by accenture when your advertising operations fall out of sync everything else follows spotify and accenture are working together to reinvent the rhythm of ad sales using automation analytics and smarter workflows to simplify campaign delivery and access better data across the business the result less time spent on operations more time connecting brands with the moments and fandoms that matter most learn more at accenture.com spotify this episode is brought to you by google chrome you think you know a browser but gemini and chrome that's new it can help you with practically anything on the web like restoring a vintage motorcycle from a 50 page restoration block or finally break down that long article you've had open for weeks gemini and chrome is here for it ready to make anything online make sense there's no place like chrome check responses set up required compatibility and availability varies 18 plus it's tough isn't it i feel like it's very easy to focus on job losses as kind of this de facto bad uh but you know you look at coca-cola just released this new ai holiday ad made with ai and this is a company that maybe would have paid an agency to do that uh in the past that employed a lot of editors and people and coke's making 10 billion in profits uh last year you know so they have the money to spend on the agency and the editors but at the same time and i know you you went to nyu tish and maybe you know you have some peers i'd be interested to hear what they think about that any of the editors or visual effects people that might have worked on that like a lot of them probably wish they could be directors or creators of some kind but because of the bureaucracy because of the if you watch the the the show the studio the existing barriers to entry for becoming an actual creative with the resources to make a video happen now you potentially can and i feel like people are talking less about that stuff i think uh really it's just as relevant the the winner of here's a perfect example he's a musician he got into like film now not by this additional route just by making something that feels great and people like want to watch it and like he got approached by people because he made something he couldn't have done before that i think it's like the greatest proof of how transformative and how job new jobs are going to be created and are created by by i would say i mean i certainly feel that way like when i've been on internal marketing teams in the past they're like oh you're just the copywriter you're just the copywriter like stick to the words bucko i'm like yo i was very nearly a film major at michigan i know more about movies than you do i can talk eisenstein i don't use premiere but in the future i'm gonna be able to make some pretty sick videos right now's the time to now's the time to the revenge to revenge yourself now you have all the tools exactly so uh alex you said you said world models are like the most interesting thing for you or like you believe in them long term uh i just think it's something everyone's gonna start talking about i think everyone's building towards that like you said um why is everyone obsessed with this idea of world models right now is it because it's how you get closer to agi is it because it makes ai physical all the above like why do you think everyone's starting to buzz about it and why are you excited about it we've been excited about it for the last like four or three years i would say we built this general framework that we call general world models um like and we made a video about the whole idea like three years ago and it's become i would say our like you know vision long term which is if you think about video models video models again are simulating the physical world and as they learn about the world they understand like how things in their world behave they're they're understanding like physics or understanding movements or understanding action and consequence and then you can technically have agents being deployed in that simulation interact with the world itself move things around um agents in the simulation now i've got the chills yeah i mean everything's just a simulation of sorts um i'm the resident matrix fan here you start to see that the models learn the about the world in ways that language models are not really learning about the world um and i think language models have gotten somehow bottlenecked but what they can scale and what they can do and a big way of unlocking them is to give them imagination is to give them like the visual capacity to simulate things like and i think world models is like a catch-all term that describes this idea that you can yeah have ai systems that can understand the world how the world works and you can deploy agents within the world you can have consequences and actions and you can see them but also the model can see them and the i think the reason why people are getting interested and more excited about it is because eventually that approach has infinite applications within pretty much Everything you can imagine. If you can simulate how things behave in the world, then you can think about like applications in like self-driving cars. You can simulate and improve like performances of like how you create algorithms for self-driving, not by deploying cars, but simulating them, simulating how things should behave. Media and entertainment, you can simulate storytelling. You can simulate like scenes. You can simulate specific things. You