291. AI in Games: Innovation or Ethical Time Bomb? With George Ng, CTO of GGWP

[Music] Georging, welcome back to the podcast. Yeah, thanks for having me again. And George, you're the co-founder and CTO of GGWP. You've been before on this podcast not long time ago talking about bots and how to survive the the bot Armageddon or not really, but uh that's how I remember that podcast. might have been totally different. Uh but today we're going to talk about the real Armageddon, which is of course AI. And let's start off right away with the most difficult question. And that is what are some of the biggest ethical concerns around AI powered personalization in games, especially in terms of monetization and player manipulation. Yeah, that that's a great question. I think that's the biggest one that AI told me to ask. that this was AI told me like George should ask me that.

Yeah. So, I mean games don't want to be involved in what uh players or the market deems as predatory practices. What that means exactly is kind of a moving target. It's unclear in some ways, but that's that's the biggest one, right? um to power AI. I think a a big question is just around privacy and if you're going to capture this data to be transparent on the reasons why you're collecting it, right? It's like, hey, I'm going to help you create a better gaming experience for yourself. And that includes targeted ads and, you know, bumping rewards for when I think you're going to turn. Is that okay? Is that not okay? Do I tell you about it that way? Am I required to? These are like pretty big open questions. Another things I think that are probably not even ethically or ethically gray but negative like what about variable rate drops?

It's one thing to be transparent in a table, right? And we've kind of heard of uh recently games being fine for this. Uh but can I change a rate drop? It's like, hey, you really need this power up right now. I'm going to pop this up to you and charge you more for it. It's like looking at data and looking at AI to build recommendation systems in the background could be abused this way or or used this way depending on how you view it, right? Um matchmaking, engagement optimization. Um we know gaming isn't gambling, right? But there's a lot of let's say borrowing in themes psychologically from things like you know loot boxes in the first place. Uh but even things like matchmaking. Is it okay to like optimize engagement? Like I remember playing, you know, the first few versions of Civilization back in the day. Uh just one more turn. And that that seems fineish, but what about matchmaking? Can I matchmake you every so often with, you know, really bad players to make sure you play another

match and then mix it up so that you never really progress? Yes. Um is that okay? Yeah. I mean, George, I feel like you're talking about things that we already do and we weren't that efficient before and now we have better tools and now we're like, "Oh, there's an ethical issue cuz I'm much more efficient at doing what I was doing before." Web stores, you've seen what good looks like, but have you seen great app charge is the direct to consumer platform backed by the likes of Supercell and built by monization pros from Moonactive, Platika, Huge Games, and others. People who know the difference. App charge powers over half a billion dollars in transaction a year for top studios. Optimizing every detail. Deeply personalized web store offers custommade design, industryleading monetization features, in-game payments SDK, and more. 70% of top grossing games already have a web store. But the one scaling fastest, they're doing it with

Appcharge. Learn more at appcharge.com. Yeah. Yeah. No, I I think that's right. I mean, one thing to understand is like when we talk about AI, we're really talking like we we conflate AI itself with the automation part, right? So like in any space, we use AI to now have more data to have more information to make a decision and then we automate that decision. So the automation or like making that decision in the first place is the issue. And a lot of things that I just gave examples are are things that we already made decisions around with or without the AI. AI just can make it worse, right? Like hey, I could have previously also just changed a rate table and the loot drop, but now I might know more about your specific persona and where you are in your customer journey. And I'm tracking this live and I'm tracking it with your friends and doing this stuff. So does that make it worse? Right. So I I think that's one

part of the question. No. And then I I guess if I took a step back and just talked about AI and ethics in general, there's like a whole body around where we are today. And feel feel free to cut me off if I go too deeply into this, but really it's like I if you think about like trainings, like LM are super popular now. uh but you think about western trainings for LM just because there's way more data available and that data was used to train like the first few generations of these foundation models. So that means inherently that hey uh if you run you know like our company we do voice and text moderation and we're basing it on these models that are just way better in English and way better in European languages also has leanings and biases towards those trainings because that's what the training was based on. Um, so that's a bias, but that's not in itself an ethical concern. But when I look at that and I apply it to like moderating voice chat in something like I don't know,

