Games

Podcast: Games and AI (with Joost van Dreunen)

By Eric Benjamin Seufert

On this week’s episode of the podcast, I am joined by Joost van Dreunen, the CEO of Aldora, an adjunct professor at NYU, and the author of the SuperJoost Playlist blog as well as the book One Up.

We dive into the profound impact of generative artificial intelligence on the video game industry, exploring the immediate market reactions to new tools and the long-term implications for creative production. Among other things, we discuss:

  • How generative AI models like Google’s Genie will actually transform the fundamental mechanics of interactive game engines
  • Why public markets reacted so aggressively to experimental AI tools by devaluing established gaming and advertising stocks
  • Whether the reduction of production costs through AI will lead to a surplus of mediocre content for consumers
  • If game companies are making a mistake by prioritizing technical metrics over the cultural and creative aspects of development
  • What happens to the cost of gaming hardware when AI companies compete for the same high-end chipsets and resources
  • How a move toward smaller, local language models could eventually disrupt the institutional power of large technology corporations
  • When the focus will shift from simple cost efficiency to utilizing AI for creating genuinely new consumer demand

Thanks to the sponsors of this week’s episode of the Mobile Dev Memo podcast:

  • ⁠⁠INCRMNTAL⁠⁠⁠. True attribution measures incrementality, always on.
  • Xsolla⁠. With the Xsolla Web Shop, you can create a direct storefront, cut fees down to as low as 5%, and keep players engaged with bundles, rewards, and analytics.
  • ⁠Branch⁠. Branch is an AI-powered MMP, connecting every paid, owned, and organic touchpoint so growth teams can see exactly where to put their dollars to bring users in the door and keep them coming back
  • Voyantis. Voyantis uses predictive AI to transform customer value into high-impact signals that boost ROAS across Google, Meta, and more.

Interested in sponsoring the Mobile Dev Memo podcast? Contact Mobile Dev Memo advertising.

The Mobile Dev Memo podcast is available on:

Transcript

Eric Seufert: Welcome to the Mobile Dev Memo podcast. I am your host, Eric Seufert, and I am joined today by Joost van Dreunen, who is returning to the podcast for the second time. Joost, how are you?

Joost van Dreunen: I’m pretty good. How have you been?

ES: I’ve been well, as I know you have. And so I’m bringing you on at a really frenetic moment in the gaming landscape. So maybe just to set the stage a little bit for those who aren’t aware, and then I’ll ask you to reintroduce yourself. You know, we had Google showcase their Genie 3 world model through the Genie tool. Genie 3 is the model. And the tool showcased the ability to render a kind of playable, call it demo or a playable environment. I don’t know how you’d really describe it, but that could be up to like some seconds long, I think it was 15 seconds or something. And that caused a lot of gaming stocks, a lot of, even in the mobile gaming advertising space, but certainly in the pure play gaming space to crash, right? So you saw Take-Two, Unity, AppLovin, and others just see a very dramatic drop in share price because it was deemed, I guess, by the public markets that this new tool would just replace them. It would absorb everything, it would devour the gaming industry. And that’s the context for me bringing you in. And I want to talk kind of more conceptually about all of that. But before we get to that, could you please reintroduce yourself to the audience?

JvD: Sure thing. And I appreciate you making the time for this conversation. Thanks for having me. Short version is, I run Aldora. I’m the CEO and co-founder. We are an AI-native BI company and we focus on interactive entertainment. You know, we try to find ways to explain these changes to companies in a way that allows them to measure and perform better over time. Particularly as they transform themselves, the lead up to all this was I was previously the co-founder of SuperData Research. We collected a bunch of data around free-to-play titles, described the transition product to service-based games industry, and then Nielsen bought that out in 2018. I was there as the head of games for a bit. And then during the pandemic, I wrote a book on some of those adventures and started really getting serious about my teaching at NYU Stern at the business school. After that, I became an investor and started Aldora. So my baby in all this is my newsletter, the SuperJoost Playlist. And that’s my laboratory for the second book I’m writing. So I like to use a lot of words to talk about numbers. And I think that that’s where our paths generally intersect.

ES: Well, can you give us a sneak peek? What’s the second book about?

JvD: The second book is about the fact that platform companies and big tech companies are taking over. Over the last 10 years, you see this massive land grab, this shift in power structure. And as a consequence, it’s become much harder for creatives to do their job. Even large companies like Sony struggle. Companies like Roblox, despite having 150 million daily active users, still unprofitable. And I trace that back to sort of the overarching power that big tech has. Everything sings and dances according to their metrics, to their thinking. And that grinds to a halt at some point. At some case, it just becomes too expensive. And so the final chapter I’m working on now is sort of like what is the possible solution to this? Like how do we get out of this rut where growth is stunted? I’m sure you’ve read Matt Ball’s report. You know, it’s like, how do we turn this ship around? Is it only because big tech doesn’t make money that the games industry is going to decline? I don’t think so. But we just have to rethink strategically and economically how we position ourselves creatively. So that’s the sort of short preview, but it’s as usual, data-driven and extensively researched.