can simulate gameplays. You can even go deeper than that. You can simulate like interfaces, which is something I think it's becoming more real. And so I think people are getting excited because it's becoming kind of like a big opportunity for more frontier research and at the same time, a huge market opportunity as well. So you said the word simulate. I know historically with LLMs, people talked about probabilities and predicting the next word. Is that different in your mind, the simulation versus the prediction? Or is that kind of just. Is that just the word for video or world models that's being used? I think simulation goes a little bit like maybe deeper. And so there's traditional simulation software. So if you think about, I know you're building a plane and you want to simulate like aerodynamics, there's a software you might use that basically you can put a geometry on an object and have like airflow and you can simulate how things will react. If you're, I don't know, doing life science research and you want to see how different compounds will interact, you need to simulate them, right? If you're building like. I don't know, some robotic experiments, you might need to simulate interactions within an environment. And these are all very hard things to do. You've coded them by hand. You've kind of are looking at the world the way we have done simulation traditionally. So you have people looking at the world and trying to decode those things of the world in formulas and with human biases of how things we think work. But really, what you want to do in like this new approach of AI and simulation and world systems and world modeling is. You don't want to describe how things work. You just want to show the models the world and have the models understand those things in the world. And this is similar to how we learn. Like I can show you, I can drop something from my hand and you can kind of tell me what might happen, what you think might happen. You can kind of imagine it and simulate it in your brain. I'm probably sure you can describe the formula and the trajectory of that object, right? And so we're taking that approach with world models, which is for models to understand how the world works instead of trying to be. All this like precise and have all these inductive biases on how we think the world works. You should just show enough data and the models will become generalizable machines. They can simulate those same things that we used to simulate before, but not in those traditional methods. You kind of like leapfrog them all together, which is wild to think that they can do and kind of like they do and it works. I'm just like so curious, have the models incidentally discovered any novel Fibonacci sequences about the universe? That people haven't noticed or discovered yet? I mean, it's just so fascinating. I mean, it's like if they're not built on these formulas per se and they're just kind of somehow doing it, it's like, are there going to be some Stephen Hawking-esque scientific discoveries to come? Probably. I think that's the bad. I think we're getting to a point where models are going to start doing really good scientific and research work. I think that's happening slowly and probably we'll see it more. What are the societal, cultural implications? What are the implications of world models as you think through it on, I mean, I do think I agree. It's probably a new medium. It's probably not quite right to call it gaming or movies. It's like some kind of combination. I'm sure someone will come up with a name, but beyond that, like what are the, what are the implications you're thinking through of how we're going to be all dealing with this? I don't know. I mean, the fact that you can stream pixels in real time and simulate every possible world and interaction feels wild. It feels strange. It feels definitely, there's going to be like, a lot of things we have might need to like prevent and this is not only runway. This is like a society as a whole. Like there's a lot of things that are hard to like fully understand until you've deployed them at scale. I do feel they're going to become the backbone of how we do a lot of economic work. And I don't think it's for one company to decide. I'm not saying like, Hey, we're building world models. We've been working it for X amount of years. We're going to decide how the cultural impact of this is. You're, I think it's. It's for society to decide at large. Um, I do feel like stuff like the film festival and other venues where you can highlight new generations might help inform those transitions. And maybe we need to do more of that. Just hear a bit more about the people at the frontier who are experimenting and kind of hear what they have to say. It's going to be very interesting what happens over the next several, several years. I'm a gamer and I follow closely all of the news about the ballooning costs of game development. Now, you know. It's taking rockstar games more than 10 years to make the next grand theft auto. It seems clear to me that they're going to be a lot more indie developers now, which is really exciting because the tools for creating these rich worlds are no longer, uh, such a barrier. But I think there's also this risk that things just get kind of bland and vanilla. Uh, I remember playing the first no man's sky game. I don't know if you're familiar with that. Of course they're like, it's got thousands of planets and they're procedurally generated