Tai, where the performance is less good, whether it's us or Google or Amazon, we're not actually um it's not obviously good or bad for TAI players. They'll just be flagged less. So they'll have fewer false positives, but they'll have fewer false negatives, right? So it's just different. And that's like traditionally like in pure AI space how you talk about bias, right? That it's less good in in this way. Now, if it's systematically biased in this unfair way and Thai players get, you know, Tai speaking players get flagged a lot more, then you have like a systematic bias and that's like an ethical concern. Um, but these models we know don't perform the same, right? We just don't have cohorts the same way. So, it's like, um, you know, female players, their voices sound different in a competitive shooting game, so you just have less samples of them. younger players in more mature games, same thing. So these models again don't work equally well in in every language, every tone, every demographic. So then you can get other biases from that. Sorry, slight tangent there. A

different form of an ethical concern. Got it. So because you will have um smaller group that is not as well represented in the ecosystem then they will won't be treated the same way as the larger group because the LLM is just not proficient or doesn't understand them as well. That's right. So your performance on estimating what they're saying for content for moderation purposes for all these purposes will just be less good be less accurate. Right? And depending on how you use that, like if it's in the variable rate drops, it might actually be a good thing because they can't figure out what to do with them. Right? So then they're not getting the powerups, you know, at these like predatory pricing as bad because they get grouped into some other group because they're a unique example, right? They're essentially outliers in that population. I understand. But wouldn't the LM learn over a relatively short period of time

to treat the smaller subset of users the same way as it treats everybody else? Uh I mean it depends on what you mean by small subset. Like if you're a game like Valerant and you like have a you know you move your server from like Singapore to Bangkok and you have like millions of players playing. Yeah, absolutely. You just get like large sample size very quickly. Um that's not true with each of these cases though. M so it kind of just depends on sample size right I mean the the reason why uh these you know like model trainers use so much compute is you're talking about vast amounts of data so if you're using somebody that's like a really unique or different demographic it's just going to take a lot longer to train and what are the concrete ways of treating different groups in a in a different way like let's say we are talking about Valerant let's use that as an example Um, let's say you have some kind of a

player base that is not well represented. Who could that be? Um, I don't know. I don't really play valor. It's too difficult for me. Um, you tell me. Like some kind of a player group that is poorly represented. How would you experience the way that AI is treating them differently? This minority group versus the majority group. Yeah. So, like I don't know, let's say you took a young female player that speaks Thai, right? Just to keep create a really small sample, it just means if you're using like voice moderation, it's less likely to catch what they're saying. So, it could work in their favor, right? Like in that stuff that they say doesn't get detected at all or could work um against them, right? Because the model's like less accurate. It like flags them all the time. One is like actually a benefit to them. One is a a negative to them. But overall, it it isn't like clear if it's like better or worse for them. It's just

different is more the point. Got it. So, if there's uh I don't know Thai at all learn the language, but if they would have a C word, that would mean totally different that would be it could very well then then she would get banned. Yeah. Absolutely. So, it's possible, right, that if they use that language or they they use terms in that way. Mhm. Um, now I I I think some of that's resolved, right, because they have like support teams, hopefully local support teams in those regions to tell you what those specific terms mean in that area. Mhm. And they're reviewing stuff. And I think it's just more important to have like um you're always like assisting automation, right? So that's kind of a balance between like what you automate and what you don't. Um, but it's e even more important, I think, for those types of edge cases to make sure that you're properly like reviewing people. Interesting. Now, now I now that I went into this cursing thing, I was thinking about the C word is actually really bad

in the US, but it's not as bad if you're in Australia or in UK. It's the same as that's just English, right? So, so language is just really complicated for sure. Yeah. Huh. Okay. Well, let let's jump into something that that most of the people are very well aware of and that is uh the AI's impact on game development all the way from asset creation to narrative design. How do you see this? Yeah. So, I think um it's universally important in early prototyping. So like uh you think about genai, you're accelerating the creation of uh concept art or level design and asset creation. So things like textures, landscapes um or or even characters designs just to let people storyboard and quickly visualize ideas. I think even in some cases animation, right? So you can automate like motion capture cleanup or uh create fluid character dynamics.

