PVP supports early stage, AI-native companies shaping the future of how we live and work. Learn more about their work at their Substack.
Connect with Songyee on LinkedIn.
TRANSCRIPT
[intro music]
Ryan Donovan: Hello and welcome to the Stack Overflow podcast. I'm here recording from the floor of HumanX, and we're talking about AI in the rest of the world, getting the international perspective. My guest for that is Songyee Yoon, who is Managing Partner at PVP Ventures.
So welcome to the show.
Songyee Yoon: Thank you. Thank you for having me today.
RD: My pleasure, my pleasure. I think it's very easy for us in the States to have a kind of... narrow, US-centric view of what AI is. But I know there is a whole world of development going on. What does AI development look like outside of the US?
SY: I mean, that's a really great question. I think since it requires such a kind of intense investment of capital, amount of the data and talent concentration, it's not easy to build the AI stack that we see in the US evenly distributed outside of the country. However, I think it's true that we're kind of experiencing this kind of whole platform transition, right? I mean, if something is going internet digitized, it's difficult for one country to kind of leave behind not getting on internet because then it gets disadvantaged in many different ways. So I think similarly with AI, I think it's other countries, even if they cannot build everything from ground up, there's a lot of demand and interest in adopting and utilizing the technology in many aspects of daily lives and businesses. And having said that, if you think about the whole kind of the supply chain aspect of AI, like semiconductor and kind of the precious materials are kind of found outside of the U.S.
So it's a global team effort to build this, to see what we have today. And also in terms of the application layer aspect of it, some of the technology, AI is a very nuanced technology. You cannot use what's applicable here in the U.S., directly kind of plant it over somewhere else and expect for it to work. It has to adapt to their language, culture, and business practices. So there is a lot of kind of fine-tuning going on to make it more relevant to different contexts.
RD: You know, we hear a lot about the Chinese AI companies, but I know there's development going on in a lot of places. And it's an interesting point about the fine-tuning it for the language capabilities because places like China, or Arabic nations have an entirely different sort of, you know, paradigm of how the languages work, right? How much effort is it to customize or build AI
for those languages? Can you do it as a fine-tuning or do you have to start from the ground up?
SY: I mean, I think a lot of it is kind of built on top of existing foundation models, so we cannot really quantify exactly how much data, what type of data and how long it takes to make it more relevant to that particular culture. But I think it can be done and a lot of things have been done. I think it's also a lot of countries are interested in building things from ground up for many reasons, including their kind of national security and other applications.
RD: Yeah, the sort of sovereign AI.
SY: That's right, yeah.
RD: How about in terms of the applications and companies that you see for investing? Are they similar to what we see in the US or are there companies that are… applying it to different areas?
SY: Oh, I mean, I think it's very similar. I think it's kind of surprisingly similar in terms of the business practices and what we are kind of trying to achieve. So the companies are solving real-world problems in a more efficient and accurate way. I think there are many applications that we see in the U.S. and similarly very popular and in high demand in other countries.
RD: I saw through a little rabbit hole through a Chinese investment firm and found them doing full AI harnesses for developing a startup where it's like, all you need is this AI, whatever it is, and you can run a company. Do you think there's more appetite for that sort of like… AI automated company, like entirely soup to nuts in the rest of the world?
SY: I mean, I think it's hard to say as a kind of blanket statement. I mean, I have seen, for example, not many years ago, we have seen like hedge fund in a box, right? It came in kind of full stack from running the firm and kind of making sure that there is a compliance and reporting in place and expense. Everything was in a box that it was easy to start a shop in New York. And I think it's a similar idea, right? I mean, if you have kind of put together AI in a box and make it very frictionless to start a company, then I think there could be companies starting with that. But we haven't heard that the most successful hedge fund started out of a box. I think it still needs kind of the human ingenuity, and the leadership of the managing partners and their kind of foresight, everything, to run a successful kind of the hedge fund. So whether you start from a box or not is not that relevant compared to: what is your vision? What is the kind of company you're building? What is the kind of problem that you're solving for the world?
RD: Yeah, this sort of soup to nuts harness sort of seems like a quick jumpstart. And then you realize you need a little more people.
