Why Traditional Game Dev is Already Obsolete

Guys, I think we should just cut right to the chase, get right into it, and really ask the question I think a lot of us have on our mind, which is, okay, there's been a lot of talk about AI. There's a lot of people with different opinions of how big of an impact AI will have on gaming. So, from your perspective, how far will this get to you? How important, how revolutionary will AI be on gaming? How far do we go? Erin, what do you think? [Music] What's up folks? This is Joseph Kim from GameMakers. During GDC, Hora, the server orchestration for multiplayer games provider hosted live podcast sessions on various gaming topics and I was able to moderate a panel on AI native games. What does this mean and what do we believe the impact of AI will be on the games industry? So speaking on this topic, we had Aaron Farre, who is a former Riot engineering manager and more recently head of engineering at

Singularity 6, but who is now currently the CTO of Jamnt Studios, one of the very few game studios building a completely AI focused game experience. On the studio side, we also had Liam Dean, principal analyst with a focus on gaming at Omdia, which is a global analysis, intelligence, and advisory company. So we had two folks who are very deep in AI and gaming. One from an operational micro level and the other at the industry and macro level. We had a great discussion and you're going to want to catch this if you're interested in AI and games. So stay tuned for a super deep and insightful discussion on AI native games coming at you right now. Yeah. So on the question of like what impact is AI going to make on the game industry uh I think it is going to have an impact. I I am very confident and bullish about that uh but I don't want to overstate it in the sense that I don't think that the way we make that

that the kind of games we have now will disappear. We're going to add and create new type of games. We're going to have new type of development processes. Things will change. Uh so I do think it will make an impact. Um but you know we'll still recognize the kind of games we have today. there'll just be different games as well and things that we can't do today we'll be doing. Um, so I do really think it will make a pretty big impact, right? And in terms of some of the the ways in which AI will impact games, there's kind of like this cost issue. There's also new player experience issue and then there's also this issue of potentially like automating away stuff. During this GDC, I have uh talked to two studios who are building like completely AI automated game studios, which I find hard to believe. But like to your point on just adding to what's there, can you talk about some of the current limits that you're seeing and why you think it just adds instead of going fully to like super intelligence replacing and destroying humankind? Yeah. So

um destroying humankind um so I want to acknowledge that within machine learning there's a concept of like the bitter the bitter lesson of machine learning like that we keep we hit the limits of of neural networks multiple times and we kind of have various times we get into an AI winter where it's like all right you know all the hype it didn't do it and often the limitation what we found is that we just didn't have enough data didn't have enough compute few years or a decade later machine learning deep neural networks come back with a vengeance and prove prove the detractors wrong. So in that in that spirit um I don't know if there will be a moment when we can AI up a game from scratch. I am currently skeptical. Uh and I think for those of us who work looking at the impact today uh there is value to be had and uh for players experience to be had where we don't necessarily need to wait until we

can have a super intelligence to take advantage of this technology to make better and more fun games. I think that as an industry we are a cross-discipline industry. It's one of the reason I love games. So we're always adding new technology, new techniques from across technology, from entertainment. In fact, sometimes we're pioneering a lot of those areas in the actual applied sense. And so we're going to keep bringing this in. Um, but I think it will become more of a rep like we have this repertoire of tools. It's going to add, right? Uh, the question for most developers is where is there an actual ROI? Where is this tool better than the tools I currently have? Right? And I think that's the question most folks need to ask themselves, right? And I think that to that the point that you're making on the the return, I I think the uh the CEO of Anthropic calls it return to intelligence, but like that question of where do you get your RTI is an important concept and for you the limitations that a couple of things that you mentioned is compute and data. So if we're like trying to assess or gauge the

limitations of AI moving to the future, would those be the the main things that you look at are the like the limitations that we're starting to see from from a compute energy data kind of perspective? Yeah. So uh to be clear for for GMT, we've been focused on how do we apply generative AI in the game loop, right? How do we make game how does this make new type of novel gameplay possible? Um we're also discovered some things about productivity improvements. I'm sure we'll talk about that a little bit. But um in that note, when we're doing real time inference, it is a very computationally intensive. Yeah. And we're very hungry. Uh we have a lot of demand for it. So you know in the game industry, I mean we all know like there's always a new GPU, there's a new CPU. We always come up with reasons, right, to use all the compute possible. And so I think that will stay true also for inference right uh for machine learning we will come up with re like even if you made it drastically cheaper

than it is today I have I have a backlog of more things I would add into it and that would be true for a faster GPU a faster CPU the AI is going to have the same issue so we are currently largely limited uh by those costs and that is something that at JMT we attacked pretty aggressively and that if we want to Liam Dean So we have a perspective from someone who's on the ground building stuff and from your perspective more on the macro basis as a mark doing market research and looking at the industry how revolutionary will AI be on gaming what are your thoughts uh I think I would have to agree with a lot of what Aaron has just said there um the limit is hard to predict because it depends on exactly what you're saying we don't know how far we can sort of push this architecture that we have that's got this breakthroughs of the last few years. But what I think is it could stop right now. Like we could see no more progress whatsoever in the foundational models