Um, now I think it's important to note, um, it's it's probably a little bit different between like a a smaller studio versus a larger studio because a larger one is just going to have like far more IP concerns. So that uh prototyping falls off earlier in the process whereas like a smaller studio might be more aggressive and use more of these things until later. Um but I I think the key differentiation there is just you have accessibility to this highquality lowcost initial asset creation without large teams is that like what's the future? Um there's meaning when I talk to executives they're all seeing that in the future the team sizes will be smaller because essentially the midlevel people will become obsolete which is kind of crazy because

I mean I understand that you'll keep the seniors and you might get some of the juniors doing really simple but what do you mean there's no midle like how do you go from a junior to becoming a senior so this is I think something that if I'm not wrong. This is something that even Zuckerberg were saying that there will be less mid-level engineers in the future in the company. Um I don't know that that that's a interesting take. I too have heard that. I haven't had a ton of time to deeply unpack that. Um, I think in games I haven't seen that as much yet, but certainly, you know, I guess taking a step back looking at like the macro funding environment and the cost of making games in general, something has to give on the AAA side. And does it come and does it come at mid-level employees specifically or is it more outsourced moving to codev entirely which would also, you know, impact the

low level? I don't know. Um I think the answer is going to be different possibly in different industries as well. Um but there's definitely going to be workplace disruptions, right? Right. And it's just more I mean going back to the things need manual support comment even about the uh looking at you know bias in languages kind of depends on how important the thing is right like meaning you probably want a human in the loop for like late stage no matter what the task is as long as it's important. Uh let let me give you an example, right? Um if I'm going to like um like I I I used to work in defense, so like I would generate reports on like where we thought uh interesting targets were. That sounds like offense to be clear, right? Well, I I can't speak too deeply about it, but yes, in in in that case, absolutely. Um, but

there's no way like you know like I would work with people and they'd be like, "Oh, I can't believe that they would drop, you know, drone strike." It's like they're not. It's going to like go through like eight levels of analysts and that's probably right because like that problem should be automation assisted, but that can't be the final call, right? Like in a game, if I like don't let you talk for 5 seconds, the pain of a false positive is you couldn't talk for 5 seconds. even if I get it wrong, do I owe it to the player to check with them or is it explainable for the greater good of the community that like I didn't let you talk for five seconds and I get it wrong 5% of the time or something like that? That's probably okay, right? But is it okay if I drop a drone strike? Obviously not, right? So like there's different bars. I mean, I'm using like, you know, some of the most extreme Yeah. between like between like like shutting somebody's mouth for five seconds to drone striking them. I think

there's there's the there's a there's a quite an escalation ladder I would say in between. Yeah. But but I mean the main point is that's kind of what gets lost in in the translation, right? Like we talk about all automation like we talk about uh AI and fairness in different governments and a lot of it is less about the AI model and about the application itself like where are you going to use this like should you automate whether someone can get their driver's license like probably not should you automate like where whether someone can get access to credit so that they can buy a house like also probably need someone to review it too right so like there's certain things where like you you need different rules around that stuff because the impact is much greater and I think that's kind of where you're you're finding that balance and even in uh going back to game design. Um you know games are super fun. I I I love games but you know have to put the false positive rate of like

games or like the false positive impact of games in like the proper perspective and context. Like if you generated this thing like if I'm the biggest risk is like you know IP theft, right? being sued for that probably or just making not a fun game. So yeah, I think even you know that you want like AI in the or sorry a human in the loop at the end to review that like the through lines make sense in the story. They're also being generated. So that makes sense why you'd have like people there at the most senior levels cuz someone has to take ownership of it. You can't say, "Hey, it was this machine that made this game, right? That that doesn't make any sense." Uh, as far as the low level, that's kind of a surprising framing to me because I would assume that the lower level would be replaced first, right? Um, but yeah, I'm happy to take a look at it and we can iterate more in the future. That's interesting. Yeah. Yeah, the mid-level part was was was very interesting and also like the implication of it. And I feel like the lowlevel like the lowlevel is a little bit of a bad word. Let's say entry level. uh so the entry-level uh

folks getting less opportunities to enter the workforce just because of a lot of the work that they would have been doing has been automated that is quite of a significant effect uh and it feels like in game development es not in game development especially game development is just very close to I I think it's the only thing I understand so so talking about the only thing that I understand I I really see what are the areas is where AI impact is more significant compared to where it's perhaps less significant. And and I feel like the area that that where the impact is perhaps the most pronounced is narrative design. Am I wrong? Cuz like Yeah. Um I think that's true. Um I mean just thinking about the direction of games in general like there's a lot of multiplayer games there that you know like I like Marvel Rivals. I wouldn't say the narrative design is like the

most important part. Uh but there's a ton of these story-driven games. Ask Ask the narrative designer. They're like that. Anyway, our focus test showed that. Shut up. That's totally fair. Um I I do think it helps easily, right? And I know that games are already using things like Copilot, using like GPT, LMS to help with that narrative design. Um, so you can spin up a couple things to say like, "Hey, give me a backstory for this person." Yes. And then this is the thing. And even if they don't use that exact thing, I'd say a lot of people are using like 95% of it and then editing it. And obviously this goes beyond games. Um, yeah. So it definitely helps and it helps flush out the thing very quickly. Um, I don't know like even without LLMs, right? Like you think about companions in recent games like um I don't know like Ellie from Last of Us is pretty well written, right? Atreaus from God of War. There's like these companions already so you can go deeper.