SY: Right, yeah.
RD: So you also come from a gaming industry background. You were… president at NCSoft?
SY: Right, right, yeah.
RD: And obviously PVP Ventures is a gaming reference.
SY: Right. [laughter]
RD: How much of what you look for in your portfolio is gaming related?
SY: We focus on three different buckets in terms of our investment focus. One is we look at AI infrastructure related technology that includes kind of use of data, inferencing models and kind of memory, etc. So, we are trying to invest in companies that continue to be relevant despite of all the anticipated turbulence in the marketplace. And second is companies who are building the data flywheel. So, because I think data flywheel is going to be very important for defensibility and long-term moat for the companies.
RD: Yeah, the data moat, yeah.
SY: So those are the type of companies that we look at. And third bucket is the companies at intersection between entertainment and AI. And that's because of our background and also our understanding in the marketplace. But investing in the companies, we're not just looking at companies that are using technology to make it more efficient or automate existing workflow. We're trying to partner with the companies who are completely reinventing what it means to play a game. So a sense of gaming is retention and engagement. And we believe that there is a lot different form of entertainment and engagement that becomes possible because of this. So we kind of love to work with founders who ask questions about like, what is possible that was not before the existence of availability of this technology. So I think it's the same principle applies to entertainment companies that we are investing.
RD: So to ask you that question, what is possible now that was not before AI?
SY: No, I mean, I think there are a lot of things, right? I mean, the reason why we like the kind of… the concept of data flywheel is that actually we have a lot of unstructured data in different industries, like the industries that we are interested in, like insurance, like legal, accounting, a lot of healthcare, a lot of data that it's really hard to find, kind of connect the dots.
RD: And a lot of legacy formats, I'm sure.
SY: Exactly. Bring out insights without the use of technology. But I think this AI technology allows us to bring out the insight, look at the data and understand it in a way that we were not able to do before. And I think that's one example of what's possible with the technology that wasn't before.
RD: And in terms of that data flywheel, do you find that these are new companies or are you finding like, you know, long-term companies that you're investing in?
SY: We invest in early stage companies. So like we are investing in AI native companies. We're building on top of the kind of embracing the new stack of technology. But then in terms of the data set, the data set could have been in existence for many, many years. So the early stage companies utilizing data, whether it's old or new or like in a position to collect, continue to collect a new set of data to train.
RD: And in terms of, you know, how you look at companies when you invest in them, I spoke to a VC last year at this conference that said AI sort of changes how he looks at structuring companies, that the ratio to developers, to salespeople, to marketing has changed. Would you agree with that?
SY: Yeah, I think that's a really great point. I think it's a... we don't see technology as just kind of automating part of the existing workflow. We want to work with the companies where we're kind of redesigning the workflow and redefine what is the role of humans, what is the role of technology in this kind of newly defined workflow when we are kind of fully engaging and using the kind of technology at its full potential.
RD: Yeah, I think there are some folks who would say the role of technology is everything except the CEO, right? I guess I have two questions off of that. One of them serious, one of them not. Can you automate the CEO? And what is actually the role of humans in the future?
SY: I would like, I'd love to see the human judgment. And I think every human is different. Like, I mean, every CEO is different.
RD: Sure, yeah.
SY: You don't want the kind of homogeneous set of CEOs running all the companies.
RD: Right.
SY: Because oftentimes, the problems that we are solving, the world that we are navigating, it doesn't have one right answer. There are multiple acceptable answers. And how we're going to choose, prioritize one answer over the other, comes from everyone's lived experience. And I think that's not something that we want to replace with a kind of trained model.
RD: Yeah.
SY: So, yeah.
RD: So the CEOs are safe. The rest of us… You know, you talk about every CEO, every founder being different. How much of the founder's sort of experience, personality goes into deciding whether to invest in a company?
SY: Again, I don't think there is a kind of right or wrong kind of personality that we are looking for. We are looking for founders because we are investing in such an early stage. We want these companies to be successful over the coming decade and more. We want these founders to have a capacity to grow with the company. We often say that companies grow at exponential speed, but most humans grow at a linear speed. We want to work with the founders who are thriving to grow at that speed that company needs for them to grow in terms of its capacity. It's a leadership and it's kind of a willingness to kind of tackle and navigate this kind of all this uncertainty.
