Revolutionizing Creativity and the Future of Storytelling feat. Cristobal Valenzuela

Speaker 1: Yeah, let's be in outer space-

Cristóbal Valenzuela: Let's do it. That's the idea.

Speaker 1: With bunnies.

Cristóbal Valenzuela: With bunnies, maybe a rabbit astronaut flying to Mars. And you can start visualizing all of those things almost immediately.

Speaker 1: And to do that, all we have to do is just type that into the prompt.

Cristóbal Valenzuela: To just type it into the prompt.

Speaker 1: Welcome to this special double episode of Ask More of AI, the podcast at the intersection of business and AI. Today we have two amazing guests. First is John Somorjai. He's the founder of Salesforce Ventures, one of the largest enterprise software investors in the world. And then we're going to hear from one of his portfolio CEOs, Cristóbal Valenzuela, who is the founder and CEO of RunwayML. Of course, they're one of the companies behind Stable Diffusion, and they just released an amazing new video gen model. Thank you so much for joining us. It's really so tremendous to talk to you. You've been at the company now for almost 19 years. You started in M& A, you lead M&A, and also you're the founder of Salesforce Ventures.

John Somorjai: That's right, that's right. And thank you for having me. It's really great to be here.

Speaker 1: Well, it's great to work with you every day, and it's exciting to talk about your amazing work with the rest of the world.

John Somorjai: Yeah. Terrific.

Speaker 1: So you joined the company on the M&A side and the company, we bought a number of companies. Can you talk about how that's gone and specifically in AI, what we've done?

John Somorjai: Sure. Well, we've acquired now a little over 80 companies over the years, and roughly a dozen of those have been AI focused. And the goal of M& A is really to marry our inorganic innovation with our organic innovation and help the company create more products faster for our customers and really bring all that innovation to our customers as quickly as possible. On the AI side, a few examples of the companies that we acquired were RelateIQ, Bonobo, which is conversational insights. We acquired Richard Socher's MetaMind, which really became the beginnings of our research team on artificial intelligence.

Speaker 1: Yes. And we had him on the podcast.

John Somorjai: And I listened to that podcast. It's terrific, and I learned a lot from it. And Richard is just a wonderful person to work with. When he left Salesforce, by the way, we invested in his startup You. com. But anyway, and then we recently acquired a generative AI company Airkit. But over the years, we've acquired tremendous talent from this group of AI companies, and we now have hundreds of people that are experts in artificial intelligence that are part of our research team, and we leverage them on diligence that we do for any investment that we're looking at or any acquisition that we're looking at.

Speaker 1: I think it's such an asset, and we were talking about this earlier, it's really fun just being a former founder myself, working with so many entrepreneurs across the company, and I really think it helps keep our culture very entrepreneurial.

John Somorjai: Absolutely. And that's another great advantage of when you think about when you're buying a company, you're also buying talent. And one of the things I'm really proud of is all the incredible talent we've brought into Salesforce through M& A, many of whom have risen to become leaders of the company, including David Schmaier, who's our chief product officer, as an example.

Speaker 1: Yeah, it's a really great point. So there's all this activity on the M& A front. We've bought a number of AI startups. I also think some of the really big non- AI companies that your team bought, Tableau, MuleSoft, Slack, they're really coming in handy now in this generative AI era, on the UI front, all the connectors bringing the data in from across an enterprise, being able to visualize all of that data.

John Somorjai: Right. And when you think about all these generative AI companies and how they intersect with Slack, and many of them are built on Slack, Anthropics, Claude, for example, runs really, really nicely in Slack. And so you see the Nuke startups are really relying on this technology as well. So that's been terrific. We also bought some very exciting technologies in marketing and commerce. So the Marketing Cloud and Commerce Cloud are really a result of a series of acquisitions we did, most notably ExactTarget back in 2013 and Demandware in 2016.

Speaker 1: It just really such foresight back then. There was a reason back then, but now it's just become even more valuable in this new AI era.

John Somorjai: Yes, for sure. For sure.