and we would still have a massive amount to work with. We would still see huge progress in games over the next few years. Just figuring out how to operationalize that core technology that we already have, turn that into tools that are effective for game development, getting everyone to work out how to get the most out of those tools. So I think we will see a massive amount of progress regardless of what happens at the the most fundamental level. Of course, if there's more progress in the models themselves, that will continue to filter down, but we're not actually dependent on that. If it turns out that we're kind of getting close to one of those limit points, even then, I still think we're going to see massive progress in AI in games over the next few years. Okay. And I I think there's been a lot of discussion about cost automation impact on people, but there's been less discussion on uh the new player experience is something that you touched upon, Erin. And so one of the things I would love to hear more from you guys is

this notion of what I'm calling like an AI native game, but fundamentally a game experience or an experience for a player that is fundamentally enabled by AI. And when we talk about that, can you guys talk about what you think are some of the capabilities or potential of having an AI native game? And I know Erin, you're working on Retail Mage, which potentially could be something like an AI native game. It'd be great to hear more about what what that is. I I will absolutely contend that it is an AI native game. Um, but, you know, we'll decide what that term actually means. So, what I think it unlocks, let me uh describe a couple things real quick. Um, we tend to talk of it as what it unlocks as improvisational play. So that the game simulation plays back with you in ways that we didn't before. Uh, by, you know, games have always been dynamic, but there are still limitations. And we can kind of think of the way that if properly integrated, the game system

itself almost acts like a proxy of the designer. Kind of like a dungeon master. I know people use that, maybe overuse that analogy, but it's not that wrong. um where you know anytime we close that feedback loop between player and designer and make it faster. We've seen major innovations in the game industry whether it went from re you know box sets that had to get you know discs pressed to live service games to rapidly increased um uh updates. This only increases that cycle and you have this kind of proxy of a game designer sitting there can modify the game state and game according to how the players are playing and that unlocks a ton of things. One of the core game design concepts that we think it creates and we are stealing a a concept from the tabletop industry here is rulings not rules. So in games typically games we have hard-coded rules that we have to follow and we can create some heruristics and a bunch of other things but ultimately is a set of rules that we follow in a game. What we can create is this more fuzzy logic of rulings where the system can look at the

game state and make a ruling on that in a way that we would not have been able to program before just never been able to program before. And we see emergent gameplay and give a few examples if we had the time in in uh retail mage uh that again would never have seen that type of gameplay before. And uh retail mage is more of a tech tech kind of showcase of what we think because when we did earlier technical demos, you'd see these amazing things and you have these AI agents and a bunch of stuff going on and we have this procedural generation. Um but a lot of folks were like that's amazing but you know I come from a game background. How is this a game? And so Retail Mage was our first attempt to show an early middle ground of some traditional game mechanics and AI together to create a game loop that does things that we didn't do before. And could you like help us understand that a little bit more concretely? Maybe in the sense of like in retail mage, what does a core loop look like and how does AI integrate within that core loop? Yeah, so retail mage is a party game

kind of dressed up as a job simulator. I like to describe it as latestage capitalism comes for diagonal. You and your friends are wizards at the mage mart and or m yeah mage mart. and um uh and in come some characters. You got 30 minutes to give uh solve these customers problems and they'll give you one to five star reviews. And uh one thing that was important for us, you see a lot of like dynamic dialogue and things like that. We have that. We have our own AI stack, but we saw dynamic dialogue as merely a prerequisite to create the other types of gameplay we have. So we also have a AI world model. When you go up to an object in retail mage, uh you get a text box. You have a default thing you can do, you know, like pick it up, put it in your inventory, but you can type what you want to do and we will try to make that happen. Anything you want to do. So, as an example, uh in brief, there was a character, a wizard who came in, wants inspiration for his spell. I walk around in the store, there's a

painting on the wall, and I'm like, "This is totally it. This is going to inspire you." And he's like, "Yeah, yeah, that that that's that's the painting. That's what I need." Well, um, for various reasons, we have to like take it off the wall, put it in my inventory, hand it to complete the sale. Well, we have an RPG mechanic, so I have like a strength stat and everything. I'm too weak to do that. I cannot pull it off the wall to sit like, no, you're not strong enough to do this. So, I'm like, well, how do I make the sale? So, I run around the store. I find a bookcase. I take a book off the bookcase. I pull a piece I rip a piece of paper out of the book. I write IOU on it. I go back to him, hand the IOU to him, and say, "Here, we'll ship it to you." And he's like, "Perfect. Thank you. Four-star review." None of that was programmed ahead of time. All of that was emergent. Wow. Right. So, the the system determined had to determine, did I have did I make the case from a narrative control perspective to make a ruling that this was the right thing to do? Can I rip a piece of paper out? There are so

