A director from Adobe was on my thesis evaluation committee, and they offered me a role in the company. I interviewed with them. I wasn't looking for a role, the truth is that my idea was to return to Chile, but they offered me this and I started to consider it. They gave me the opportunity to stay in the United States, to work in what I liked, they could support me with a visa, it was basically everything I wanted. But I told them no. Now, why did I tell them no? I don't know. I think that sometimes one enters into decisions that are more rational and decisions that are more emotional. This was a completely emotional decision. Rationally it makes no sense, I had nothing, I had no other option, no other offer, no other opportunity. It was either everything or nothing. And I left with nothing. But emotionally I felt that this idea of Runway might be something more. We launched a model a week ago called Gen 4.5, which is the best model in the world. The model we launched that generates videos surpasses the models from Google and OpenAI.
This is Mass Peach, a podcast from DF Má where we learn the stories of those who founded companies. Not only the successes, but also the failures, the difficult decisions, and the lessons learned that don't make the headlines. We seek intimate and profound conversations, without pretense, without prepared speeches, real stories told by the founders themselves. My name is Juan Pablo Silva. Welcome to More Peach. Today we have with us Cristóbal Valenzuela, founder and CEO of Runway, one of the world's leading companies in artificial intelligence. It has raised more than $500 million. It is valued at more than 3.5 billion and has received investment from funds such as Nvidia, Google, General Atlantic and Softbank. Cristóbal grew up with a kinesiologist mother who raised four children almost entirely on her own. That's where the idea that accompanies him to this day comes from: figure it out yourself. As a child he liked movies, video games and recording videos with a camera that had belonged to his grandfather. He wanted to study film and his mother didn't stop him. She told him something simple: do whatever you want, but be very good. The big turning point came when he learned about Google's Deep Dream experiments. He became obsessed with artificial intelligence and if artificial intelligence wasn't just for data or cars, but for creating stories. In 2018 he founded Runway with two partners without a manual and learning to start a business as he went. Well, welcome to this new edition of MPach. Today we have a very special guest who comes directly from New York to talk on this podcast, this conversation thanks to Activa la Rainbial, who brought him here for a seminar, and we're going to talk about many things. Who is Cristóbal Valenzuela, the co-founder of Runway? But also his entire vision for media companies, what he is creating. Cristóbal has a blog, you can find it on Google, which says, among many other things that we are going to talk about, that Runway is not an artificial intelligence company, it is a media and entertainment company. And in fact he says, I think the era of AI companies is over, not because AI has failed, but quite the opposite, because AI is becoming as fundamental an infrastructure as electricity or the internet. Welcome, Cristóbal.
Of course, thank you so much for having me.
Hey, let's go way back, before the Silicon Valley glamour, New York, rounds with Nvidia, etc. Are you a business engineer? Why?
Why am I a business engineer? Well, because I studied that. I didn't really know what to study when I started college. I think business engineering gives you the versatility to be able to do anything, and that interested me. I had interests in art, film, engineering, and business, and it was a career that allowed me to be quite flexible in where I went.
I read somewhere that you once told your mom you wanted to be a filmmaker and she said, 'You can do whatever you want, but you have to be the best.'
But you didn't want to be just an artist.
For me, making movies, or at that time, which was one of the things I thought about doing, was a way to channel my interest and creativity more than anything. My mom, more than anything, encouraged me to be very good and to take things very seriously. And today I work with some of the best directors in the world and it's an honor to be able to stand alongside them, and I've been able to do that. It's not a very traditional career, but because I took a different route, in the end I still managed to reach that destination somehow.
So, deep down, linked to cinema, you always looked for a different way to be linked to cinema.
I think I looked for a different way to be connected to what interests me more than film. Film is one of the many interests I think I have, and being able to work in that industry wasn't a specific goal; it's more a result of pursuing what interests you, of curiosity, and you inevitably arrive at that destination and many others as well.
You studied business, I understand you were the first in the program. Among other things you were an assistant to the current president of Corfo, vice president I think it's called. And at some point you went to study at NYU. Why NYU? Because deep down, in Chile you won't be able to fulfill that need for creativity you're looking for?
The truth is that it was also a little bit driven by curiosity. I graduated from university here in Chile and I stayed teaching at the university and I worked at a consulting firm for the other half of the time. In the other half of the time I had like three lives. I acted as an exhibitor and worked with professors and artists doing specific projects in some museums, for example. And little by little I became more interested in this idea of research and working in artificial intelligence. Now, we're talking about 2015. In 2015 the biggest progress in terms of algorithms and artificial intelligence models were models that could basically classify objects. Sorting objects is basically you take your cell phone, search on your iPhone, put in photos and type dog and only pictures of dogs appear. That technology that we take for granted today was a very significant advance 10 or 12 years ago, and I was very interested in the idea that the way we got there, which was a technique called Deep Learning or neural networks, could do that.
And it was a Google paper, right? The first one?