ES: Well, one thing I want to promote your blog and I’ll link to it. But what I love about your blog is that everything is very exhaustively supported with data. And that’s very rare. And I really appreciate that because I try to do it and I know how much extra work it is. I remember when I first started blogging, I remember I kind of set this standard for myself that I need to show my work for everything because I was writing about a topic that a lot of people were hostile to at the time, which was freemium, free-to-play. This was, you know, I started my blog in 2012. It’s funny to think back on that, but people like, freemium is a fad, free-to-play is a fad, this is never going to scale, gamers don’t want this. And I said, no, the new kinds of gamers that will be created as a result of this do want it. And you can’t think about gamers being this stable, unchanged profile from the time of the Nintendo NES to the time of the smartphone. Like you actually build a bigger tent and you onboard more people and they look different than you and they engage with content in different ways and they have different priorities and they have different tastes and preferences. But anyway, because people were so hostile to this idea, I kind of gave myself this assignment of like, I have to show my work. If someone wants to attack me, they’ve got to come up with different data. They’ve got to show why the logical kind of sort of conclusion was wrong from the data that I supported. And you always do that. And I really appreciate that because I know how much extra work it is.

JvD: I appreciate you pointing it out. If I’m truly honest, it comes from a deep-seated insecurity. You know, it’s like, am I getting this right? Am I just making this up? Am I reading something that… and so and that becomes sort of a self-referential check every time. Like, can I justify whatever it is that I just said with some kind of evidence? And look, let the evidence be rudimentary at first and then improve in a sort of scientific way. But it’s much more comfortable for me to speak from a place of data. Let’s just take a bunch of revenue numbers for publicly traded companies, figure out their profit margin and say the number went up or down or whatever. And so we can compare two numbers to each other and then have something to say about that. So often, particularly in creative industries, people have deep emotional connections to things. You see this all the time in the news where it’s like, well, this title launched, this executive said or didn’t do this thing, and people have strong reactions to it. And say, yeah, but none of this is numbers-driven. Why don’t we look at the digits? And what you discover, and I know that you know this because free-to-play was one of these examples of the industry that came from a premium model having to understand freemium, free-to-play monetization. That’s a mental model. There’s an inertia into thinking about it. And I find those kinds of questions fascinating. Like, why is it that we can’t see it until later in the development? Why does it take so long to pick up on these macro-level trends? And that’s where data is essential. But it starts with I’m just totally insecure that I say something nonsensical.

ES: Well, insecure a lot of times brings out the best of people. And in this case, certainly it does. Let me start with a very just kind of big picture question. Do you have a broad overarching view on the role that AI will play in game development in the medium term? So let’s call medium term next two to three years. Like, we hear on LinkedIn, everyone’s talking about we use AI and dah dah dah and we’ve integrated it into our workflow. What’s your sort of broad overarching view on the role that AI will play in game development in this kind of immediate term?

JvD: I like that question. The short answer for me is that when it comes to the games industry, AI and the technology that it represents is catching up to the business case, right? And so the creative process where we are, and you know, I’ve written about this extensively, but the short version is that I look at the industry over a course of decades. And what you identify really easily is an oscillating pattern between innovation centering around content and content creation and economies of scale. You know, when new technologies come around and people start to make three-dimensional games and they start to do all this sort of cool new stuff because they have new technological affordances. And then it’s what follows generally is a phase of innovation around distribution where we come up with pricing schemes and new channels and new ways to get things to people and so on. And so currently we’re in a phase of distribution innovation. It sort of kicked off in 2023, 2024 around the time when the Super Mario movie came out and Nintendo realized that they should totally be in the business of making film productions, when people start using user-generated contents to enrich the experiences like Roblox and Fortnite. And so you see a broad range of activities nowadays that weren’t really standard five years ago when everybody was still consolidating and looking for economies of scale. AI is a technology that arrives at a time when the industry is looking for new distribution models. And while naturally the conversation has been a lot around, oh, this is going to make things cheaper to make, we can now generate cool stuff at the same level as, you know, XYZ AAA titles, it really comes down to my mind, how do we deploy AI as a distribution innovation? How do we make it so that we for instance democratize the production process, right? So that we give tools to people that we allow it to be something that’s more than just making things that are interesting but kind of cheaper. And so the medium term for me very much revolves around, you know, the 20 years or so that AI has been a significant part of the production process in gaming. Now it starts to kind of manifest in that sense and like how is everybody geared up for this? How’s everybody set up to make the most of this? That makes it fundamentally different than say, you know, some of the precursors to this, like when people were talking about the metaverse or crypto or VR. Those seem to be more exogenous developments where they kind of came from outside or were sort of imposed by technologists. AI has always been part of gaming, you know, procedurally generated content, machine learning, natural language processing. A lot of large language models start with like the training models provided by chess. And that type of thinking I think now manifests in gaming like how do we use this today as ways to make things more fun, more accessible to people. And so the short version of that then, and I think that that’s sort of the crux of this conversation, is where AI emerges as a tool for everyday people to make things in effect is sort of something we’ve seen in the past, but it is often misinterpreted by investors as this holy disruptive phenomenon. I think it’s an extension of what was already there and an enrichment. But so that’s the discussion around like how do you value AI and its disruptive capabilities today?