Speaker 1: So you joined the company in the early 2000s. In 2009, there's a decision to potentially start a venture arm of Salesforce. Can you tell us that story of how that happened?

John Somorjai: Sure. So it was the middle of the financial crisis and we were talking about it towards the end of 2008. We had so many partners that were unable to raise funding and there was a risk that they were going to run out of money-

Speaker 1: So these are app exchange partners.

John Somorjai: App exchange partners. And as you know from your background with the app exchange, we created this incredible ecosystem of partners around Salesforce that would tightly integrate with all of our products and operate globally and is really important for the company's growth and for our customers to be able to find solutions wherever they are and whatever they needed all pre- integrated with Salesforce. But these companies are having a very hard time raising money. And so what we decided we should do working with Mark is create a venture arm that would fund our partners and where we in many ways could become... We were the anchor tenant of the round and help them pull together a round of investors. And some examples of that era, and it's hard to believe, but DocuSign was a company that we invested in their series B. And had we not done that, that was a company that they were struggling to raise funding. And it's hard to think of that now given all their success over the years. But back in 2008, it was a very grim time and a lot of companies were wondering, " How can we keep things going?" And so we invested in Box, in HubSpot, in DocuSign, and a number of SI companies. Appirio was a notable success there and it really took off from there. And what we realized is that we have a few advantages as investors. One of them is we really understand what our customers are looking for. We listen to them and know... We see the buying signals and have a very good sense of what are the types of companies they would want to buy from. We also have all this in- house expertise that understands software and understands enterprise software in particular and how to create the best products for the cloud that are secure, that are reliable and scalable. And so we leverage all of our employees with this expertise to really help us make better investment decisions.

Speaker 1: It's really special, and I can attest to it having been on both sides of the table now. Having been a founder who raised money from Salesforce Ventures, I really felt like the team was very hands- on and helpful. And then now being at Salesforce, getting to work with these amazing startups every day that your team brings in.

John Somorjai: Absolutely.

Speaker 1: So could you have imagined 14 years ago that this fund you were starting to help hold a lot of our partners over and get them through this tough economic time, that it would become one of the, if not the preeminent B2B software venture capital arm?

John Somorjai: At the time, no. At the time, we were just trying to keep our ecosystem alive and in a very, very challenging financial market. But then really the world changed and the cloud started to really take off. Enterprises realized that they could run their companies more efficiently and save money by moving more of their systems to the cloud. And so not just Salesforce, but all these companies that grew up around us really benefited from the changes in the industry. And so where we are today, we've invested in over 550 companies.

Speaker 1: How much capital deployed? Do we share these numbers?

John Somorjai: We've deployed over$ 5 billion in capital. We have had 33 IPOs. Over 150 of our companies have been acquired.

Speaker 1: Incredible. And how large is the ventures team and is it geographically dispersed?

John Somorjai: It will surprise you that the team is actually only 33 people in total.

Speaker 1: Very efficient.

John Somorjai: And that includes our operations team as well. And so we have offices in San Francisco, in New York, in Tokyo, and in London. And our team is really, they are very efficient, but they also have the ability to leverage the power of Salesforce and to meet with experts like yourself and get your take on companies that we're looking at and your assessment if they make sense or not.

Speaker 1: It's one of the most fun parts of my job is getting to work with you and your team.

John Somorjai: And I think that other executives feel the same way too, because it also gives them a window into the innovation that's happening in the industry. And so I think everyone that participates in the process really learns from it.

Speaker 1: Yeah. So recently you launched a $ 500 million generative AI focused fund. Can you talk about that?

John Somorjai: Right. Super excited about this, and I think that generative AI is one of the most transformative technologies that we've seen in a long time in our industry. And when you really started to see interesting business models take off with this technology, that's when we started thinking about, wow, we want to be at the forefront of deploying capital into this industry and make sure that we're getting into the best companies and also building an ecosystem of partners around all of our AI efforts that we're doing internally. And so we announced a$ 250 million fund at the beginning of March, and we very quickly ran through that funding. And so we decided to double down, and now it's a $ 500 million fund and we've invested in some amazing companies.