many I have dozens of these type of stories where we see emergent gameplay that we never would have programmed before. Uh that allows people to push the story within bounds. We do think there are bounds we want to keep. Um but the way that they want to go and it plays back with them in a way that we couldn't before. For a lot of games there are specific win conditions and it kind of sounds like in this case it's more about having emergent experiences. Yeah. And having more experiences rather than winning. Is that would you say that's fair or Yeah, in retail mage we tried to find a middle ground and some things we did and some things I mean we learned things not to do honest in retail mage too of trying to like find a hybrid between these. Um but we did want to create these more creative experiences. We take a lot of inspiration from tabletop games, but also from experiences like immersive theater and improv where you're having this participatory experience between the consumer and the producer of the content in real time creating together and that can there can still be win

conditions that we're moving towards um traditional ones or new kinds of more fuzzy win conditions um but the experience is very different. How I get there, the journey that I go on can be very different. I think there's an interesting comparison here with if we look back to when uh online multiplayer games first started emerging. I think what makes that compelling as a format compared to traditional single player games or what did make it compelling is the unpredictable and emergent nature of playing with other players and that that created another whole segment of games a different kind of game that didn't really exist meaningfully before. But of course to your point as well, it didn't mean that traditional singleplayer games went away. We still have them and some people prefer that experience. Other people prefer the online multiplayer experience. That's actually turned out in that case that that online multiplayer experience in one form or another is now the dominant form in the market. But it didn't cause what we had before to go away. So it can, you know,

that's at least there's a potential here for something to emerge that's a different kind of game that allows people to engage in a different way, but it will because it's different. It will be to some people's taste and not other people's taste. Um, but the interesting thing is, is this going to be the kind of thing that hooks people who don't currently get hooked by the kind of games that we have? like does this add on another segment to the market or does it just kind of cannibalize existing formats? And that's something we it's going to be really hard to know until we see some of these games in the wild. What kind of people play these games? Are they people coming in from outside the games market? Are they casual players? Are they people who kind of until now got their kicks from playing against other humans, but actually maybe AI agents are more fun and they don't call you as many names? Um so it there's a lot of kind of known unknowns there. Um but uh there are you know some of

these scenarios are really pretty revolutionary for where the market could be. Right. And Liam I I think that when I try to think of at least the companies that are public about being AI native you guys are certainly one of them. You've got Series Entertainment. You've got a few others out there. But I can only count a singledigit number of games that are at least public so far. And the kinds of AI native games that we're seeing might be like yours where there's potentially emergent behavior and kind of interesting kind of dynamics from a user experience perspective. There might be games like series where they're using AI to like create content for them and it's more on the back end rather than having a different player experience. I I guess a couple of questions. What are you seeing right now and what kinds of AI native games do you think will emerge in the future? And I I think the one example that I've been waiting for for years is like the you know you mentioned Dungeon Master like somebody I'm sure someone at some point will make a

Dungeons and Dragons AI native game where the AI is the dungeon master. Uh but wanted to like hand it over to you and get your thoughts on all this. Yeah, someone is going to do that. Like it's h it's such an obvious concept and it seems clearly within the limits of what can be done. So this is just going to come down to you know we're going to see someone try that on a scale where it gets in front of enough consumers that we we have a sort of a test case there. Um but you know how well are they going to do it? How well is it going to work? How are people going to respond to it? You we have to actually see it. Um, so that that's just one of those things we we need to kind of have people out in the real world kicking the tires on it to see does this concept work as well in reality as we're imagining that it will, but I have no doubt that we're going to see it. Um, and to your other point about there not being that much of this activity visible, um, I mean that's

definitely true and um, just the fact that the reality is that AI is a controversial topic and it's especially controversial in the games industry more more so than it is kind of in the rest of the world and you know it's not lacking for controversy there either. So I think there is more happening under the surface than there are companies who actually want to talk about what they're doing. Um they don't want to risk blowback for being a company that is doing AI experiments that some people are going to get upset about if they don't have the product at the end of the day. Do you guys want to talk about this a little bit? Do do you want to do you want to put on your We were kind of joking like when we talk about AI ethics, you got to put on the legal face and then you get go into like a boring discussion. But I think with you Aaron when it comes to AI ethics or uh rights and things of that nature what what is your perspective on that? Yeah. So a couple places of from where we come from um we are game creators. We are creatives. We want to create a healthy

vibrant future in this. We think this is a way that may enhance our our ability to create experiences but not replace it. Um, we have a narrative writer on our team, right? Like that that has not that's not a job that's been replaced. Um, we intentionally currently we we do not use AI generated art. Um, we are largely using uh large language models or other kinds of similar uh machine learning models. There are some important questions around the use of fair use for that date training data. uh we see interesting and we keep track of uh models that are have a bit more uh under bit better curation of their data sets but it is not a clearcut so there's the legal questions and there are then the ethical questions right uh you know my background before games was actually in open source where we really believed a lot on how we use technology and the