ES: Yeah, and I think that that’s under-appreciated because I think the most obvious application is to see a lot of these generative tools just as a way to reduce the cost of art production in a game, right? Which is sure, you could do that. I mean that’s certainly one thing you could do with these tools. Another, you know, similar application is with ad creative for user acquisition in the mobile space, which absolutely that’s that’s happened. But that’s not really creating a lot of value. That’s just reducing cost, right? I mean maybe you could reduce some head count or you’re maybe spending less money and you’re going to be, you know, but that’s not producing a better more engaging, more retentive product, right? And I think you’re right, it’s like well actually that but that is going to be where consumers benefit where probably there’s more commerce as a result is in those kinds of applications of thinking through like, well how how do we utilize these tools to make just a better product? A product that keeps people engaged for longer or allows the worlds to be more immersive or or larger? And that that I think is like the next leg here.

JvD: Yeah, this is the forward and the rearview mirror type of approach you see so often when new technologies surface, right? Now of course there’s a caveat here is like the the that AI has been developing gradually and slowly for years as opposed to some of the more recent ones I mentioned crypto or Web3. Those based on just the research and the effort behind them were much more explosive, right? It’s like nothing happens for a while and then all of a sudden there’s all this attention. This has been a much slower, more gradual development and so it comes with a lot more weight behind this. But the immediate interpretation is great, that one line item in my P&L that says headcount that is the most expensive piece of everything, what if I can cut it by 20%? What if I can cut it by 50%? Right? And I think that that’s an incredible fear for a lot of creatives and correctly so. I think it’s also a mistake for a lot of large companies to jump into aggressively because game companies are not tech companies. That’s kind of the fundamental thing is like you are in a creative business. The key mistake I think that the games industry has started to make over the last five years or if not longer is that they have started to mistake themselves for a tech business. And I get that, right? Is as games as a service became popular, everything was measured by metrics like MAUs and ARPU and it all sounds like and walks like the tech industry so maybe it is the tech industry. It’s a software-based business but it’s something that has a cultural component to it. And that I think is a different model and that is where technology plays a different role. And so for those reasons I would argue that AI can be conducive on the short term for like small and medium-sized creative outfits to say, we can now do much more, right? The person that does the level design can also help with the marketing messaging because we just have like a small language we have some custom GPT thing that we built and trained on our messaging and it allows us to just send out more effective tweets or whatever and go that route where you know you can just do more. You can increase productivity on the per capita basis. At the same time, you know this is only an efficiency, this is not necessarily growth. And I think to mistake the two, that’s a critical issue that’s happening right now where large companies are asked to first reduce their headcount. That’s been going on, the layoffs are well documented. Last year we had this record of 15,000 people discharged. 2024 and 2025 was like 9,500 according to Matt Ball’s report together with Amir Satvat and it’s the, you know the number of people losing their jobs in the face of sort of an economic decline is one thing, it is accelerated by this technological promise. But my guess is that two years from now a lot of these large companies are going to be rehiring people that have gotten really sophisticated in their use of AI because that’s going to be the new thing. All of this to say, at what point when the smartphone first became popular, you had games like Angry Birds that popularized the device? That was the the unlock, right? It showed people how to use it. It was where it created demand. Currently AI doesn’t really create demand. It just makes things more efficient. So what’s the thing that’s going to make AI be this must-have consumer application? This must-have thing? And I don’t think they’ve answered that yet. And so to dive into it thinking that that’s going to be the answer to all your problems I think is naive.

ES: How do consumers stand to be impacted? So I think the impact on game development to date really has mostly been kind of represented by these layoffs unfortunately, right? But I mean you there I I do believe now there are some game developers who can make a credible case for how they’re like producing more stuff, more and better stuff as a result of the application and the integration of these tools. But how do consumers stand to be impacted? What’s the consumer-facing application?

JvD: Yeah, so it’s not great. So the front end of that is there’s just more content for them to navigate. Discovery has already been this endemic problem and AI at least in its current manifestation just adds to the problem. It’s just all this stuff for me to now figure out if it’s worth my time. And so you see this natural resistance where people don’t really like games that are very much reliant on AI for their creative output, right? And I think that that’s, you know, it’s it’s sort of like a natural thing, but so AI threatens to flood an already saturated market. And so now we have all this cheap stuff. It’s like, oh that’s I mean I guess it’s good to have stuff, but is that exciting enough for me to spend my time and or my money, right? And so you start to see the reaction where they reject it, you know, say like oh this is clearly AI-generated, no thanks, I want some kind of farm-to-table gaming experience, I want to know the human. And so that’s the natural next thing is where it’s going to be more influencer-based, it’s you know it’s going to be more about the talent behind it. It’s like I want to support this creative or this creator as opposed to this machine language model or whatever. So that’s the front, the back end, which they might not connect the dots there as much, but it’s already the case that we see a lot of the hardware and also the chipsets being bought out wholesale by these incredibly data-hungry and compute-hungry companies. And so it’s driving up prices for hardware for consumers. The reason we start to see all these numbers go up for handhelds, for consoles is of course because of tariffs and because of value chain inefficiencies sort of this global tariff debacle that goes absolutely nowhere, you know is exacerbated by the fact that you have these large AI companies spending billions to build warehouses for their compute needs. And so you’re now competing with Nvidia to buy chipsets. You know even the most forgotten SD manufacturers have skyrocketed in value. And so they’re all expecting the hardware components and the chipsets to sell well, which means that I as a consumer will have a harder time getting things at a cheaper price. And so both the discovery problem, the glut of content as well as the increase in cost is the first line of offense to consumers.