Speaker 1: And you really have invested across the tech stack. I mean, let's even just start with a model layer. Can you talk about the companies you've invested there?

John Somorjai: Yeah, so I'll give you three examples. So Anthropic is one. We talked about the Claude model. They are really emerging as one of the most trusted generative AI general foundational models. And so when we think about, by the way, all of our investments in this category, we want to make sure that they follow our values around developing responsible AI because trust is so important and there's a lot of fear about this technology and what it can do. And so we've really applied the values and the ethics and the way that Salesforce thinks about it and approaches AI to our investment philosophy here. And so Anthropic was a company that had a philosophy around constitutional AI where they really think deeply about what is the humane use of this technology in everything that they do. And so that fit really nicely into our thesis. And also we've been able to partner with them very closely with our AI technologies, as you know. Cohere was the next one, which was more focused on the enterprise use case and they build specific enterprise foundational models for companies. And then Hugging Face was one that we led the round actually in August, which is the largest open source repository for models and growing incredibly quickly. And so those are the three that we're super excited about.

Speaker 1: Those are all, they play such key roles in our open ecosystem. And Hugging Face in particular is a community where our own research team has been open sourcing models for years, and then now it's neat to see customers just giving customers more choice in their models.

John Somorjai: Absolutely.

Speaker 1: Yeah. Okay. So that's the model layer. We've also invested in applications. Can you talk about some of those?

John Somorjai: Right. Well, I'll give you a few examples there. I think you're about to talk with Cristóbal from Runway.

Speaker 1: Yes.

John Somorjai: So I won't go too much into that, but-

Speaker 1: And I'll ask you about him in just a moment.

John Somorjai: ... But I thinkthat's at the text to video application layer and super exciting what they're doing. We invest in a company called Typeface, and we led that round back in June. Typeface is a close partner with our Marketing Cloud, and what they do is enable a marketer to have a contextualized image that's personalized for the advertisement that really preserves all the brand integrity of their company when the image is displayed. And you can talk to the image, you can add things, change things, but in every time you do it, the brand is always preserved. And the way that the company cares about how that brand looks and how it will be generated to the user is really important. And so they do that in a very unique way, and it's so fast and saves a marketer so much time, that that was really impressive when we saw them. The CEO of that company, by the way, was the former CTO of Adobe-

Speaker 1: Abhay.

John Somorjai: Yes, exactly. And so he really understands this space deeply. And by the way, a big part of when we make our investment decisions are based on the leadership and their expertise, their talent, their qualities and things like that. And so great example of a very strong leader. Another deal that was recently announced was our first generative AI investment in the UK in a company called AutogenAI. And what they do is they take the very, very complicated RFP and bidding process and they help you automate that. And so that you can generate a document that might be 100 pages for an RFP that looks contextually at what were your successful bids that you've done in the past-

Speaker 1: Oh, that's so smart.

John Somorjai: ...And what has this agency, what are they most likely to respond to? And then they create that RFP and that bid really, really fast. And so it saves the writers so much time-

Speaker 1: Oh my gosh, those are so time- consuming. Nobody likes doing that.

John Somorjai: And what all of these companies have in common is how much productivity they add back to the company. And if you think about that, McKinsey has the stat that AI is going to save$ 4 trillion every year for companies. It's an enormous productivity uplift. And that's why I think these companies are really on the cusp of just breaking through and becoming very, very successful large businesses because when you can drive that much efficiency to a business, who's going to not want to buy it?

Speaker 1: It's really remarkable, isn't it? And it's all happening so fast. You mentioned earlier bringing our values to our investment. I'd love to hear about how we're doing that. I know your team lives this every day, but just to talk about for our audience, how we're bringing our values to these investments.