legal frame to create communities and create public good but it is based on a clear understanding of copyright law and how we work together. So I hope and expect that we will evolve in this perspective. We also take a pragmatic choice that these are technologies are very powerful. Uh sometimes described it as a bit of a Napster moment right like there are market forces that will push some of these technology innovations into our world. The question is I wanted as a creative in the space to be part of that conversation to guide it into a place that I hope is somewhere I want to live as a creative um and not merely have it be I think one of the one of the things I worry about in our industry is is if we're not willing to sit at the table and engage and figure out how we move forward with it. those choices will be made by people outside of our industry and the and the economic and market forces may adj may adjust our industry in ways we don't want. Okay. Do you have any thoughts on this or Yeah, I I think there are some really good

points there and in a way it's become more difficult to have these conversations because it's become such a polarizing topic that people increasingly feel pushed to take one side or the other. I think there are a lot of people out there who maybe a year ago at this event were a bit AI curious and now they've they've shut themselves off to it because it's just it's become more difficult to take a nuanced view on it. Um and I think you you've made some really good points here on you know we we need to have a practical conversation about this to make it something I think that is attractive to developers. Um, you know what they want to hear is how exactly is this going to not just make my job easier, but make the games that I'm making better? How am I going to take this and use it to actually make a better product, make a better experience? And I think it's time to have those kind of boring conversations about how it's going to work. um because that's really the only way that people are going to see how it benefits them

rather than just something that um you know maybe maybe it's going to take their job or maybe it's just going to create a kind of work that they're not interested in being involved with. So you know let's see how we can actually get to the place that developers want to go. Yeah. All right. Well, maybe so shifting away from the boring part of that conversation, I I think the the uh the I I think the part that's not boring is when we talk about loss of jobs and things like that, but we can get into that. But maybe before that, I'd really like to get your take because, you know, I'm I'm an AI maximalist. So, I believe AI is going to just fundamentally be bigger than the internet. How much bigger is it? 1x, 10x, 100x, I don't know. Are we gonna get to super intelligence? Who knows? But I do think there's also with any technology shift, with any new platform, there are winners and losers. And I think that we can at least extrapolate even today who would

potentially benefit, who wins from um from AI and the impact of AI in terms of game game studios, game developers. And so what do you guys think? What are the kinds of companies or the kinds of people that will win in this environment of AI having an impact on games? I mean, the obvious answer is the people who work it out first. Uh, so and an interesting thing to talk about is we've uh we've gone around the topic a few times now of AI native games, but there's a separate concept of an AI native studio. And that is something that we're definitely seeing happening. The studios are being founded and they're building their whole tech stack from the ground up around AI. And they're not necessarily making um a completely novel AIdriven experience like Aaron is. They could be making a traditional game, but they have built it from the ground up to be built with AI. Is that true though? Can I can I challenge you on that a little bit? Because Okay. So a lot of the folks I

talk to in bigger companies are like like you know they they basically say I have no time to to learn about AI. We're going to fast follow. We're just going to wait and see what happens. And then the startups you know like we have a person dedicated to AI like an AI engineer product manager type of person who's actively experimenting with stuff and trying to integrate AI into workflows. But when I talk to other companies ex except for Chinese companies, okay, the Chinese have got their together. Okay, but like US startups, many I don't know of any that are really pushing AI. They just kind of trying to they're they're actually just out trying to get more money to finish their game, right? Like we were talking about Look, I mean, yeah, that's completely the point you're making about the big companies is absolutely right. They don't want to lead on this. They want to follow. Yeah. Um, and I'm not saying that this is like the median

startup studio right now. Most of them are still following traditional models. Uh, but there is there is a new model emerging there. And you someone has to be the the guy that everyone else ends up following. Um, and it is easier if you don't have all of these legacy tools and workflows to work around. So there there is a model emerging there. Um, and I I think we this could be the year even, you know, within 2025 that we, you know, we see kind of the proofs of concepts start to emerge. You see something come out of um a studio that you wouldn't have thought could have done an experience of kind of a particular scale or quality and that kind of that could be a big moment for catalyzing people to kind of actually start to make the effort to make that change at their studios. Yeah, I to add with that I absolutely agree. I don't think we could have done what we did in the last year or so uh if we weren't a startup where we started from this from first principles. It's not to say other

organizations can't do it. There are lots of reasons why it's challenging. Uh but at the same time, you know, Retail Mage, when I was here at GDC last year, if you had asked me if I was going to ship a game that year, I would have said no. We conceptualized and built Retail Mage largely from scratch in 5 months. Now, it's not a huge game, but it is a multiplayer uh Unreal 5 game with a AI back end. And we and we didn't even also set out to largely leverage a bunch of AI pipelines, but the integration of that technology absolutely made us move faster. And we're a small team. We were eight people when we shipped that. Yeah, I I I'd have to agree with you. I do think that we are going to see because of AI a lot faster product velocity. And just to give a little color of my own experience, our studio, we've been getting our ass kicked for over four years working on this like mobile extraction shooter game and uh we're probably going to soft launch in 3 to four months. We just started an AI