ES: Do you think that there’s a credible possibility that so I had made this point when the Genie 3 got showcased that, you know look it’s the whole point of it and we’ll get to the Genie stuff in a second, but like the whole point of a game engine is to make sure that rules are knowable. And you know they don’t have to be static. I mean the rules of a game can change over the course of a of a game, but they need to be knowable. And that’s just fundamentally impossible with like the output of of a Genie type model. And that’s just a game theory constraint. That’s not a video game characteristic. That’s a that’s a game fundamentally like a game as a broad concept. And and someone asked me like, well what about what do you think about the idea that people would just like to vibe play, right? So you know how like there’s these generated playlists on Spotify? They’re just random and people like listening to them. They’re popular, they get a lot of… and and yeah, maybe that’s maybe that’s an outcome. Maybe that’s something that players will will grow accustomed to or or which will onboard a new class of player that was not previously being satisfied right by the form that interactive entertainment took. They will find a way to become a gamer. This will onboard them into gamerdom because hey I actually just wanted to vibe play, I just wanted to kind of sit back and I’m playing this game and it’s not deterministic and so the rules change, the physics change, you know my character’s stats change unpredictably, but that’s fine. I I like that. It’s it’s very sort of like low mental cognitive load. Is that a possibility?

JvD: Against better judgment, I’m going to say yes. Because as you’re describing it, it sounds a lot like channel surfing. You know, like when I first moved to the US, I discovered this overwhelming aesthetic that broadcast television is in in North America. You know, I come from a small country in the Netherlands where you have like a few dozen channels and so like okay within minutes you’re back at the beginning. You’re like alright that was it, that’s that’s all there is to see right now. In the US you have 1200 channels and you could sit there for hours and so and you’re just kind of stumbling through different forms of content and that’s the process, right? It isn’t that I’m watching anything specifically by topic or theme, I’m I’m sort of just perusing and that’s the point, right? It’s like window shopping or you know something along those lines. And I think in your description like AI would allow for exactly that. I could just sort of like switch off my brain, like click on a bunch of stuff, sure why not. Oh look look at this, I’m a dragon now, wow that’s interesting. You know, it’s like so sort of real-time procedurally generated random experiences that sounds like there could be a market for that. I don’t know if it would be accretive in terms of like adding a whole another segment of the audience that previously wasn’t a gamer or wasn’t interested, maybe. I don’t think it’s going to be a billion people, but that there could be an audience for that where people sort of just passively play the same way that people are kind of watching Netflix in bed with one eye open by sort of like sure, it’s it’s the lowest form of consumption in some ways. I don’t know that that’s a sustainable revenue model for anybody that’s interested to investing in all that. That sounds expensive to build to very little revenue, so unless you do it at scale and and throw in ads or something but it doesn’t seem like it would be the immediately investable trajectory to go. And then I would argue that the fundamental thing that makes games and sports and entertainment in general so interesting is sort of the emotional connection. Whether that’s you and a character or you and other people or carrying a an avatar through a series of quests like in a sort of pen and paper D&D setting, you know, what’s the emotional connection or the opportunity to connect emotionally in an AI-generated universe? And I don’t know what that is. Like if I’m just randomly thrown into something, and let’s say I do have this deep meaningful enlightening experience and then do I get to save my game and revisit it or is it just once and that’s it? Is it some kind of, you know, I mean that would be heartbroken to have this immensely meaningful experience and then suddenly I can’t go back to it because it only existed for this one session. So I find that it isn’t necessarily compatible with how people play games and what makes games meaningful to them.