John Somorjai: Sure. Well, I think one of the greatest things about Salesforce is how much we value giving back to the company. I think it's a really key part of our culture from when we started the company with the 1- 1- 1 model where 1% of employee time is donated to volunteerism. We give 1% of our product away for free. And then most importantly, we set aside 1% of our equity into a 501( c)( 3) foundation, which has now been able to give away over$ 700 million in grants. And so you can just see the incredible impact that has on philanthropy. So we've brought that to our portfolio. We have now more than 200 companies that have adopted the 1- 1- 1 model. And so they're building on the success that Salesforce has had here and helping their communities in all these ways. Another example is our focus on equality and diversity and inclusion, and we have now 55 founders in our portfolio. I think it's of who are female, and I think it's one of the largest portfolio of female founders in the world.

Speaker 1: Well, it's got to be. That's incredible.

John Somorjai: Yeah. Yeah.

Speaker 1: Yeah.

John Somorjai: And then we also invest in many, many underrepresented minority founders. We funded the Black Venture Institute, which trains new Black venture capitalists on investment philosophies, investment best practices and things like that. And so I think everything we do with our values internally, we really try to bring that to bear in the companies that we're working with and also make sure that the founders share our values too.

Speaker 1: It's just such a powerful way to scale our company values and it reminds me of what Mark always says about how business is the most powerful platform for change.

John Somorjai: Absolutely.

Speaker 1: Yeah. So we're about to have Cristóbal from RunwayML come on, as you mentioned. What drew you to this company and made your team want to invest?

John Somorjai: So Cristóbal is an incredible entrepreneur and he and his team created, by the way, the Stable Diffusion model. And they are really at the forefront of thinking how to apply generative AI to video and visual effects. And we were really blown away by their demo when we saw it-

Speaker 1: Did it have visual effects?

John Somorjai: Very strong visual effects that you can basically just talk to and create by typing with a machine and just the creativity that that can unleash and you can see how much money it would save in video editing. And for anyone that's doing video advertising, movies, television, very, very powerful use cases for generative AI and how much productivity it can add.

Speaker 1: It's just incredible. So I mean, you've had this remarkable career and I think a lot of people associate you with just so much innovation in B2B software. What was your background? Were there early experiences that led you to doing what you're doing today?

John Somorjai: Well, I started as a chemistry major, which was one of, I think, the interesting-ness about my background. My father's a professor of chemistry and I grew up around chemists. And I think when you're tinkering in science labs, I think that definitely gets your creative juices flowing quite a bit. And so that was my early training. Strangely, I switched to law and midway I decided to go to law school and become an attorney and try to apply a lot of that creativity to M&A. And I was an M&A attorney and realized that I much prefer the strategic part of the job. I much prefer thinking about, what are the right companies to acquire and then how do you put them together? And it's like when you do M& A integration, it's a bit of a puzzle and how do you take the best of both companies and marry them together and hopefully something better will come out of it.

Speaker 1: It's like a formula and then there's a reaction.

John Somorjai: Exactly. Right. That's a great way to... So I do think that scientific background really played a role in everything that I've done.

Speaker 1: I did not know that. Well, I'm sure you would've made an amazing chemist, a wonderful lawyer, but I'm really glad that you chose to keep going and it's amazing to partner with you. Thank you for everything that you and your team are doing.

John Somorjai: Thank you. It's so great to work with you and appreciate the time today.

Speaker 1: Cris, it's so great to have you on the podcast.

Cristóbal Valenzuela: Thank you for having me here.

Speaker 1: So I want to talk about Runway, but first, I know Stable Diffusion, which is such a pioneer in multimodal models, that was a collaboration that you were a part of along with Stability AI and LMU in Germany. Can you talk about how that came to be? How did you guys meet each other and what did each of the companies contribute?

Cristóbal Valenzuela: Totally. Yeah. Stable Diffusion has an interesting story and origins. It was originally a paper and a work and a research collaboration between Runway and the University of LMU in Munich who published the code, the work, and released the models almost two years ago. Since then, we've constantly improved it and changed it and modified to make it even better, mostly improving the quality and the resolution of the images that you can generate with text. Around a year ago, we got a compute donation from another company and used that compute donation to improve the quality of the model, and that was the origins of Stable Diffusion. But it's a model that's been out for now a couple of months and I think has completely changed the landscape of AI and we're very grateful to be part of that initial collaboration that that research and the model that I think changed a lot of things and has continued to change a lot of things.