I think um where people are trying to go is to more personalize these NPCs. So like now I'm coming in, I'm talking to them and now they know that I'm I'm George as soon as I play the game and they remember the conversation I had with them last time or they remember the conversation or the interaction I had with them. So that can make the world feel a little bit more real. So I I think that could add to the immersion. Um I think it depends on the type of game. Yeah. Like it's not Diablo where you have three options of of where the conversation could be going, right? So it is possible like in a Diablo 5 where you could have these like branches and memory and history around it. Yeah. Yeah. and and actually, you know, find out through a conversation. Like you can really start a narrative like that is the true narrative of of the game where you can actually dig deeper and have a conversation with a character and find out who they are, where they came from. Maybe like some surprising things maybe like viral sharing moment based on what

your conversation undercover like uh what you find in the conversations compared to those sort of a you know very strict and uh and detailed answers that they can only have. Yeah, that's right. And I think um if you move a little bit outside of games to like what I'd call games adjacent, you already see stuff like that where like whether it's like I don't know character AI depending on whether you view it as an app or like a platform to use like a replica which is like is it like a therapy slash I mean some people use it for like a dating sim. Um so lots of different things that that you have there already. Do you think it will ever be fully autonomous? Like the characters will actually evolve and feel real? I mean, once the memory issue is solved, of course, cuz I think there have been these situations, but they run out of memory and you kind of have to retach them who you are. Yeah. Um,

I think it'll be much greater depth than today. I think it depends on the level of depth that you're looking for, right? Like I I still have, you know, friends that like speak so fondly of like, you know, the old Final Fantasy games. And to them, those characters are real, like no different than like any other entertainment experience from that standpoint. So, will they feel more real? Like, yes. Um, I know people were like talking about committing suicide when some of the uh the replica models were changed, meaning they turned off some of the functionality cuz some of them were using them effectively as like I don't know how PC we have to be here, but they were using them as like uh I mentioned dating sim. I mean, you can imagine some of the things that they might be typing or saying to their Yeah. like basically and then the Italian government came in and said, "No, you have to turn it off." Which government? The Italian government. Italian. Yes. I mean, the EU uh there's there's things

like uh DSA, GDPR that like are uniformish across the EU, but then each country has its own local privacy safety laws too, right? Italians, the Italians came through. They're like, "Listen, hey, this is too much. No more sex dolls. That's right. But yeah, I mean that like like people were super attached to their characters and now they're like, "Hey, my AI is acting cold to me cuz they like, you know, they added some co code to prevent them from like behaving that way anymore." So it's like, oh, and then from that detachment alone, you know, felt real enough for like them to like talk about how they have suicidal thoughts, how they're seeking mental health, right? because of this change. Um, so I don't think it's that far away and to people it does feel real um or real enough, right, that it causes that mental pain.

That's so sad. Not that it like I think it's sad from all the angles where somebody would have genuine real emotions towards a software that is literally coded to create those emotions. You know what the you know what the hoax is and you be like can't can't resist, right? But but that last part I think is what makes it tough, right? Because it's coded to create that kind of response. an engagement, right? And I mean, I don't know, like Black Mirror has a new season coming out depending on when this airs. Um, and there's like a whole episode on Black Mirror from it. So, it's not that far-fetched already, I would say. H Yeah. when when we were talking about these NPC platforms, AI NPC platforms, uh I think there I forgot the name of the company was one really big one, the

Lightseed joint. Um do you feel that they will be at some point like fully autonomous like evolving characters that feel as real human beings or real human players? Attention app and game developers. Are you ready to unlock the secrets behind the world's top performing apps and games? With Sensor Tower, you get the ultimate competitive edge, whether it's breaking into new markets, fine-tuning your UA strategies or boosting monetization. Dive into data like never before. Track global trends across app stores. Analyze market leading inapp purchase and inapp ad strategies. and benchmark your performance with realtime insight. The best part, Sensor Tower doesn't just deliver data, it helps you turn it into action with intuitive tools and expert analysis. It's everything you need to make smarter, faster decisions.