ES: Yeah it it feel it’s susceptible to to training a generation of nihilists just no connection to anything just you know servo-mechanisms. They’re just like pressing buttons for satisfy you know. Actually maybe let me let me kind of stay on the topic though. Fundamentally what is a game engine? When we talk about a game engine, so I grew up in Texas in in Houston but that’s that was the epicenter of game development in in Dallas, not Houston, but Dallas, actually in the suburb of Dallas like the Plano area is where ID Software was, it was 3D Realms was there, Ensemble Studios which made Age of Empires was there. I don’t know if you remember this John Romero’s God Games when he did the publishing and they did got Daikatana? Daikatana was great. Yeah it it that was interesting. Um they made the other game what was the other game? Deus Ex, Deus Ex was a fantastic game. But that was real. Daikatana was like a little bit of of a Frankenstein’s monster of a game, but I enjoyed it. It was very I don’t know if you remember but it was it was like it was very controversial that you know what what he was this big ambition and it got you know sort of whittled down but but and so I just but I was I was just awestruck by these people I these these sort of like these characters right like were sort of larger than life to me and I follow I don’t know if you remember the .plan system where you would do like it was a blogging system as .plan it they’d updated .plans and their all these gaming websites would have like the finger and they would go and it would they would fetch these .plans and they got updated and so essentially it was like all the game developers were blogging but on this sort of like on this non-centralized or decentralized platform. And so like John Carmack was doing that, you know John Romero all these everyone. Famously John Carmack after every release would just go and hole himself up in like a cabin or like a hotel room or something for like three months and then he would write the big update to the Id engine. And then the next game that came out whether that was like Quake 1 to Quake 2 to Quake 3 Arena which was like a big upgrade would be the result of that that just like solo kind of development experience. And so that’s kind of how I always envisioned an engine and it was like it was the visual upgrade, physics were better, like there was sort of like more polygons and and now you’ve got you know Unreal Engine is obviously like a SAS product and you’ve got Unity. In your mind the conceptually what’s a game engine? What what is the purpose of a game engine? What does it do?

JvD: So in the context of game development for me it’s a foundational codebase that makes up the interactive world. So it’s the middleware and it generates the two three-dimensional places and spaces in which we then create some kind of play scenario and that includes of course everything from rendering physics to audio, input and sort of the larger system, right? So yeah I’ve always explained it to my undergraduate students as like that’s the word processing software that you use to write your book, right? This is so it’s the thing that you interact with that that makes the thing. And so when it comes to engines in general they have specific needs they have specific iterations like if you describe it well I think that you know what made that era in gaming unique and particularly you know things like the Quake engine like in my mind was like that they were iterate and get better over time through the incredible amount of work and the clever approach of distributed computing and sort of like have this modding network almost where everybody would contribute. So this this is this collectively shared text, right? And we all contribute to it and that’s what makes it better. That’s what helps push the boundaries. That’s what progresses it from 2D to 3D or improves the lightning effect or makes it more realistic in some way. And I think that sort of collective human experience or production effort has a chance to be catalyzed or accelerated with AI but I don’t think it can be replaced, right? Because in that process as you described it I think you’re right is like at some point there needs to be some kind of moment where we just shut it all down and like a group of people or a person some kind of author some kind of architect needs to make some hard decisions as to what it can and cannot do, what the priorities are for the next batch or patch or version. And that I think you know is seemingly absent from the conversations around game engines now when you talk to investors. They think oh it’s this sort of everything machine, you press a button and you have GTA 6. Yeah not not really. I mean perhaps it’s best explained with like where Project Genie is great, right? At generating these explorable 3D environments, that’s cool, but it’s 20 frames 24 frames a second. You know it’s like can you imagine like that that is like that is some 1990s level quality video. So in other words it’s like okay so they have all that but then it’s also you know never mind it’s expensive and inaccessible right? It’s $250 a month for the Google Ultra subscription. So as opposed to a freely available shareware version of an engine or something that has like an on-ramp for like small indies that they can use out of the box that’s been honed over decades. So I think that it falls well short of what creators would need. I think it’s a great concept. It’s perhaps a good challenge that might make for a good plugin somewhere along the line. But I I doubt to see that it’s going to really be disruptive in that sense. I think the the the world-building models or the engines as you describe them like those are larger collective you know how do you say like a sort of collaboration between people and to have a machine kind of generate that I think that incurs probably more technical issues in the short term at least than it solves problems.

ES: Right and I mean you mentioned GTA. To me GTA 5 which is like endlessly replayable I mean it has still mid 100,000 concurrents on because of GTA Online right? But what makes it endlessly playable and from the single-player experience is just the humor. It’s so funny. And like you know someone the storytelling the narrative. I mean that to me is the essence of GTA 5. It wasn’t the engine. The engine didn’t even look that great. I mean I no I shouldn’t say that. At the time at the release you know more than 10 years ago now it did look pretty fantastic. But it really what kept me playing the single-player game which was I don’t know how many hours I mean it’s probably hundred hours or something to complete like it was just the humor and getting to your point like it was that emotional connection with the characters, right? Which I don’t know the other side of that argument that someone might take is like look you’re talking about the limitations now. Yeah it’s low frames per second. Yeah it’s only a few seconds. Yeah it’s expensive. But think about in a couple years all of like let’s just hand-wave all that away right? Like let’s just say that you know all of those constraints disappear because fundamentally they’re technological constraints and you can just they’ll things will get cheaper and dah dah dah. I don’t know that you get LLM say to write something that is that funny that it becomes a cultural phenomenon. You can get it to write something funny that like kind of hues to the current norm but can you get it to write something that funny and that compelling that it becomes a phenomenon because that’s by definition outside of the bounds of the the norm right?