The neural networks approach to using deep learning in the field of AI dates from long before that, from the 40s even, much earlier. However, in 2012-2013 there was a series of papers and significant advances in how these algorithms were trained and the possibility of scaling these algorithms over time. Before that, there were other techniques for solving similar problems, but they generally never worked very well. In 2012, 2013, and 2014, they first started working really well, and it was the first time they used graphics cards to train the models. That's where the field started to take shape. The moment we're living in today is basically the result of those early years of experimentation, starting mainly in 2012. There are some papers from the University of Toronto by someone named Geoffrey Hinton, who worked with Ilya Sutskever, the co-founder of OpenAI, who later did CS. That's where we started, more than anything, as part of the boom we see today. I was fortunate enough to see that and part of that beginning of those papers in 2014 and 2015, and that's when I got interested. I said, look, there's something here. There may not be anything, but I'm interested enough to see if I can get a scholarship and study this for two years, at some university that has the capacity to show me a little more about what it's about. So I got in, I was accepted at NYU and I spent two years studying basically this idea of artificial intelligence and specifically how to apply it in creative industries.
In what context did you read those first papers? What were you doing when you started reading this? Because it wasn't a fad. Nowadays, of course, someone might say, 'No, I read a paper on this or that, or I read Cristóbal Valenzuela's blog where he's talking about the new Pixel Economy.' What was your context for reading that?
Well, I was doing research. I was working at the university, and being curious, I actually found some papers hidden in internet forums. There's a very close-knit, specific, and rather dense community in terms of research in AI. There are portals and websites where people generally share a lot of science. If you look, you can find them; perhaps they aren't discussed more at a mainstream level because they're quite specific, and that generally happens with any scientific industry where the transfer or dissemination of that knowledge takes time. There are many things happening today, I don't know, in nuclear energy or semiconductors or AI, that people working in the industry have known about for a long time, but being able to communicate the impact of that and for people to be more publicly aware of the consequences takes time, because generally there is a transfusion, a diffusion of technology. And today, many of the things that have happened over the last 10 years, generally within the AI research communities, have become known. It's more that today they are becoming known to the rest of the world, but in general, if you want to look, you'll find it.
And then you go to NYU, you start studying—I understand it's like a mix of art for engineers—and you meet your partner, Alejandro, and Anastasis. But they were researching?
We were just playing around. I don't think we were researching per se, but the truth is, we went to New York University, the three of us met there, we were part of the same group, the same class, and this school is quite multidisciplinary. The best way to think of it is perhaps as an art school for engineers, an engineering school for artists. So, people who go there generally are people between 25 and 34 years old with work experience, who come from some technical field—they might be engineers, researchers, scientists, or people with recognized artistic careers—and they come to test ideas, to experiment with cutting-edge technology. When I arrived, many of the things they were doing were VR, Oculus. Oculus was just starting out, it was quite small, and when I arrived, I started to get interested, along with a few other people, in this idea of artificial intelligence and how we could work with it. Again, this was very much a new field; there was nothing like what we know today about how these models work, how they scale, how they're used, but it was enough to start doing a couple of projects, get some funding, and we gradually started working together. That was basically two years of work until we graduated, and that's when we started the company.
What year was that? Graduation.
We graduated in 2018. We worked at the university from 2016 to 2018. And that's where it all began.
I read an anecdote about your thesis, one of the people who examined your thesis was from Adobe. And he saw what you were doing and offered you a millionaire salary, a visa to stay. Tell us that story.
Yes, of course. So, the idea for Runway was born out of that work, that experimentation, that game we were playing at the university, and my undergraduate thesis was for Runway, it was called Runway and it was about this idea of taking these models and being able to use them in more creative, more interesting, more different ways. And, within the evaluation committee for my thesis there was a director from Adobe, a multinational company that makes things like Photoshop or Premiere, which are like the industry standard, the standard of the creative industry today. And they offered me a role in the company, I interviewed with them and well, it was a super important position. I wasn't looking for a role, the truth is, my idea was to return to Chile, I had to return and be here. But they offered me this, and I started considering it, and it was very interesting. They gave me the possibility of staying in the United States, of working in what I loved, they could support me with a visa—it was basically everything I wanted. But I said no. Now, why did I say no? I don't know. I think sometimes you make decisions that are more rational and decisions that are more emotional. This was a totally emotional decision. Rationally, it doesn't make any sense because I didn't have anything, I didn't have another option, another offer, another opportunity. It was either I offer you everything or I offer you nothing, and I left with nothing. So rationally, it doesn't make sense, but emotionally, I felt like my intuition was telling me, we have to see if this idea of Runway could be something more. And when I said that, he immediately told me he needed more people to take it further, and he started talking with Alejandro, with Anastasis, and we said, how do we do this? We take it and make it stronger. I'm on that track.
And on a more personal note, you said it was a totally emotional decision. You're with your wife, partner, what do you say to her, and what does she say to you regarding rejecting that?