Thousands of app and game studios, from indie creators to global giants, trust Sensor Tower to help them win. So why not you? Start your free trial today. at sensortowwer.com. That is sensortowwer.com. Yeah. I mean, given the example I gave with replica, I think they will feel that way to people, but if you deliberately scrutinize it, like from a more mechanical angle, you'd be able to tell that it's not in my opinion. Yeah. Um, ultimately I think look AI is like already there. I'd say if you said, "Hey, generate a story that looks a lot like Marvel, right? And it can draw from like rich history." So, it's really good at that. But what AI lacks is like the intent and understanding of nuance and in emotion. Mhm. So if you're looking for that

experience like you know some of these people were using like a replica then yeah it makes sense that you get drawn into it because you're not deliberately looking for the flaws in it. But if you take a step back and you're deliberately looking at the flaws in it then I could see um it'll be some time before you have that separation. Yeah. I mean on the other hand there are people built massive webcam businesses. So what do I know? It's it's the same. It feels like same thing. It's like really you sent $20,000 to this lady that is that needs to, you know, that will be coming to see you any day soon. Okay. Yeah. And I mean, arguably that's not any more or less real to you from a practical perspective. Exactly. It just uses human human humans to essentially create those emotions. Um, how do you see AI driven procedural generation evolving? like could it at some point totally replace some of the functions or

do you still believe in game development? I mean, but or do you still believe that there has to be that sort of a human backs stop on on every every aspect? Yeah, I think um I think can and should are probably two different I'd give two different answers for can. I I already see like the first generation of games trying to like basically have the AI be like the co-director or if not even the co-director, the director of the game. I haven't seen any of those be successful yet, but I think because of the cost of making games, you're going to see some, you know, smaller studios, some people bootstrapping stuff trying this out. And I can imagine some of them even having like some indie level moderate success uh in in in the near term. Should they though? Probably not, right? Like I would think if you're the game director or game designer, you want to take ownership of your creative vision. And as mentioned like AI does have gaps, right? Um ultimately it's not creating

new stuff. It's regurgitating old stuff which a lot of uh creation is anyways to be frank. Um, and and that's kind of in the balance, right? If if you're talking about like something like totally novel and making sure it all fits, again, I go back to the point about, you know, AI doesn't really understand the intent or the background. It's just an amazing summarizer and it can reconfigure stuff, too, but it might not make perfect sense together. So, it's going to miss little nuances in like timing. With that said, I think it can do a pretty good job. like right now like if you wanted to do derivative work like it's like let's say me and you want to I don't know reboot you know the hundth version of Tetris it's like we could do that or like literally we could have any number of ways that to to piggy back off AI to like help us do like you know 95% of that work right and we're just adding a layer of polish to help differentiate it. Yeah, I'm I'm actually we started doing a little bit of like little app

just on the side just to see how fast it is to build it with AI and um and basically just concept test it in a in a in an incredibly short amount of time and just and when when designing an app of course app is not the same as a game. So it's much more um how would I say controlled environment like you're you're you're having the the user go through specific paths and the monetization is subscription based but the the speed of development when you know what you're trying what kind of a service you're trying to do is pretty intense and now we're going to go in like we did the design in just a couple of days on everything and now we're going to go into coding and I can't wait to see what happens once we begin coding an app through uh with uh with with AI. So uh and then we'll do all the marketing assets with AI. So uh interesting interesting stuff. Yeah, for sure. And who knows? We'll we'll see. And maybe that you know 90 95%

approximation is like good enough. That's insane. That's that it's uh yeah, when you know what you want to do, the AI is kind of just helping you to make it faster. It's like I want this, I want that. Can you tell me 50 of these? Cool, cool, cool. And then just start putting it all together, right? And you can even tell it like again with data sets, right? Like with sample size, you can say, I want to make the game like a Blizzard title, like from this era in this way, and it'll take that in account as well. Three Vikings, let's go. [Laughter] Um, what's the future look like with AI? I mean, it seems like the entry barrier to do anything is just lowering down. And um as as a context, like I'm I'm thinking about the um GameCraft podcast and when they were talking about AI, they sort of painted two different scenarios. A bull bull and a bear case.