RD: Yeah, I have definitely seen companies outgrow their founders [laughter]. So I think we talked earlier, China was obviously an area where there's a lot of AI development. I assume Korea is a big investment. Where else is sort of the places that you're seeing AI startups?
SY: I mean, I think that recently there is an exciting company came out of the EU, right? And like the UAE that you mentioned and like Japan. I think it's most of the countries who have resource. I think it's trying to invest in this kind of future technology and build their new tech stack on top of it.
RD: And we will talk about the gaming angle for a little bit. I know a lot of our listeners are avid gamers and I keep hearing that people got into tech because of games. What are the new experiences that you see being unlocked because of AI?
SY: I mean, I think there are a lot of new attempts, so it's really hard to see what is going to be. So there is no complete transition from the… old existing type of games to new type of games. But for example, I think one of the companies that we work with, Operative Games, as one example, it started by a senior executive from Disney Imagineering has obviously brings kind of in-depth experience, the intersection between entertainment and technology. And they're building new experience. Like once you start playing game, you don't have to kind of keep sharing a monitor to solve a mystery. You get phone calls, you get messages, you get kind of Zoom call invitation on your calendar to talk to characters to collect the clues. And I think it's a kind of coordinating all of that and kind of storytelling and very natural interaction. I think enabled by their kind of AI stack they're using. And it's one very early, but it's very early, but like an example of like... your type of engagement experience is being brought out because of the existence of technology.
RD: Yeah, I think I heard somebody mention earlier in this conference that, you know, games are just one sort of like entertainment experience. And there's sounds like there's a blurring of the line between what games are and what entertainments are.
SY: Right. I mean, I think that it has been happening for a very long time. For example, I think it's… Unreal Engine that was made for a game development. Triple-A game development is widely used in many movie studios. It's very well known, right? Special effects. I think if you go to Vancouver, which is a kind of hub for all this post-production and... there are talent who work for a gaming company and then also they work for the movie studios because the kind of underlying technologies that we use are very similar and sometimes it's interchangeable.
RD: A lot of gamers' experience with AI in games will be the sort of AI slop because it's very easy to create games at this point. It's much easier. So it's much easier to create bad games. Do you think about the sort of, you know, the structures that the gaming industry will need to filter and curate better games?
SY: That's a very good question. Gaming itself always evolves. It always changes. The type of game that we played when you're young or like even before internet age is very different from the type of game that we play today, right? But we've been playing games all the time. It just look different. Gamers are early adopters. They would like to use and they'd like to experience what this new technology can unlock, what kind of experience that make it available for them. Not necessarily the most polished experience, but I think there are different type of gamers who kind of want to have that, like... very kind of movie-like, very polished, immersive experience. But there are a lot of gamers who would rather be on the kind of rough edge of the new technology and kind of experience what's possible. So just like gamers are not homogeneous, type of games that's going to be embraced and appreciated are not homogeneous. And you're right that with the availability of these tools it's much easier to kind of make good enough games. But just by making good enough, it doesn't mean that there will be enough of the fan base and the users who make this business endeavor sustainable. So gaming industry's responsibility is to continue to reinvent ourselves and come up with a new type of experience that can satisfy the early adopters, our fans, who are looking for what is the next evolution of the innovation.
RD: Okay. You know, speaking of some of the more, you know, cutting edge things that have been positive about games, are you looking at world models at all?
SY: I mean, I think it's a really, it's a fascinating technology. You know, we are just kind of looking at its evolution with a very keen interest.
RD: Yeah. You talked about the infrastructure, data, and then the sort of entertainment, and they all seem to have like a through line, right? When you are investing in companies, do you– is part of your strategy to pair them up or to get pieces in the portfolio where it's like, well, this company could help this company?