The truth is, my wife has been quite supportive in that sense. We always make decisions together, in terms of what's best for what we want to do, not what's convenient for us. And the truth is, again, for me it wasn't a premeditated decision to say, 'I'm going to do this, then advance some money, I'm going to start a company.' Basically, it was like, look, if there's something here, we're going to discover it and invent it, and if there's nothing, at least I tried. There's a framework or a way of thinking that I sometimes use, which is basically this idea of minimizing regret. It's about minimizing future regret, and I think that's what I was thinking at that moment. If I made a decision that went against starting Runway, and someone else built Runway, I was going to regret it a lot and say, 'No, that should have been me.' And I think that idea, that way of thinking, really comes into play when making decisions. If I make this decision or that one, which of the two am I going to regret if I go down one path or the other? And I think that in that decision, at that moment, I felt that I was going to regret it too much if I didn't at least give it a chance, and you never know, so it's better to give it a chance. I'm still going to regret it.
Totally. Hey, so you started Runway with Anastasis and Alejandro, but they were three friends just playing around. How did it become what it is today? Tell us, obviously a long story, but they probably didn't even know how to be a company.
Yes, there's this idea in English, like, 'it's an overnight success that took 10 years.' It was a success from one day to the next that took 10 years. Well, the truth is, we didn't start from scratch. We started without much in the way of connections or experience, and we just worked really hard. If you work really hard, have an obsession, a vision, and a consistent ethic with how you want to see the world and how you want to change it, you can achieve anything. And for the first few years, I think, we were just going at it really hard, working day and night, every single day, Monday to Sunday. And we were small, there were three of us. Then we hired an engineer, we were four, then five, six, seven. We started launching products, people started liking them, and we started iterating more. We started to grow, and there, more than anything, I think the last two years have been more about public visibility, more about people getting to know Runway and what we're doing. And I think that's partly because the models are working quite well, and the models are quite useful and quite visible in terms of what they can do, so it becomes a much more public conversation. But within certain industries, in earlier times, it was kind of known. There were people who knew Runway in small research or product communities. Nowadays, I feel it's much more mainstream, much more well-known, and that's mainly because of the work we've done over the last eight years.
What was perhaps the tipping point moment? When did Runway explode and start to become more mainstream, as you say?
Well, I think it's been less of an explosion. For example, we launched a model a week ago called Gen 4.5, which is the best model in the world. That model that generates videos outperforms Google's and OpenAI's models, it's a very good model. And we've had moments like that before. So, there are milestones that basically mark the company's growth, and those milestones are consistently moving, shifting the curve. And there are milestones that are more important, like this one, which is very important, the one we did a year and a half ago called Gen 3, which is like the third version of our models. It was also quite important because it was the first time the models worked really well. And then we had other moments with other models before that. I think it's been more than just one moment; it's been the sum of many of these releases and updates that we constantly do.
Compound interest, basically.
Clearly, clearly, clearly. And what happens is that you think of compound interest exponentially from that point of view. Going from one to two doesn't sound like much, but from two to four, from one to 8, 16, and you start to scale like that, going from 500 to 1000 and it's quite different.
Right. I imagine it's also helped that AI has become very mainstream. Totally, perhaps for better or for worse, you ask anyone, probably any generation, and they'll say, 'Oh, chatbot.' So, nowadays it's something that maybe everyone in this room has on their phone. I imagine that's also helped you with this evangelization, but I also imagine it's brought you, I don't know if problems, but we're not like that, we're something a little deeper, I mean, we don't even compete with that.
Yes, that's a good point. Runway doesn't exist in a vacuum. Runway exists in a very competitive industry with other companies. The best companies in the world are doing similar things to what we're doing. And that's been a recognition from the industry in general that we, along with many others, have been able to demonstrate that this technology is quite useful and necessary, and that has had implications that, like you say, everyone knows it. Now, there was a moment in the last two years that marked a pretty big milestone in adoption, which was when ChatGPT was launched and went viral. And I think as a consequence of that, ChatGPT became synonymous with AI, like Kleenex. In Chile, at least, we say toilet paper, we associate a brand with the category. In the United States, something similar happens with paper products like tissue or sniffing paper, which is the same thing. You don't say 'like paper,' you say Kleenex. It's become synonymous with an industry. And what happens with that is, it's a bit like what you're saying, that many times a company is confused with an industry, and a product is confused with an industry. It's like saying that Amazon is the internet. Of course, Amazon isn't the internet. Amazon is a large part of the internet, but to say that Amazon is the internet is simply incorrect. And nowadays I think that's been changing a bit, and it's more like spaces to discuss and be much more specific, but I think at the beginning it was quite obvious, it was like people would make that comparison with AI and the only thing that came to mind was OpenAI or ChatGPT. Nowadays I think that with the competition and the models that we and others have launched, that's become a bit clearer, and it's also been necessary because you need much more diversity than just one company doing one thing.
You guys have raised a lot of capital with incredible funds like General Atlantic, Google, Nvidia, Salesforce Ventures, Softbank, there are probably more. Let's talk about that path. Because, of course, you start doing this as your thesis, maybe just for fun, and suddenly you realize it can be your way of making a living, and then you realize it can be your way of being in the elite of the world. Let's talk about it. From that journey. How was that moment when maybe you sat down with your partner and said, 'Hey, we've got something here, we can play in the big leagues, let's go raise capital.' The three of you probably looked at each other. You've raised capital, right? Do you know any VC funds? No, how do we do it? Let's go down that road.