and then you go to one and there's nothing. And so I wonder what that new job is going to look like. People don't talk about prompt engineers anymore, but I think as long as there's some intent and some human ingenuity that's being put, uh, into these requests, you know, for new worlds, new content, new textures, new, uh, new ships, new costumes, then it's going to continue to be interesting and we'll just have more good stuff. In most film projects, you have like this, this Bible of search, which is a book that basically. Defines the world that you're like doing or building. It has art references, like, uh, like script references, just description of characters like that is for, for the crew becomes like the reference point on how you make a lot of things like imagine feeding that Bible book to the model and being like, okay, here's the world they want to hear the rules in which I want you to like operate. And instead of describing every single interaction and every single thing, you might just provide the right references. You're telling the model here's like the world they want to see. I want you to envision with the references I've come up with now that you need aspect then of like how you make it. It's not the engine that's making it. It's like, what are the things, what are the references? What are you trying to do? And we just come back to where we're maybe speaking before, which is at the end of the day, it's about the story. It has about like, what do you want to tell the world? We, I get this constant feedback from people where they're just getting to AI where like you show them all these amazing videos and images and they're amazed by it. And then they, you give them the, you give them runway and give them the access to all of it. And the first thing they have, they struggle with is like, what do I make? You know, you've handled them like the most powerful machine they've ever had and they're like, uh, a cat and a dog, you know, they're like, and then you start to realize that the bottleneck is like, not the way you make it. It's like what you want to make. And that's not going to change regardless of tech. I think there's going to be a new golden age for idea guys like me who are jacks. You know, I mean, it's not like things are perfect with movies or certainly with Netflix right now where everything is just so blonde generic and that stuff isn't even AI made yet. I think part of it is just very formulaic, which is like, you know how, you know, you know how it works and you just want to repeat the formula. I think that we're tired of that. But to maybe connect this to what we were talking about earlier too. Do you have advice? Like when someone comes to you and is like, are you here? You're about someone who works in a production house that's adopting runway and, or they freelance for them and they're like, okay, my hours are going down now because the amount of time that it takes to make things is being, you know, rapidly decreased by tools like runway. And this whole kind of gig freelance economy that is traditionally powered entertainment and gaming, like it's really hurting as a result of like a lot of this change that's happening. What advice? Do you have for those people that I think rightfully do then have angst about AI and taking jobs and it goes into the water debate and all that, but like fundamentally what's happening is like work that people had is going away. Yeah, I think that's a fair point. I do agree. Some jobs are going to go away, but also a little more are going to be created. I think the advice I have is look, there's, this is like, if you just think, put things in perspective, video models in the way we know them today are like a year and a half old. So you're like. At the beginning of one of the biggest and most important like technological revolutions in your industry, you have the choice to think about this as a thing you will never use or like the thing you just have to learn how to use. And so the only advice I can consistently give people is like, I would choose just like go and use it, experiment with it. Like the rules and how you use it, the combinations, the workflows are still there to be invented. Like people are figuring out how to. Best make things with it and you can be the one like pioneering a lot of like techniques. Not enough people are realizing that. And I think when people ask me like, how is runway going to be used with Avid or like with their like traditional way of working? I'm like, it's not, it's like, it's, it's, you can have fax machines or you can have the internet the way you're going to transition from fax machine to the internet. It's not by connecting the fax machine to the internet. It's just like learning how to use this new thing, you know? So your advice is go. use my app essentially who you say I just I mean of Of course, you just run away, but go use, like, by now, if you haven't used any, like, AI model, like, I think you're missing, like, I'll be like, come on, like, these things are insane. They're so powerful. I think it's true. Yeah, you have to. But, I mean, the outcome that everyone hopes for that would be the best outcome is, like, people can do more work and people have more opportunity with these tools, right? I think everyone who's building these models, for the most part, thinks that way or tricks themselves sometimes into thinking that. And I'm not saying I disagree with that. I'm just saying, like, that needs to happen