And then the the bull case was um well, the let's start with the bear case. The bare case was that there will be so many of these AI startups in gaming specifically that will just flood the market with subpar products essentially variations of each other and whenever anybody gets traction somebody will just put in another another 15 titles of that same thing in a matter of days if not hours. And then there was a bull case uh where AI is just going to make everything much more efficient. So larger companies will be able to create immersive worlds, just immersive universes pretty much and then smaller companies, smaller indies will be able to build much larger games, something they weren't able to do before because of the uh you know the resource restrictions and now they will be essentially uh unhinged to use their creativity and get that into a software. and they leaned towards that bare case

where we get the flood of subpar products that only help the incumbents to become more successful because they stand out with the IPs that they owe and all these shitty copies are only making the original product better. Like you you can have Marvel and you can have hundred other comics that are copy of Marvel and all you're going to be is like, "Oh my god, I really want to read Marvel because this is kind of shitty." like it's it's 80% of where Marvel is. Why don't I just read Marvel? Yeah, that that that's a interesting take. I was going to say as you describe the bull and the bear case, my views would I don't see those as different states. I can see both happening at the same time. Like that bull case is already happening in the background. We as just consumers don't really see that as much, right? We are making like designers more efficient like early stage they're already using even at the large studios they're using it for like the early textures and stuff that we described earlier. Yeah. Um I

haven't seen the bare case yet just because I don't see every game. So I I tend to see more popular games. I'm only more aware of it as like an angel investor and as like a a tech startup. We come across games like this, right? So I see like a handful of games that have received funding that are trying to do it totally based on AI. Um so it's harder to see this copycat world today from me. Um but in success in entertainment in general and in gaming you you see copycats with success. Yeah. So if one of these hits the consideration too is well they didn't have that moat in the same way. If you made it quickly with no moat Yeah. then obviously someone's going to fast follow, right? And we saw this like on every platform you think about I don't know Facebook games early generation or first generation of like mobile games and the number of copycats of like the same types of title like reskin. Yeah. You'll probably see that here too. Yeah.

So I can see both but I don't see them as being mutually exclusive. I just see them as like kind of targeting different markets in ways. I I don't remember the copycat approach working in gaming. Like I was there in Facebook games. I remember our company like where I worked at Digital Chocolate, we did ship reskins and the reskins performed worse and make made actually the original game perform worse. Uh I remember American companies like Kabam, they were all in the resins business and collapse their business essentially with just making reskins of their same fours back to back to back with with different type of IPs. And this type of approach where you where you just make it faster seems like a quick spiral to the uh to the bottom. Yeah. And I think that's true in a sufficient time horizon, right? If you were like at Kabam at that time or uh like um a good friend of mine like started Plato and he started GGWP with

me. Um if you catch it at the right time, you can do very well as a company. But you're right, what you're ultimately doing to the market and to the overall or to the player perception of the market is not a good thing. I remember our our head of studio back at Digital Chocolate. He's a you know since then he's done phenomenally well, super smart guy, super nice guy, but this was long time ago. He was head of studio then just needing to give a caveat like he's not a dumb person by any means necessary. I remember like do did a presentation digital chocolate and he showed like we had a top title. I think it was like Millionaire City was called that. And then he was like, "Look, here's our strategy. This is the game. We're making good bucks. Like we're going doing great. Next year we launch two copies of this and look look how much our revenue goes up and the year after that we're going to launch two more. And look at our revenue just going." And you're like, "God damn it." Like that's the strategy. Let's let's go. Now I was I

was young then and I looked at it. I'm like, "Huh? Huh. Okay. I guess it works like that. Didn't did not work like that. Yeah. I I I think a consideration with market dynamics is one one around preferential attachment. I'll get to that. And two, just we're in an era where games have like the longest legs that they've ever had in history. Like meaning you you think about like the top titles, right? Whether it's like Fortnite, you're talking about like seven, eight years old now, or like League, Counter Strike, you have these like games with these massive legs that people still play. And that means it only leaves like a very small percentage of the market for new IP each time. So independent of any of this new AI creation or AI boosted creation, you already have these types of dynamics. Um, and I mentioned preferential attachment. Like a lot of these games have these long legs because other players are playing it. So you have