Speaker 1: So what do you think it was about that paper that was so unique and really allowed Stable Diffusion to take off the way that it has?

Cristóbal Valenzuela: There are a few innovations in the paper that were transformative in the way image has always been approached and how the challenge of generating images with diffusion model was tackled before. It's built still on the shoulders of giants and a lot of other research communities who've been experimenting and exploring with the idea of reducing the size and the time it takes to be more effective and creating high quality images. I think since the release of the paper, there's been a lot of improvements on both the speed at which you can generate, but also the quality. I think that paper in particular, which is actually called High- Resolution Image Synthesis with Latent Diffusion Models, short is Latent Diffusion and then eventually Stable Diffusion had a few interesting things. Primarily I think the most interesting one was this idea of using a latent space to generate the pixels of an image and then growing from there. I think that the main impact though was I think the open source version of it and the ability for people to tune it or modify it or work on top of it. I think that definitely changed the game.

Speaker 1: And for our business audience, can you describe what a latent space is and how you're using it to generate pixels?

Cristóbal Valenzuela: Sure. A latent space is a multidimensional space that's been reduced or simplified. So it allows you to have a representation of a much larger space and you can move across that space. And the benefits of that is that you don't have to move in a much more highly dimensional space. And so it's sometimes easier, there's some compression around it. But it's a way of most models, more of AI models these days have latent representations that are easier for are more interesting or more, I guess, beneficial for generating, for example, images these days.

Speaker 1: Yeah, as well as language. It's this ability to collapse into a fewer lower dimension vector space to generate all kinds of things. And also for our audience, could you talk about how are large language models different from a diffusion model, for example? And why would you use the latter for images?

Cristóbal Valenzuela: Yeah. I mean, there are different techniques. These days, language models are mostly based on transformers. Video models and image models are mostly based on diffusion. Before diffusion, we had GANs. Before GANs, we had other techniques for image generation. Although I think those things are converging these days. There's vision transformers, so there's ways of using the core backbone of a language model. Also to drive a generation of an image or a video. Diffusion is mostly now can also be used to visualize that representation. I think we're early on, I would say in the stages of understanding how to really make the best type of outputs of both multimodal systems. And sometimes there's no one single answer of only by using a Transformers architecture or diffusion architecture you achieve, but you need to achieve. Sometimes it's combining those and making sure that you can work with a multi approach stage or using different models and mixing them to get the results that you want. But diffusions tend to be, I would say the state of the art these days on pixel generation.

Speaker 1: It's just so interesting. So Stable Diffusion focuses on images. Now Runway, you have moved on to video and talk about the new challenges that video presents relative to static images. And talk about your new models.

Cristóbal Valenzuela: Totally. So Runway has always been a company devoted to storytelling and creativity. And so if you think about the goal of expressing an ED or telling a story, you need to understand it from a perspective of a multisensory approach. And so building multimodals has always been at the core of what we do. We're trying not only to create the best possible models for both image and video, but create the best possible models for every domain that would allow you to take some idea that you have in your head and express it in the most effective way possible. And so we started with image at the beginning, and so we've published and made some great advancements on both research and product to make sure those models are usable, reliable, safe, and of course the best quality possible. And now we're moving into video. We released only a couple of months ago, our first iteration of our video diffusion models, Gen- 1. We now released Gen- 2. We've now released Gen 2. 1, which is a major improvement on the quality. These are models that are going beyond single frames, which is images to continuous frames to create a consistent video up to close to 20 seconds of videos you can generate of basically anything. Also in high definition. So we have models that 1K and 2K, and that would only continue to increase. Now the challenge on the video side, it's not only the ability for you to prompt or generate an image, it's also the ability for you to consistently control how objects are moving within that video. And so that's a new set of research problems that had to do a lot with user experience and the primitives that you can create an event to come out with ways of controlling these models in a great way.