for all this to work out, for people to not really hate AI, especially in creative fields. Like, they need to get work and they need to not be working less because of AI. And there is this, I think we're just in this kind of messy transition period where people are adapting or not, and they're not sure what's on the other side. And I think that's the thing. That's the angst that I get asked about all the time. My assumption is, like, well, things are going to continue to change really more rapidly. And so you either get in or you don't. But things are going to continue moving at breakneck speed. And so it doesn't really matter if you think it's not. It will continue to, like, scale and change. And so just try to get in, like, fast. Try to understand that there's nothing preventing people from, like, experimenting. It's the greatest experimentation era for artists. It's, I mean, like, I was, I met this. Like, filmmaker, I was in LA a week ago. I met this filmmaker that is probably my childhood hero, like, someone I admire my whole life. My whole creative career has been somehow inspired by all of his work. And now he's, like, he wanted to meet with me. So he wanted to finally, like, he's, which felt like a dream, but. Plug him. Who is this? I don't think, hopefully he can tell at some point. But basically he wanted to, I think he went to a point where, like, this is, I've seen enough stuff. I experimented with enough with Runway. I get it. This is. This is the same thing I saw 30 years ago with, like, computers, with detail, like, compositing. This is the moment for me to start breaking new ground. And that feels to me like perhaps an ethos that more people should, like, breathe. It's like, this is a time for you to imagine entire new ways of making stuff. And it's there. It's, like, up for anyone to grab. If I was an artist, like, just, yeah, I would just do it. I can't believe PTA told you that, Chris. I have one more question. Why do all my startup clients. Why do all my startup clients want to write and talk like you? You know, a lot of them send me your blog posts as examples. I mean, I guess part of it is you have an arts background. You're not just some PM turned founder, are you? But, I don't know. Clearly you're doing something right. I mean, what do you think is resonating with your thought leadership you put out into the world? Look, if I have to be honest, I don't think of, I don't think of what I write as, like, thought leadership. This is just literally what I believe. And if I have to be. When I was writing all of these things, like, eight years ago, no one cared. Like, it just felt like a very niche, obscure thing to write about. And maybe if I think about it right now, you're asking about it. Why do you think I think people might resonate with it? It's, I think, because it's just authentic. Like, this is, I'm not trying to, like, appeal to someone or, like, you know, write the right thing for Twitter. I think it's just, like, truly what I believe. And I think one of the hardest things in AI these days is have original thoughts and authenticity. Like, this is who we are. This is the company. And I think Runway has, we've done this for the last seven years. It's, like, the most consistent thing you hear from people is, like, the culture is very unique. It feels different. And I think we've cultivated that unique culture and that manifests itself in how we speak and what we do. I don't think we're trying to pretend. We're not performative. This is, like, literally, like, how we think about the world. And the fact that, I guess, the fact that that resonates with people, it's great. But we didn't change it before. It was the same before that it is now. And maybe more people are realizing it in a way. So maybe just be authentic, you know. Try to have your own voice, which is sometimes the hardest thing to do. Word. Well, Chris, we appreciate your time. Thanks for coming on the show. Of course. Thank you guys for having me. Great chat. Thank you, Chris. Thanks to Chris for coming on the show. We're new here. We're a new show. Please follow, like, subscribe everywhere you get your podcasts. It means a lot to leave a rating or a review. We're also on YouTube, where you can watch this whole show in full video. So check that out. Access Pod. And if you're feeling so kind, please share with a friend. We're new out here. It would mean a lot. We need that dark social. We got to get that thing humming. Dark social is important. Get the algo boost. Does anybody talk about dark social anymore? No, but they should. You can find my newsletter at sources.news. Please check that out. And you can find me at Hamburger on Twitter and at meaning.company. Access. Part of the Vox Media Podcast Network. Our producer is Shiraz Dume. We'll see you guys next week. Bye. Bye. Booking.com is the easiest way from a day surrounded by noise to a stay. Surrounded by nature. That's nice. Go on. Book it. It's easy. Booking.com. Booking. Yeah. Fall is the perfect time to refresh and reorganize your space. At the Home Depot, find power tools and tool sets starting at $50 to help tackle DIY projects, home updates, and more. Whether you're drilling brackets to support, new shelving, or sharpening your hedge trimmer blade with an angle grinder, the Home Depot has the tools you need to check projects off your list. Shop Labor Day Savings at the Home Depot and gear up for fall projects with the right tools to keep your projects moving.