JvD: Yeah it’s I mean it’s a word guessing machine, right? It’s it’s predicting the most likely thing to happen next constantly. And so what it does is produce a lot of mediocrity. And what distinguishes successful blockbuster game releases and pretty much everything else in that in entertainment is that they are new, that they bring something that is unexpected and outlier of sorts. And they do it not just on that merit alone, it’s also the entire production and marketing and sort of the the movement around it that has to speak to a generation. You know, and I think that that was part of what GTA has had as a franchise for a long time. It was sort of this counter-cultural thing, right? It’s it’s the game your parents don’t want you to play. And every news item on it was like, well but you can steal cars from people and shoot gun. It’s like okay that’s precisely why people play this because y’all don’t like this. It’s the same thing you see nowadays by the way with things like Skibidi Toilet, right? On Roblox, like I invite you to go venture into that for like an afternoon. You will come out of that say like your mind will have been scrambled completely. But that’s the point, that’s the point of it, you know. It’s a play space for like 12-year-olds that are just trying to like have their own online space where they can hang out. So what GTA 6 will be is not another game release, it is a platform upgrade. It’s an operating system upgrade, right? So for Take-Two or for Rockstar they use Rage. You can see the you know the same process you described earlier is like you can see the the breadcrumbs, right? It isn’t just that Rage works really well from GTA 4 to 5 to 6. You see also in all these adjacent games like Manhunt and you see also Max Payne. You see the targeting engine in one and you see sort of like the stalking and the other and then all of those bits and pieces and components eventually make it into Rage or they’re part of that whole effort. So that they don’t just make it better in the context of this one franchise they make it better across the entire company all of their work. And so I would welcome anybody to invest the computational effort to emulate or repeat that process, but I doubt that you’d come up anywhere near as funny or as exciting or as culturally relevant as what franchise like GTA have accomplished. And so for that reason alone it’s like yeah yeah sure you could do it but no one would care. It wouldn’t have the same meaning to it. There would be no emotional connection to it or at least not in the same way that we’re used to you know having blockbusters now. So I think that that’s the the ultimate argument in favor of creatives. That’s the space where they excel. Like they the tool serves the emotion not the other way around.

ES: Right so divorcing the you know or trying to trying to look then at the disparate workflows that AI can impact in the game development process versus just AI creating games becomes a lot more of of the task, right? So like so how do those creatives get superpowers? Right? To to do more work and and push out more content but that is directed by their skill versus like do I type in a prompt and get a game? Like I just it just feel like no one in in that’s in that’s currently where we’re at in the AI pessimistic case because it’s like oh well you know I don’t know if you saw yesterday Perplexity released some new product and everyone and everyone say oh well this you know the Bloomberg is dead because Perplexity’s has just replaced it because you can now just write a prompt and recreate Bloomberg, right? And it’s like no that’s not you can’t do that. That’s not actually replicating what Bloomberg is right and everyone’s kind of coming back saying like that’s not if you think that Bloomberg is the charts then you don’t understand what Bloomberg is used for, right? That’s why people pay $20,000. That’s a great metaphor. That’s exactly that’s exactly what it is with games too. Right and so it’s like what actually think about how do these tools help you know whatever buy-side analysts do better analysis? Right? Don’t don’t think about this replicating the software that they use with a prompt think about all the specific decisions that they make every day that maybe are like better as a result of these tools. And for a game you mentioned Max Payne and I remember like when I lived in in Helsinki, like those people are seen as like the artists, right? Because like Helsinki was kind of like a free-to-play games town. The Remedy people were over in Espoo and they were like the artists like they were very artisanal with how they approached things and they didn’t really have this kind of free-to-play sensibility, but like they were all extremely extraordinarily talented. And if you look at a game like Max Payne, I mean that to me was a masterpiece when I played that. That was one of my favorite games of all time up to that point. And it just took something that just the dark character that game could have only come out of these truly skilled people that grew up with you know three hours of daylight in the winter like it just it just emanated from the depths of their soul. So like you’re not I don’t know. I mean I don’t want to be like minimizing like the the growth curve here of these tools but like at the same time I just don’t know how you could recreate that.

JvD: I’m with you on this because I mean so let me start by echoing the same sentiment. There’s a Joost version before and after Max Payne 3. You know I just it changed me as a human. You just you’re like bye you hit the edge you’re like yo. I had to I had to sit with that for anyway but so that comes from the storytelling, right? Of course that was also the third iteration and here’s here’s the thing with all this tech babble is that what it does is like yes it’s important of course and it’s favored by investors because it’s such an easy thing. Like if you have the gizmo then you’re better than the company that doesn’t have the gizmo whatever that might be. But it becomes almost the only thing to a lot of conversations where it’s like like you say cultural background and relevance, geographic location, you know distribution innovations etc etc. What’s the business model look like? You know can when you run several cents per conversation does that actually suit your revenue model as a company? There’s all these other questions that also have to be answered. And so like yeah sure maybe AI is having a moment running ahead and at some point we’ll get that but it requires that all these other factors and variables have to catch up also. You know it’s never just this unilateral thing. But this is I mean I guess back to my book a bit but this is always the case in these creative industries. Like the what suits the investors and the tech companies is just overwhelms every conversation. And it minimizes everything else. And so then you get exactly this because it’s like how do you invest in Finnish artists? Like that’s not an investable thing, right? Only when they figure out how to do user user acquisition really well because you can measure that. You can show increases and so in other words I find it encouraging that these creative components are also sort of like the free zone from all this. That’s the area where a lot of this can’t go and so that’s why a lot of these engines will come up with bunk and just slop that nobody finds interesting. And it won’t be a life-changing thing. It won’t have the cultural cachet, it won’t have the background from which the really meaningful experiences emerge. And I think that that’s a that’s the same for AI as it’s been for everything else, like VR, the metaverse for that matter. Like these are all great ideas if you’re a technologist and an investor. But how do you actually make this make sense in the real world, right? Like who’s gonna pay for this stuff? Are we going to see AI being distributed freely as a common utility for everybody across society or is it just going to be something for people that can pay for it? Is it going to be like different consumer classes for this thing, you know? It’s like how do you make revenue work here? And so the larger conversation you have to create demand as you create production capability, right? And this is sort of the example of Ford where he says sure we can build all these cars but now we also have to make sure that people have jobs and can actually afford the car. Otherwise we don’t build a cycle here. And so what AI is doing really well right now is creating a product and justifying its own existence but it’s terrible at currently is creating demand. Like what’s the consumer base that they’re imagining here? And I haven’t heard a strong answer to a lot of this.