Speaker 1: Can you share more about how you're cracking the code on that?

Cristóbal Valenzuela: It's a very interesting challenge because these are things that have never been done before. And so if you take the examples of how we thought about media for the last 30, 40 years, we've come to realize that we're building assumptions and metaphors of creative software based on analog, like ideas. And so the camera has been the way we've worked and manipulated videos and films. Now using something like a inaudible model or a multimodal system to generate video, audio, text, and images might not feed really well with traditional paradigms or metaphors of how you manipulate this, the outputs of these videos and these models. And so you need to come up with completely new set of ideas. Generating frames. For example, prompts. I'm not a huge fan of using language to drive video. And the reason for that is that trying to tell a story, even if it's just five seconds, using a prompt, it's extremely hard. Specifically is you have a lot of movements and objects and occlusions and subjects and actions. You need a much more nuance, a much more detailed interface or ways of controlling the models. And so we're working towards making sure that happens. And the way we tackle that is by actually having filmmakers work alongside the research team. And so we have filmmakers and producers and people who've been working in Hollywood for years, sitting right next to someone who's inventing a new set of research models. And that creates the right, I would say, ideas and incentives to build these new tools.

Speaker 1: So that collaboration with the end user who really understands the domain and what good looks like in the finished product.

Cristóbal Valenzuela: Exactly. And so we have a production company within the company that sits right next to our research team and helps them figure out the right things that need to be figured out when it comes to interface design and research design and how to train the models effectively to not only create the most compelling and high quality pixels, but also to have them be very expressible. We're building tools for storytelling, for media, for film, being able to control your tools becomes a requirement in any art form. And so for us is an obvious and much knitted approach to build the right things.

Speaker 1: And your tools have been used in primetime. I mean, one of my favorite movies, Everything Everywhere All at Once, uses RunwayML. Can you talk us through the process of partnering with the team and what you did?

Cristóbal Valenzuela: Totally. So yeah, we have around 30 different tools for not only video generation, but things adjacent to filmmaking. So for example, rotoscoping, which is a very baseline and fundamental aspect of video that used to take filmmakers hours and sometimes even days or weeks to make, can now be almost fully automated or simplified with tools and models we have in Runway. And so the interesting thing about that movie is that the editing team and the VFX and visual effects team behind the movie was only made of seven people. And those seven people got together to make something so amazing and so unique and so beautiful that, I mean, they won seven Oscars as a result of that. And they were very, I guess smart about how to optimize parts of that process of editing and making the film by using tools like Runway. So they use a tool we have called Green Screen that does automatic video segmentation or rotoscoping, and they managed to save theirselves a lot of time by automating those tasks and following a much more convenient approach to storytelling, which is you don't spend time doing this very repetitive things or very boring or tedious task, and you focus right now primarily on executing your story and making sure you can edit it as fast as you can. So they've used it and now some of the editors in that film, and there were people who work now are using it almost in any other project they've been working on.

Speaker 1: What did you learn from that project?

Cristóbal Valenzuela: I think that's for me is a sign of what's to come when it comes to what it means to make great stuff, great stories, great art, great films. It's not that you're constrained because of the tools or the budgets. It's a low budget movie that made a complete impact in a lot of, not only festivals, but it's a favorite of a lot of people. And I think that demonstrates that with the best tools and with the best AI tools these days, you can make things that used to take companies and people hundreds of thousands of dollars, over millions of dollars and a lot of people to make. And the fact that seven people edited that film, it's a great representation of what's coming next where we're probably going to go where to a moment in time where someone, a singular person is going to win an Oscar, Academy Award for making and editing and shooting an entire movie. And I think that's great. We're going to start hearing from people we've never heard of before. I think we have a saying at Runway that we keep repeating, it's like a mantra of ours now, which is, " The best movies are yet to be made. The best stories are yet to be told." And that's because these tools are going to make storytelling much more convenient for everyone else.