native game and the development cycle there is going to be 8 weeks. And uh whereas our company is like 44 people, we have two part-time people working on that game now. And so to your point, like I I do think this is something that's actually occurring right now for a lot of companies. It's just the the ability to create some kind of game, some kind of experience very very fast using AI. Yeah. And to your earlier question around you know what companies are taking advantage what's the ultimate impact from an AI maxist perspective I mean ultimately in our industry what I hope what I believe will be true is that the is players will decide right and most of them don't care they don't really care what happens behind the scenes they just want a fun experience um what I think will happen is I mean we already have a a lot of content available um for some we'll see is this a type of content that attracts either different players or others. I also suspect that it will continue the trend

that we see with Roblox and UEFn that the barrier to entry to creating game content will continue to get lower, right? That will have a number of impacts. In fact, I'm not even quite sure what that world looks like completely. Again, I don't think that that we'll we will still see AAA games, right, that are that are unattainable without large budgets. Um, but I think it will supercharge some of the indie movements to be honest. Yeah, I actually have a concept which I call like large vector to small vector development. But to your point, like by vector I mean like time, resource, budget and to achieve certain types of game experiences, it used to be you need this big large vectors to get there. And to your point on Roblox, what Roblox has done is by encapsulating a lot of functionality, it opened the funnel for different kinds of creators to create and made that large vector into a small vector. But I think that's also to your point what AI will do is to

similar to Roblox open up the funnel for the kinds of creators that we see, but also turn what was previously large vector to small vector. Um but maybe I can just to push on this a little bit more in terms of the kinds of winners and losers that we'll see because this has always happened, right? Any big technology disruption has always created winners and losers. Who do you who loses? What do you guys think? I'm going to swerve the question a bit here. I mean, I think the premise is absolutely right. Um, and I also think that I don't think there's any reason to think that structurally the the games market or games development is going to be worse off as an industry because of AI. There's no reason there should be fewer people making games just because it's possible to make more game content with AI. But the shape of the industry could be very different. You know, there is there's a shape of game and a shape of studio built around what we now understand as a AAA experience. Will exactly that kind of scale experience um

or that kind of scale studio be making a game that's 10 times bigger in a few years time or will there be no real market for a game like that and we won't have studios at that size anymore. So there, you know, we're too early to be honest to say how that's going to shape out, but we have a kind of a set of categories that I think is not going to map on to how things look when the amount that you can produce per person looks. So maybe I could ask you guys then because I have a specific theory. Maybe I could get your reactions to this theory. And so I believe that there's this discussion about the impact of AI to organizations talking about how today you've got like these fixed organizational structures and with AI we will move to a completely dynamic structure. Now I believe that's incorrect. I think that in we're going to get to a midpoint where there's what I'm calling an organizational meta meaning there are certain limitations strengths and weaknesses and based upon

um what I believe will be like the the collapse of teams and the collapse of roles that you're going to have a certain organizational meta such that if you have a specific team structure and orientation you will win. But the companies that cannot adapt to this, like big companies that are very fixed with monolithic game development practices like uh well, they're they're pissed off at me anyway, so I'll just say Ubisoft. Um you know, then that those companies are dead. They are just completely gone. What do you guys think? I think you might be right, but I mean you could have said similar things about, you know, look at the internet that that killed a lot of big companies that couldn't survive the shift. H but also a lot of those legacy 20th century companies are still with us. Some of them made it. Uh so it's very difficult to point at any one big lumbering organization and say that one won't survive the shift. Some of them will, some of them won't. Um but it's you know

it it's definitely likely that not all of you know actually it's basically certain that not everyone is going to make it but who exactly is going to make it um you know we can't call that right now the I'm not sure how it's an interesting question around the the dynamics and organization uh a dynamics of an organization and how it responds but I do think that it will there will be a difference in organizations that can keep up with the speed. So, we've talked, you know, I've described I describe it as like improvisation or participatory media and and there's really two aspects of it. One is the DM aspects like in the game, the game's playing back with me. It's more participatory. It's improvisational, but it also decreases the gap between the player and the developer because the cost of incremental content is so much lower, particularly once you have these things that now. So, you know, we look at changes in other media where the cost

of content changes the relationship between between the consumer and the producer. You you don't get diss tracks from Hollywood because you can't rapidly create a featurelength film in order to respond to that kind of conversation. But as we've created other forms of video that you get to like Tik Tok and YouTube, we do see that type of back and forth. Similarly, games often don't have a, you know, if you got a four-year release cycle, you can't be part of the today's zeitgeist and respond a game that takes advantage of that today because that release cycle is too long. When the release cycle moves from weeks to minutes, yeah, the kind of conversation I can have with my players, right, changes and the way I can like follow with trends and follow with a meme and come back and play back with them by releasing new content and new levels or new characters, that just changes my relationship. And I think companies that can keep up with that speed will be I think you're exactly right on this Aaron and and the way that I would characterize it is is it's

actually like when you think about the objective of a company or a team ultimately it should be about shareholder value. Now what is going to drive shareholder value the most in this new world of AI I would argue is product velocity to your point right taking that player feedback adjusting to it and so the the product velocity is both a function of the direction and the speed and with respect to this speed then it's these teams that are dynamic that understand the inputs to being able to speed up development and product velocity. So someone that understands the the factors of production that will go in and how to optimize that that understands how to organizationally structure a team to be as fast as possible and that understands the this notion of RT return to intelligence and knows how to apply AI to their team or to various capabilities. I think those are the teams that are going to win.