ES: I read recently about the background or the Michelin. The the tire company? You hear the story? You familiar with this? I think I know where you’re going. Okay go ahead. So basically I think it was like it was a family business and it was they made tires and it was struggling right and so what I think the son inherited it or whatever and it was like on the brink of bankruptcy. And so what he did was he created the Michelin Star system to encourage people to go on road trips because he had this Michelin travel guide essentially that was the source of this Michelin Star system to then go rate these restaurants. Well if people needed a drive to go visit these places, so these Michelin guides, these tour guides gave people the reason, created the demand for the tires to then go and you know visit these places and yeah I think you’re right. That hasn’t happened with AI yet and all we’re hearing about is like how some you know non-trivial proportion of the economy will will be sort of economically thrown overboard and and well who’s going to do the buying then? I don’t think that question has really been answered in like the very macro sense.

JvD: 100%. I mean I could talk about this for hours but the the encouragement here is like of course that you know the games industry is always one of the first ones to experience these changes, you know it’s always a good lens and I think what is forgotten a lot of the times here too is that and we mentioned this it’s been going on for 20 years in the games industry. This is a conversation that’s not new, this is been this is endemic to game development. And so I think a lot of these industries would do well to kind of take a moment say okay how do we integrate this the slow way? Like everybody’s rushing towards some unknown future when in fact you know the games industry’s been leveraging these tools for a long time. And they’re very piecemeal about it, you know they’re they’re very I mean why wouldn’t Rockstar use AI 10 years ago? It was such an incredible productivity boost, right? In an industry with very little like you know labor laws and a lack of unions across the board and you know headcount being the most expensive piece in any kind of studio, why wouldn’t they have embraced AI for like this massive productivity increase? Well there’s a reason, there’s reasons for that. It’s because it doesn’t work. I’m I’m curious to see if they can prove me wrong in this current moment but feel like the excitement is overwhelming sensibility.

ES: Yeah I just keep coming back to you’re talking about Max Payne, like Disco Elysium. You can’t make that game with AI. Like there’s no ChatGPT is not creating the narrative structure for that game and like the dialogue for that you just know it’s not going to happen. It’s not going to be as dark and gritty as as the actual game.

JvD: Oh yeah. No it’s yeah I think people looking in the wrong corners for these things. I think it’s a you know there is definitely… so here’s another way to to look at that is like I had the same skepticism around the metaverse and VR is like the companies that are currently leading the charge for this allegedly disruptive technology are the incumbents. It’s just a naive to think that they are going to undermine their own business model. It’s just not going to happen. They are just not doing that, right? It’s it’s a land grab, sure, it’s a you know trying to elbow themselves to the front and of the line. But they are not these lean clever companies that are going to uproot an entire existing industry in which they have dominant market share. And so you know we’re going to see I mean it will produce exactly what it proposes which is just more of the same. That’s exactly what AI does.

ES: I want to end actually maybe let me shift a little bit into assuming that sort of most optimistic outcome takes shape, right? And so like what types of game experiences or what type of sort of productivity gains do we get from AI that sort of support the you know core creativity of interactive game experiences while also maybe just making content cheaper to produce and let’s say that you know GTA 7 takes a year to make, right? As a result like so if that were if that were to happen and we come back this time next year and it’s like hey GTA 7’s already live and Rockstar said you know we didn’t lay anybody off, we just supercharged everybody with AI and and it just so happened that we could make GTA 7 in in like in like six in like six months we were able to make it versus you know the decade-plus. What would have been the use cases that were found? Like what what is the most optimistic sort of outcome with with respect to game development?