Speaker 1: That's so interesting. So you're saying we can take this recording that we're doing right now of this video podcast and run it through Runway and I can type in a prompt, " Have a frog jump out of Cris's mouth." And it'll just work?

Cristóbal Valenzuela: It will. I mean, eventually it will work like that. Right now you can prompt it and you can generate videos that will get you very close to that. And so if you want to, for example, generate B- roll, I can speak about if you want to speak about animals like frogs or peacocks or dogs or bears or penguins. Everything you want to generate and all combination of animals you want to see, you can just generate them and have consistent video being generated about them in all sorts of styles or combinations. And then control is exactly I guess the next stage, which is exactly what you're saying, which is great. It can generate a frog, a very colorful one and amazing one, a great shot with a macro zoom exactly what I wanted. But now I want the frog to stand or to jump or to do X, Y, or Z. You are going to be able to do it very soon. And that's the next stage of control that we're speaking before.

Speaker 1: Okay. So I don't know what B- roll necessarily means. So I actually want to take this... I seriously want us to do this. We're going to take this recording, we're going to run it through Runway, and you can say I can have animals in the background or I can change your background?

Cristóbal Valenzuela: Yep, yep.

Speaker 1: Okay.

Cristóbal Valenzuela: Or generate any type of video. So yeah, we can use the examples of the list of animals that I just gave you or anything at all. You can have a dog flying through space or an astronaut cat or whatever you want-

Speaker 1: Let's do all of the things.

Cristóbal Valenzuela: ... I mean,it will generate for you.

Speaker 1: Yeah, let's be in outer space-

Cristóbal Valenzuela: Let's do it. That's the idea.

Speaker 1: With bunnies.

Cristóbal Valenzuela: With bunnies, maybe a rabbit astronaut flying to Mars. And you can start visualizing a lot of those things almost immediately.

Speaker 1: And to do that, all we have to do is just type that into the prompt?

Cristóbal Valenzuela: To just type it into the prompt. Or maybe you have a drawing maybe you made, and you can not only type it, but you can use that image as the baseline, as the starting point. And so you start with an image, you don't have anything to prompt. You can start with a real image as well of something you've actually seen or saw in the real world.

Speaker 1: And the video can be created from a scratch. It can also be overlaid onto an existing video like the one we're recording now?

Cristóbal Valenzuela: Exactly. So we can do, and this is great for control, there's this idea of video to video. So most people have heard about text to video or text to image. We're talking about Stable Diffusion, that's a text to image or mostly used as a text to image model. But we've pioneered this idea of video to video. So video to video is, this is video, we're recording video. I can clap my fingers like this and turn myself into any world that I want. And I'm using the structure and the consistency and the depth and the camera angles of the existing video to turn myself into something else. And so people are using this to create worlds. And so you can flip or move or travel across infinite sets of worlds just by using the original video as a driving source and a prompt or an image as a reference guide.

Speaker 1: So interesting. So you grew up in Chile and you did your graduate work at NYU. How did you end up in this space?

Cristóbal Valenzuela: I mean, it's a passion I've been working on and thinking about almost for the last eight years. I grew up in Chile. I studied a bunch of different things, but I always had a soft spot for engineering and art. And I came to NYU to study exactly that eight years ago. I saw something was happening in deep learning at the time. There was a lot of breakthroughs. Nothing and no one was really thinking about the idea of taking this algorithms or models into an artistic environment or artistic context or thinking about creative tools. And so that's what we started building with my co- founders at school, and Runway was actually our work initially at school that we started as a company five years ago. And it's the tools we've always wanted to build. We also consider ourselves artists and have had an artistic practice. And so it's very much an obvious thing for us to think about and do. And I think always the company didn't... We didn't founded the company. I think the company founded us. And this is the only thing we know how to do.

Speaker 1: That is just incredible. So you talked earlier about just the incredible efficiency and productivity gains that you've seen production teams go through. And in parallel we've just watched what's happened in Hollywood with the strikes. What do you think this means for the future of jobs and what should business leaders and creative leaders be doing now to prepare society for that future?