Uh oh, go ahead. I think that's absolutely right. But what kind of what strikes me when I'm when I'm listening to you say this is it's um it's a point that would actually be true even if the revolution hadn't happened in the last few years. Like this is an industry that is not set up to work efficiently has not kept up with like best practices in like software development anyway. So like there there is this uh even if we weren't grappling with AI and we are. You're telling me Ubisoft is dead. Is that what you're saying? For the record, I did not say that. Okay. Um, so may maybe also Erin, I I' I'd like to also just kind of maybe go into a little bit more nuance in terms of how you're embracing AI as as one of the few AI native game studios out there when it comes to where do you apply your resources in terms of the model in terms

of like Yeah, I'll just let you. Sure. Sure. So we have a hybrid process. So we work in Unreal. Um, we that's primary engine we work with, but we also have our own AI backend. Uh, and then one thing that was really important is working in the engine. If that was one trick, if one bit of advice for anyone, we didn't sit around and just do a bunch of text prompts and be like, "Hey, that was really awesome. We got that. Let's figure out how to put it." All of we actually built custom real time like while the game is running. I can do prompt engineering testing and change that in real time. We built those tests in engine so we could always prove that the experience is in the game and I can see it um to your so that was one important lesson learned um to answer your AI native question and and even the DM question. I'm going to actually give a quick example and it would also maybe cover costs. A year ago at GDC we had a tech demo uh that uh features about well many more things than retail mage does. 32 mage is about a third of the AF features we prototype. Um, and it was

very very rough. It's a tech demo, right? Um, but it was so the premise is uh you're in a village, there's villagers disappearing. It has our AI technology. Uh, we also have a procedural generation AI technology that we use. And so there you're in a village and uh villagers are disappearing and the question is why? And so we had one set of players, they were actually had an improv background, so very creative. They talk to the villagers and they discover that uh it might be werewolves. So they're like, "All right, okay. There's werewolves scaring uh kidnapping people. What do we do?" They decided they were going to bake cookies for the werewolves to pacify them. So they talked to the villagers. The villagers are like, "All right, sure. We're going to roll with this." So they say, "What do we need to do?" Well, we need to gather ingredients. So we spawn a new level. We have a procedural generation process where we combine classical procedure generation with machine learning to build a new level on the fly that follows that narrative context. creates an environment where they go off, they collect the ingredients, put them in their put them in their inventory, they talk to the villagers, "All right, what are we gonna do?" It's like, "Well, there's an Elvish bakery nearby. Let's go bake the cookies." So,

spawns a level with an Elvish bakery, an oven. They go in, they put the ingredients in the oven, out pops some cookies that have in-game effects attached to them. They're like, "All right, let's go find the werewolves." They go find the werewolves, they throw the cookies at werewolves. Werewolves were not impressed. But all of that was emergent gameplay, right? Right. I never programmed cookies. I never programmed an Elvish bakery. We have an ass a game asset library that the system can pull from to to build these things on real time. So I don't know if from the machine uh from the dungeon master side whether we'll be an embodied DM there. I think we look at it more as like the AI director that we saw from like Left 4 Dead and that kind of tradition. It's the game is playing back with you. It's a bit more invisible, but I think we're very close to being able to do this. If I add one more part to it from a cost side though, when we were first prototyping this and using some of the now the API costs have dropped, but using API costs from like OpenAI and uh claw uh and some u some of the other uh

cloud vendors, a four player session on that would run the cost API cost of a ticket to Disneyland. Like you can't do that. Um, part of the work that we did last year is to drop that from hundreds of dollars to dimes. Um, but that required us getting our hands much get our hands dirty, a lot more than I expected and building our own AI stack, building our own using our own uh well foundational models that we that we uh fine-tune uh a lot of collaboration with Amazon uh in order to reduce those costs. And I think that's another area where developers understandably will look at that and and and extrapolate and be like, I I can't there's no way we can build a game if it's going to cost $200 every time a player plays. I think we as a startup would we're just going to move forward. Um, but that is a limitation and and I don't think the game industry needs to be on the forefront of of this

partnering with some of these other AI companies because uh that's the only way we'll bring down these costs and make the and make the operations and such that it that other studios will be able to follow behind and build the kind of experiences we want to see. Yeah. And to your point though about where you have been applying your resources because if you're working on your own model and things like that, you know, one of the interesting things that I've discovered is just so much is happening on a daily basis. And so to this issue of how dynamic the AI space is because almost daily something new happens like oh you know, bite dance released this open source model. Oh, Sesame just open source and this and that. So then how do you think about where do you where do you apply your resources and where do you stop like when you when you think about okay we're going to use this when something new is popping up every day. Yeah. Yeah. I take some solace in that that that is not necessarily a new problem in technology. you know, we we always have I I have often been at the cutting edge of new technologies and you have to kind of decide uh where those