SY: Yeah, I mean, I think that happens a lot too, right? I mean, I think it's not the primary motivation in investing in companies. We want each company to have its standalone chance of success. But when we see a lot of synergies, we make introductions and help them each other. And also because of our experience working with big tech companies, we introduce companies to those companies because one hand, companies, big companies are looking for where is the innovation forefront. And also the startups are looking for their future customers and collaborators. So I think that introductions are also very appreciated.
RD: Yeah. Some people listening may think of a VC's role as just, you know, money. But I don't think you would agree with that.
SY: [laughter] I mean, I think VCs would like to build a portfolio of investment, portfolio of companies to generate predictable return. I think similarly, founders need to think about what is their portfolio of investors, right? So like some investors will bring one thing and then other investors will bring other things. So we tend to be more available for the founders when they need to, because I mean, oftentimes being a CEO is very lonely. And there are not many people that you can talk to about all sorts of things, such as HR matters within the company, and also strategic choices about their product, technology, and even some personal matters, too, like how to navigate and balance their personal life with their professional lives. And I think because we have experience of being an operator like running companies at international scale and also kind of understanding technology from our academic background and collaboration with universities and innovation hubs we can be kind of personal therapist and board member when needed. And I think that's something that's appreciated. And I think each VC offers a different type of support to the founders.
RD: Now, have you, I mean, I know the sort of AI companies that you're looking at are solving similar problems around the world, but are there differences in sort of working culture and how they view investment, that sort of thing?
SY: Yeah, I mean, I think it's a different founders with a different background kind of certainly has different approach. I mean, founders who are like serial founders who have run to have like exited a couple of companies. They have their own lessons from their kind of hard learned lessons. And versus– versus founders who when like, this is the first time starting a company. So, yeah, I mean, again, I don't think there is a one right answer. I think each founder has to find the kind of model and each company has to find a culture that kind of fits what they're doing. We're here to support whatever they do.
RD: You said there's not one right answer. Is there a wrong answer?
SY: [laughter] I mean, I think that growth mindset we talked about is very important. I think we are living in the world, there's a tremendous amount of data available, which means that we are also expected to kind of learn and grow. I think that complacency is something that could be not very helpful for the growth of the company. I mean, there are many things like that. But again, I think each company brings different culture.
RD: Yeah. And, you know, you mentioned a few places that looking at that have the sort of resources to have innovation. Are there places that you're looking to break into that are resistant to sort of investment? Are there other countries where it's like we would rather, like culturally, we'd rather bootstrap? Or we don't have the sort of tech pipeline– you know, I've heard of places in South America that have developers, but they don't have the sort of startup culture where people cash out and are able to start their own companies.
SY: I mean, I think as a fund, we try to invest in... companies and the market that we understand what the growth trajectory is going to be and how it can be exceeded and the return can be realized. So we tend to focus on the market that we have a good understanding of.
RD: Yeah. So with all these pieces, do you see... the approach to the sort of building the software, the software engineering itself changing within companies?
SY: Yes, I think it's just like any other workflow. I think it's the kind of process of doing software engineering is changing, right? I think it's how it starts and how it's evaluated and what's kind of AI being used, not just for writing code, but also doing QA and kind of how to check. So the role of the humans is more of a kind of orchestrator and using those tools. So I think there are a lot of changes in kind of how the workflow is actually designed. But what doesn't change is that I think it's the same thing with the class I taught. It was the first class that I had to publish AI policy, right? I mean, I think a lot of students may use AI tools for developing their kind of working on their project. Ultimately, individuals, students, have to be responsible for everything they committed, everything they kind of committed to their repository. So I think similarly, I think it's coming up with a workflow process that's designed to really fully realize the kind of potential of this technology is important, but also it should come with a responsibility that everyone who commits has to read it and be responsible for what they have committed.
RD: Yeah, I think you talked about earlier that the value is the human decision, right? And somebody has to be responsible for that decision. All right, so if people want to learn more about your fund and your work and your portfolio companies, where can they go?
SY: Oh, so we have a website, principalvc.com, and also we have a substack that we publish our kind of latest thoughts and ideas quite regularly.
RD: Okay. Well, thank you for listening, everyone. We'll talk to you next time.
[outro music]