Cristóbal Valenzuela: Yeah. And that's something I've been having a lot of conversations with these days with people within Hollywood and the media space, spending a lot of time helping them understand the magnitude of the change. And the metaphor I always come back to that I think perfectly encapsulates I think the magnitude of what's coming is to think about AI as a new camera. The camera 150 years ago changed everything. It changed hard, but it also changed entire economies. Entire cities were born just because of the idea of using a camera to record life. Shows, movies, series, everything was born out of a very interesting, unique technical advancement that was allowing us to capture light within a device, an optical device. This for us feels like a new camera. We are opening the doors for a completely new form of art that's going to create an entirely new form of storytelling. The camera created the seventh art, a new form of art. Cinema is a new art. We're heading to a world where AI will be a new form of art and that will change radically how we think about media and how we think about stories and how we think about art in general. And so I think the best state of mind to be in these days is to embrace that new tool, that new camera, and to really understand that exactly as a camera, you're not going to become a filmmaker just by having one, in the same way that you're not a writer if you just have a pen or you're not a painter just by having a brush. It requires an understanding of the craft. It requires an understanding of how you use it to tell a story. And the tool becomes an augmentation of that. And that's why we speak a lot about augmenting a creative process and not replacing it. And I think that's really our mission at Runway as artists ourselves, is to help people understand how to use this new brush, how to use this new tool, this new camera to make amazing, amazing things. And again, it's year zero, so we're still really early on on driving that conversation.

Speaker 1: How do you find time and space yourself to be creative?

Cristóbal Valenzuela: I think building the company has been my creative space, to be honest. I used to think of myself as I'll pursue a career in art and do something different, but it's much more rewarding these days to just be creative in what I do in Runway. I think creativity is a state of mind. I think we normally associate creativity with an artistic practice with people painting, but you can be creative doing anything you do. I mean, soccer players can be creative and the best soccer players are really creative because they think about the game in a way that no one else thinks about. And so really, creativity for me is a state of mind, is a way of looking at the world. And when you have tools like Runway, these tools can help you look at the world in a much more creative way.

Speaker 1: So question I always ask our guests is, in this changing world, how should we be educating the next generation, our children differently so that they're prepared to be as creative, as productive, as fulfilled as possible?

Cristóbal Valenzuela: I think that's always a relevant question when we speak about change and technology. I think being very specific and optimistic about change, it's a great mindset to having inaudible. We're living through a magnitude of change that I think we haven't seen in a long time, and things are starting to accelerate even faster. And so one thing that's something's always tricky for everyone to really adjust is getting used to that change. Getting used to change is very hard because we as humans always try to rely on things that we know. So we actually have a lot of programs in Runway that work alongside film studios and universities and schools and high schools and helping teach this next generation of students and artists and filmmakers and creatives how to benefit from this and how to use it not in that feels... In a way that augments them and creates them possible inaudible. And so I think by investigating, asking the right question, having the literacy around how the technology works, we're going to make that great step forward.

Speaker 1: Well, Cris, thank you so much for all of your amazing contributions to the field of AI and to making the world a more creative place. We're so proud to be an investor in your company and I can't wait to see those movies that are yet to be made.

Cristóbal Valenzuela: Yeah, amazing. I'm excited for the future as well. Thank you for inviting me.

Speaker 1: What an amazing jampacked episode. Three takeaways from me. Number one, Salesforce Ventures was founded in 2009. Initially it was meant to hold over our app exchange ISV partners who were struggling through difficult economic times. It's since become a$ 5 billion, one of the most successful enterprise software venture capital endeavors ever. Number two, Cris from RunwayML views AI as the new camera. What does that mean? It means that art and creativity and movie making will be open now to a whole new set of people who wouldn't have been able to create in the past. Number three, this means that it's time for us to rethink work and how we spend our time, and also how we educate our kids to get comfortable with rapid accelerating change. That's all for this week on the Ask More of AI podcast. Follow us wherever you get your podcasts and follow me on LinkedIn and X. Have a wonderful holiday with your loved ones.