are, but we specifically so we use openweight models um and uh we specifically actually have built a set of benchmarks and unit tests that we have so that we can onboard new models and test them to see are how do they perform uh because we're anticipating I think the important part is to anticipate change and and build a platform form a technology stack so that you're able to onboard new technology. Um it's not easy. We'll get it wrong a couple of times. I certainly have over my career. Um but we shouldn't be surprised that next week a new model will come out, right? So it does require a little bit of forefront for like a little bit of planning uh in that regard. Okay. Yeah. In some ways this kind of takes us full circle back to the start of this conversation. Uh so the the thing I said there was even if there was no further progress we would at the foundational level there would still be a lot more work to do. That's definitely not the same thing as saying it doesn't

matter whether there's more progress at the foundational level. You know every single time that uh the the models move forward that the hardware moves forward that's enabling use cases that don't you know you might be able to dream them up but they're just uneconomic right now. uh but every step forward unlocks a whole set of new things. So there's already a thousand things we can do uh but each you know each additional uh piece of news that we hear whether it's deepseek or whatever coming through that's unlocking a new set of things that you can do in reality every single time. Yeah. So the other question I think is important to ask for me because I have a lot of friends in the games industry and some of them are like all the stuff you talk about AI is pie in the sky. Come on, show me the impact it has right now. So maybe the question I could ask you guys is in terms of what you're able to do now versus or maybe even more short-term near-term. Um so

for example I'm very excited about um MCP and Agentic and the potential ability for leveraging both of these those things to automate workflows for example. So, you know, a specific example would be like, could we automate a battle pass season completely, for example, and with MCP, you know, one one of the things that's been exciting that we can do right now is use like Windsurf and MCP and Claude to like have it audit an entire code base. Um, have it evaluate specific developers on how they code, find risks in the code and things like that. And so from a right now or near-term perspective, are there specific breakthroughs in AI or capabilities in AI that you guys think the audience should be aware of? I said the the areas where kind of we're already there and all that's left to be figured out is implementation from my point of view. Obviously 2D asset generation um that's you know uh there's

stuff to be worked out there but it's all tooling and pipelines really. um live ops like you say um basically worked out kind of uh you know fully implementable uh right now dialogue voice sound there's a lot of stuff that kind of is already there and really we're just waiting to see the game that uh leverages this stuff to the max. Um so there's there's a huge amount that is you know we're we're just waiting to see it. It's um we're not waiting for a breakthrough in tech. We're just waiting for a breakthrough in delivery. Yeah, I mean there's a lot of things, you know, uh there are a lot of things that I think being so close to it I start to take very advantage take take for granted like you know our team uses co-pilots that does not necessarily mean that all of our code is generated but there are times we we are learning how to h how to as as pragmatic engineers who want to just get things done um you

know but have a high quality bar uh you know we're learning how to use that and and we use we use a bunch of AI tools across the board to improve you know summarization and meeting summariz kind of productivity tools. What surprised us was as we integrated in our a in our AI workflow our engine that we have as we integrated that uh so actually better way to say it I think that one day what we're kind of building to is a hollow deck- like experience an environment now put the AI or put the VR AR side of it uh which I also think would be amazing and cool but that kind of responsive world that responds back to you within whatever parameters it feels alive like we as games we've always been seeking that whether it's GTA or other kind of games right we're we want to move towards these kind of hollow deck like experience if you have technology that enables the holiday that means your tools have hollow deck magic too and

your tools will take advantage of it so I think one of the issue with some of the AI tooling right now it's too low level if all you give me is an LLM the the real hard work is the integration into the rest of the stack and the integration like you give me dialogue, that's great, but how does that affect my quest system? How does that affect my in-game economy? All of these kind of questions. I think the answer is actually going to be higher level. I suspect that someday what we'll be integrating with like the way you integrate with these a or work with these AI NPCs may might be more Westworldish, maybe with a little less murder, but maybe more Westworldish less murder. Uh yeah. Uh you know, where you're talking to them and and maybe using sliders. We see some of that magic now just by having English language as part of our essentially programming language which means that I've have team members who are writing unit tests in English that we use because we build out a system for that that they never would have written unit tests for because it's so much more accessible and folks who can just dive in and because they can change it's just so much more accessible

and we didn't set out to do that but it was a consequence of integrating this technology into our stack. So we have uh five 10 minutes left. So I wanted to also give the audience a chance to ask questions to our esteemed speakers here. Are there any questions from the audience? Um thank you for the panel and I have two questions. So one is you know about AI games right now. AI native games right now. I mean, you talked a little bit about, you know, what you're building, but what I'm talking about is like, you know, games that you can actually already play that are AI native and that would be impossible to do without AI, right? I can think of one example, and that example is suck up, you know, that viral vampire game where you where you know, you know, when you're a vampire, you know, you cannot just enter a house, right? The idea is that you have to be invited by a victim, right? And that that game like became viral like like a while ago. Super funny, hilarious, right? And it it it

went viral on YouTube, Twitch, Tik Tok and things like that. Are there any other examples like that? That's that's point one. And the second point is um I think Joseph, you you mentioned that at the beginning a little bit. Um the impact on um on uh like human resources, right, on people, right? That's why we are at GDC at the time of the recording and my understanding is that the floors are like full of job seekers, right? So, do you think that and this is something that I'm I'm getting asked also. So, what do you think is the impact on the industry in terms of you know staff sizes in term like in in in all of these studios? So, are we still going to see like 1,000 people working on AAA games for five years for instance? Yeah. So, uh, I'm happy to get you a Steam key for Retail Mage because I would argue that is one of the only other AI native games out. In fact, the Proxima folks are amazing. Uh, we took a page from, uh, Suckup. uh we we share some investors and uh one of the reasons we shipped that game was to prove out that there were AI native games that

were able to be possible today that are commercially available because I agree one of the issues is we see to be frank too many like just demos uh or they're only available on YouTube or they're at a conference and uh we wanted to give something that uh players could experience right now. Um, I'm looking forward to hopefully seeing many more. Um, but there are not that many right now. And uh, and I I I'd be interested to see your take of whether you think it passes the muster uh, for for an AI native game, but there are definitely game mechanics in there that we could not have done without AI. Um, on the other side, uh, and I'm, you know, curious on both of your your response on these, I would say the game industry is already facing issues with it's a great time to be a player. It is can be a difficult time to be a developer because we have so much content available. Um, AI will probably not make that problem any easier. I think it's up to a lot of us to

understand you know how are we focusing on again like doesn't matter whether it's technology what it's doing is this making better and new experiences for players. Uh if we think we focus on that aspect it will have an the right kind of impact that we want a positive impact to the to the industry. Yeah, maybe I could also jump in here because I as an AI maximalist, I believe there will be a pretty significant impact to teams, people, and even organizationally. And one of the things that we've seen with AI is that it basically is attacking work from the bottom and the top, right? Like so the most mundane repetitive tasks it should automate away and that's a good thing. But it's also taking very complex problems that are very difficult for humans to wrap their heads around, right? and being able to manage that complexity much better than a human can. So when you think about the the way that I think about it is that AI will enable a certain set of capabilities and it will take over a certain set of tasks.

Now what does that have the impact of doing? Well, it will make certain roles um compress meaning the the the things that they do, the work that they do become smaller. So what does that mean? That means that I think roles will compress. So you may not have an engineer and a product manager anymore. You might have a product engineer. You may not have a marketer anymore. It may be a product marketer. And so this goes back to this notion that I talked about around an organizational meta. that specific uh design of the team and an organization will necessarily shift depending on limitations, capabilities, weights. Very similar to how you know when Riot changes the weights on champions in League of Legends, there's a new meta. There's going to be an organizational meta, a team meta for specific kinds of products and things of that nature. And I so I do think that

that's going to happen. And I do think, and by the way, I want to apologize to Ubisoft. I I love their products. I I crit I criticize them. I criticize them in love because I want them to improve. But I do think that um a lot of people who are not able to adapt to this, these big organizations that have fixed structures that cannot have um the ability to adjust are going to be in trouble. And I also think that there's going to be a certain role that is kind of like a organizational orchestrator that well kind of an AI and organizational orchestrator that's able to understand AI, understand RTI, the return to intelligence and apply that to the the team and organization and design the the most optimal meta. Those are the teams that are going to win. And I think that's what's that that's how it's going to impact organizations. Go ahead. Yeah, I I think those are all great points. I

mean, the only thing I would add to this is I think it's just it's a misconception to think that it follows from that that there are going to be way fewer game developers in the future. Like that's just lump of labor fallacy. Um so there's going to be disruption to roles. Certain roles, certain specialisms that exist now might not be important in the future. Uh but I don't think there is any reason to expect that the industry as a whole is going to be a smaller industry than it is in the future. It's just going to have a different shape than it does now. Oh, one one other uh maybe organizational impact that I've been thinking about as well is one of the things that I noticed when I um years ago when I was speaking to Roblox developers is that I noticed that they have a little bit of a different way of organizing around game projects, right? And it goes back to I don't know if you guys remember the the nature the the famous essay on the nature of the firm like why do we have companies and the thought there is that it's because you have all these external

transaction costs that exceed the you know so for those reasons you have organizations um and what I saw with roadblocks is that there were a lot of developers that would self-organize around projects rather than having companies and so I do think that potentially One of the other changes that could occur depending on how far AI goes. Who knows how far it goes? But if it goes far enough, then I do think also there may be an organizational impact where you might have people that operate similar to Roblox developers where instead of having big companies, you might actually just have bands of people that kind of combine together. Um, you know, I'm I'm not a big crypto guy, but may maybe they use crypto to organize the the rights of the project, but then they may self-organize and that may that may have an impact on organizations as well. All right, with that, I want to be respectful of time. Gabby, we are right at 12:00. My god, what a what a moderator with the

time. That is very impressive. Thank you so much JK, Joseph, Aaron, and Liam for uh for sharing their wisdom. Uh, I think it's there's a lot of stuff to follow up on. So, [Music] [Music]