locking in a few different ways from your social networks, right? It's like, hey, my friends play this, so I'm going to continue to play this or I just have a ton of assets in this game, so I want to keep playing this game because there's something of a sunk cost, right? Like if you look at their, you know, like if you look at like leagu's uh retention rates, right? If someone like bought a skin, it's like disproportionately high, right? So you see a lot of these types of trends um that does make it hard for anything to break through. So then for the AI boosted examples like if you get a good game and good is like tough right? Um that's probably like the biggest mo if you just have something that a lot of people are playing there's like critical mass and it's not so easy to like move a person. It's more about like moving an entire network. It's not as extreme as like social networks as we have right because that's like pure social but gaming is like highly social already. So it's like even if I don't have friends, it's like who do I flex the skin to if no one plays this game and I would just migrate off this game. So that's a

consideration I think in these market dynamics too. Last one. What do you think about these fully generated games? I think Microsoft had an example of it. I know there there came in I know a few startups that are working on it where they essentially can generate games with AI and they do these demos for me like look we're generating games and my answer has been like nobody cares about new games especially new games that are not that good that are copy. So I'm I'm trying to understand like who needs this. Yeah. I I think from that standpoint like the the bar for a player is I just have so many things competing for my attention. Yeah. Right. At the end of the day it's like whatever the best game is, it doesn't matter if it's AI generated or not. Uh same with shows, right? Like if there's a really popular show, maybe I put down my controller or my keyboard and I stop playing this game for some time to like watch the show. Um that's a pretty high bar. So if it's

made this way, it has to be good. to your point, it's not just about, you know, being cheap. Um, I think AI can do it, but again, with some of the the gaps that I mentioned, if it's a story, I think there's going to have some gaps. Um, I think people talk about it a lot because it's really costly to create new content, right? If you think about Yeah. live ops, like multiple releases, it's like, can you at least assist it with procedural generation? Can you make like, you know, we we mentioned Diablo earlier, so like um you already have like procedural generation without AI. Can it be better with AI? Can you make like more worlds where people would like want to play different variations on these dungeons differently? Um if you can, it saves them a ton ton of money because it costs a lot, right, every time they do like a new patch or a new update. Um and create content for people to keep going through. That would be valuable once you already have a title, I think. Yeah. Um I think it's much harder to go the other way when you don't have like that player base to get

them on board that first time. Um so I I do think it's possible to make a fully, you know, AI generated game today. Um I don't think it'll be as good. Um mentioned, you know, like the extreme example of like a Tetris skin. Like yeah, I think it's like super easy to make something that already exists because AI is great at copying identically. M uh but if you want like some spark and you believe that that's going to differentiate you I think that's harder to come by probably if it's about expanding an existing thing and keeping it running I think yeah probably you you could probably le leverage a strong existing IP brand and existing platform to extend it. Yeah, I was thinking a lot about puzzle games in this example or slots games. Uh both of them have when it comes to level design or slot machine design what we consider on the uh on the outside like how hard can it be? And in reality in both of them there's like 20 people in the world that are really good slot machine designers or

puzzle level designers. Like literally 20 people in the world. That's that's where we are at. And if somebody says, "Yeah, I can do a match three game, no problem. AI will just spit it out and be like, "Yes, yes." It's like it's like making bad ice cream when there's good ice cream available. Like nobody's going to eat it. Yeah, that's right. Great. Well, George, thank you so much for joining the podcast. Um, and as a parting shot, like where do you think we'll be we'll be at with AI in 2 years in game development? Yeah, in two years I actually don't think it'll be that different. I think we need to answer some of these uh IP ownership type questions to make it a little bit more pervasive and clear because that's when you'll have like the larger studio buy in. Otherwise, it'll kind of be at the fringe, right? It'll be a part of the indie story or just being components like yes deeper NPCs

but not in a meaningful way my guess small small tool sets that will add 15 to 30% efficiency to studio operations but nothing that is world changing. Yeah, that would be my guess at least for the near term. I 100% agree with that. And on that agreement, ladies and gentlemen, thank you so much for listening to this podcast, George's connect information below, LinkedIn profile, and um catch you next time. Thanks for listening to the full episode. Now, did you know that Deconstructor of Fun isn't just your favorite industry podcast? We're also a bit of a consulting powerhouse. Now whether you're launching your next hit, scaling your live ops, or betting your next acquisition, we've got you covered. From market research and go to market strategy to live operations tuning and due diligence, our team of experts works

with top names as well as the rising stars of the industry. So if you are building, publishing, or buying, don't do it alone. Do it with us. Email consulting deconstructorofund.com and let's build smarter together. [Music] Georging.