JvD: So the precursor I think of is the printing press. I want it to be a printing press, right? And by which I mean back in the day before the printed word you had manuscript culture which is dominated by the church in you know in Europe and so like in many ways it was quite literally a direct line from you know heaven into monasteries where you had scribes and priests and they were the only ones who could read and write and they would tell the story of the Lord to sort of everybody else. And so they dominated basically the spiritual experience through these means of production and distribution. The printing press came around and all of a sudden anybody could print and people learn how to read and it scattered and crumbled a lot of these like existing institutions. All of a sudden the church lost a lot of power because now we have much more smaller and more localized versions of whatever spirituality is, right? It’s not this monolithic one one-dimensional thing. It was also the cause of like the nation states emergence in Europe and it created a whole bunch of free thinking, right? And it was initially considered very weird like why are you sitting there quietly staring at a page with ink? That’s weird, nobody does that. You go to church and you go talk to people, you go listen to the sermon whatever. But it’s over the course of several hundred years it slowly brought to its knees this institutional power and changed it for something I think is possibly better, right? It’s the Enlightenment age was accelerated by this and you know a lot of new ideas came as a result of this. So we’re currently probably in some version in some early year of that with AI is like okay maybe this is the thing but in order to do that it has to be democratized. In order to do that people have to be literate in this. In order for this to work it has to be basically free to use in the same way that a book might cost a little bit as an entry fee but it sustains and is available and you can give it to your friend and they can read the same book at no additional charge, right? And so I think what we’re going to see is AI is going to initially be sort of the show pony of a lot of large tech firms and then over time become disruptive to those same tech firms when people figure out how to do it on the cheap. When people figure out how to run you know not large language models on these cloud servers that cost a fortune, but something that’s local that runs on existing technology. I’m a big fan of small language models for that matter because it allows to take place on edge computer and on edge devices where I can now run an entire seven billion parameters on my phone as opposed to seven hundred billion parameters in the cloud where I pay 200 bucks a month for the access. And I think once we get the small enough versions of this, once we get the low power FM versions of this where anybody can read and write I think that’s when it’s going to be really disruptive and then you’re going to find that a lot of these companies building their models are going to have a horse have a hard time making their numbers because people will naturally pull away. Say like you know look I get it, it’s important, but it will be the you know the decline of their institutional power over time. So that’s how I see it play out over the next give it 10 years 20 years.

ES: You have students now? What do you tell them? How do you prepare them for the industry that they’re going to be released into?

JvD: That’s a good one. And so I allow AI in my classroom. At Stern they gave us the option, it’s either you do, you don’t, or something in the middle. I said yeah I encourage it. You need to learn how to use this. This is not going away. I do believe that there’s a sustained… so I believe in a future that has AI as a a big part of it. It’s just we disagree on timeline and application. But whatever. And so students I tell them you must learn how to use this in the same way that when I went to college one million years ago, nobody asked me if I could type on a computer. If I had that figured out. But my mother got a secretary job in the 60s or whatever. To get that job she needed a typing diploma. Like you had to go show that you could type 60 words a minute to get this secretary job and it was just typing more words per minute or whatever. And so but by the time I hit college nobody cared because it was sort of assumed that you’re a master of that part of life. AI will go the same way. You’re sort of able to navigate this technology on your own. How do you leverage this? How do you use it? And so I encourage them to use it, but not be lazy about it. So the fundamental component is like I see my students so often from countries where English is not the first language, and then the moment they start writing papers, all of a sudden they’re prolific. And it’s this beautiful language. It’s like, okay, you’re overusing it a little bit. Like you have to master it or the horse is running you, right? It’s you’re you’re not in charge of the technology per se. So I encourage them to learn about it, I encourage them to use it and use the confines of a classroom to kind of experiment and find out. But it’s definitely, you know a permanent fixture in their career. And I think the better you are at just knowing what it does and how to apply it… you know can you really generate a P&L or financial analysis if you’re going to be a sell-side analyst or an investor using AI? You probably could but you might as well just Google things then you know for a living and then base your decisions on that. So like you probably want to be in charge of some of the fundamental components. And that’s where I think it can be a catalyst but not a solution in and of itself. So that’s the conversation I have with my students and they they take to it well. The papers have improved in quality structure overall writing. The creative thought seems to fly as a result. They have a much better sense of confidence. They can really do something that they couldn’t do before because they can generate an outline for instance very easily. Now they have to control it and not be reliant on it, not become exclusively dependent on it.

ES: I yeah that’s that’s the discipline that all of us are going to have to learn. And I think it’s a barrier right? It’s a barrier to being productive. That’ll be I think the determinant right of success for a lot of career paths. Can you utilize this in a way that’s additive and not become overly reliant on it to where you’re just one among many right producing similar quality output? Joost I embarrassingly mispronounced your first name when we started. I asked for confirmation over your last name’s pronunciation not your first name at the outset. I apologize for that. Where can people consume more content from you?

JvD: On my Substack, the SuperJoost Substack is one of them. You can find some of my work at Aldora.io. Or just Google SuperJoost, that’s probably the easiest way to do it. I’m headed to GDC if you want to say hi, and I’ll be at South by Southwest to do a talk as well. And big fan of the show, I always appreciate everything you do too. So you know, let me buy you that coffee or beer next time we find ourselves in the same geography.

ES: Next time we collide we shall have a coffee or beer. Joost thank you so much, this was fascinating. Really appreciate your time here.

JvD: Thank you.

Comments: