Panel: Beyond The Algorithm - Will AI Redefine Player Experiences?

[Music] So, I'm going to let the uh the the panel um moderator come up. That's Dave Bradley who is COO of Steel Media, but also more importantly for this is the editor of AI Game Changers. So Dave, come on up and I'll let you introduce your panel.
Thank you, Bradley. I'm the COO of Steel Media here, the creators of Pocky Gamer. Uh and I also curate the AI game changers newsletter that we do um every week. Um uh in which I get to interview a lot of people who are working with AI at the moment. And this is a perfect opportunity then for me to do that live on stage with two uh very important guests and we're going to run a panel

called beyond the algorithm. So before I get started I'll invite to the stage we have from ABA and Otto from Supercell. So come on up and we'll we'll get started. Please give them a round of applause. Go to here. Take a seat. You can be in the middle. Great. Um, so we're talking beyond the algorithm, um, about how AI might define, redefine player experiences. Um, we've spoken a lot already today on stage about how AI, uh, can be used in the workplace, how it can be used in the workflow. I don't know whether you saw Tommy Thompson's talk this morning. He was, it was a great overview of where things are, including sort of the difference between where AI is already being used to speed up processes and where it might be used in kind of, you know, AI native games for experiences. So we're going to talk a little bit about that about what I AI can do for us and how that can be leveraged by studios and how the technology can uh can really change the way we play games as much as the way make we make games. So first of all then I'll invite my esteemed guests to introduce themselves, tell us a little bit about their business. So uh

here go we'll start with you. Do you want to Yes. Tell us tell us who you are and what you do.
Yeah. Okay. Hi everyone and thank you so much for uh uh being here. It's uh and it's super to see so many people as well. Uh I'm Hega. I'm the CEO and co-founder of IBA. And what we do is we have uh developed an AI powered moderation tool. Uh so we can help you to keep your community safe and nice from all the toxicity out there. And I'm very interested in humans. That's why I'm working with AI.
Yeah,
that's great. Uh very interested in humans. That's excellent. Otto.
Cool. So, hi everybody. I'm Otto and I'm an AI lead at Supercell. Um, I'm an old nerd. I've been using computers since a kid and I started programming at 10 and did all kinds of computer stuff. Um, I studied machine learning around 20 years ago, but it was a bit too early and I got a little bit disillusioned because you couldn't really do all the kinds of

cool things that that the promises were already back then. So, I turned to the dark side. And uh, for the last 15 years, I've been a founder. I've started a couple of companies. I've sold a couple of companies. Um the latest one uh was a speech recognition and natural language understanding company that that was sold to Roblox uh two years ago. And after that I took some time off and I was supposed to do some fishing and hunting and doing some cool stuff with my family but I couldn't help myself. I was just tinkering all the time with AI stuff and uh I was uh about to start my next startup but then the Supercell folks came to lure me into a recruitment process. I hasn't been in one in 18 years. So, it was a bit weird, but they got me lured uh to Supercell and now I've been at Supercell uh for 10 months looking over all of the AI things there. And um yeah, it it feels a bit weird. I'm probably the the person in this room that has the least gaming industry experience, but I know a thing or a few about AI and and startups.
Great to meet you guys.
That's great. Thank you very much, both of you. and really distinct experiences

um and uh great companies doing some interesting things in the space that that that we're going to talk a lot about. The topic that we're here today is will AI redefine player experiences. So I'm going to start with you. What does that phrase mean to you? I mean you know that could mean a lot of things to different people in the room you know about what what what player experiences are.
What do you think we're talking about today and what does that mean in your business? Well, um for me that means like the whole experience from the beginning or when you start planning your game into thinking about how will you do it? Uh how will you take play care of the players? How will you get it into the platform? How will how will you make them stay on your platform? Um etc. So from the birth to the end.
Right. Right. Right. Um, and um, we'll come back to some of that because I know that a lot of what we've been spoken about on stage during the course of this event is about AI for us as developers in in the room, but we're talking here about the players experience from start

to finish. And we'll come back to that. Otto, what do you think we're talking about here? What do you think AI and player experience means to you? What's the player experience?
Yes. So I do agree fully that it covers the whole spectrum but but where I see a lot of opportunities especially in the actual game experience and I just see a lot of opportunity to innovate those those really like game experiences and and how how people you know find fun in novel ways and how they actually want to get engaged uh with the game. So I'm just very bullish bullish on that that space. But of course it has a major impact on the whole life cycle.
Absolutely. Now, um I'm going to just dive in, stay with you, and I'm going to come to sort of expanding on that question because, you know, a player experience could be all sorts of things, immersion, the actual gameplay itself, their personalization, moderation, you know, their experience of the game. We'll come to that. I just want to say one thing. We're going to leave some time at the end for questions from the audience. So,
be thinking of that and we can u make use of our experts here. But also, I'll try and keep my eye out. It's a little bit blinding. So, if anyone wants to kind of leap in at any point, try and track my attention, put your hand up and and we can take questions as we go, too. So, let's make sure we do that. So,

Otto, staying with you then. So, can you give us some examples of how AI could impact the player experience? In fact, any real world examples that you've seen that um either from your own work or elsewhere that you think is relevant to this? What what has impressed you with how AI could impact the experience?
So, uh I'm just now focusing on the actual games because like I said, it can impact everything and I would have examples of everything, but just to to focus this conversation on the actual games. So, we're now starting to see the first kind of AI first games emerge in the market. they're not hitting mainstream yet, but we're seeing the first examples. Uh at Supercell, uh we are heavily investigating the opportunities of of what AI actually allows you to do and how it allows you to re-imagine the player experience. Um already now in the market we've seen a lot of examples of different kinds of NPC interactions with clue extraction games and and and different different kinds of um novel tests of gameplay but we haven't really seen the killer new AI feature. Uh what I personally enjoy a

lot I like those claw extraction games but there's one specific game just to give you an example of like totally novel kind of gameplay. Uh how many in this room have uh tried a game called Suckup?
Suckup. Suckup zero a one. Yes.
Congratulations. Sorry.
Yes. Nice. Nice. Nice. So, that's a great example of of a totally novel kind of game play where the player actually goes around a town, a 3D town, and knocks on doors and and speaks with the NPCs and tries to lure them to actually invite them in with all kinds of excuses. Right.
Such a marvelous. I really recommend you to test that. Even though the game mechanic is just an NPC interaction, it's a really cool novel kind of gameplay.
Yeah. And it's good because it's given you an objective that you're trying to use your conversational skills to manipulate it a certain way.
Perfect. I've got an example to add to that, but hey, I'd love to know what what you've seen. What examples could be something you've done yourself, but what what have you seen in the real world of of AI in player interaction?

Well, um um I recently met uh a founder in uh in Germany who made a game for teenage girls where they could uh chat with an AI. And what made me a little bit worried was that in that game uh the girls, what did they talk about? Well, they talked about self harm. And uh and I asked him but what what do you do when they talk with the AI about self harm? Well uh we don't have the expertise in this. So we try to giggle it away and make them talk about something else. So I also see that AI in games uh we also have an um responsibility to take care of the people that actually comes into the game to Yeah. to live their lives because it's not the difference between what happens in the real life and in the game anymore. It's a part of your life.
Absolutely. And we're definitely going

to come back to talk about guardrails and things in a minute, but you have just reminded me of of a game experience that I want to share. Actually, I saw it at GDC this year. Uh I met some guys who their background is actually screenwriting in Hollywood and they've created a game called Operative. I don't know whether you've come across operative. It's actually not been released yet but it's but I had a go on the proof of concept and essentially they're using speech generation like 11 Labs and it creates one of those kind of artificial reality kind of situations. So you go on and you join this game and you're given a number to call. So you call this number and it's a person. It's obviously artificial voice but it's very good and they say I'm in hospital right now. really need you to to go and email this thing. And you're sort of pulled into this spy story where you have to kind of do real world actions in order to interact with this person. And I thought as a way of narrative that's quite interesting, right? That's a different way of of doing things.
So that was quite fun. So look out for that. That's guy's called operative.
Um I'm going to skip ahead actually. I have got lots of questions, but you raised the point there about our obligations and about god R and things. Let's go to that question. Sorry.
No, no, it's good. No, it's good. Um throughout the player experience then whether it is
having a conversation with an AI NPC or

whether it's you know, um, interacting with other users somehow and so on. Um, and and an AI involved in that on on forums or whatever, but particularly when you're talking to to an AI. One of the problems that I guess people perceived very early on was you're relinquishing control to to the to the model, right, to have a conversation and uh, and that comes with a responsibility in terms of what it is. I mean, how do we deal with that? Is is in your mind, I mean, this is the way you work.
How do we tackle that? How do how do how seriously do you take that responsibility and what can be done?
Yeah. Well, I mean, uh, and this is where technology and especially AI shines, that's where they can really do their job. Uh, so I think by using, uh, AI for the good stuff to protect the players in that specific community, that's uh, what we should do. And also the AI can be trained on detecting the really bad stuff and the good stuff. So we should not forget about that either that uh when you have a player in your

game for example that uh um experience toxicity most of them they go out and never come back
right
but if you also in your game they have a very good experience they will come back and they will also bring their friends so yeah uh using AI to uh to understand what's going on in your community and in your game is something you should do.
Absolutely. And in fact, you're right, the the toxicity is has a huge impact on retention and I know there's actually a talk about that later on today as well, but that absolutely is is important. I mean, also, do you have a have a take on that? As someone who comes from a from, you know, a tech background and so on. I mean, how do we ensure that when people in this room perhaps are developing with AI and they want to have these player experiences,
is this the most important thing they should be mindful of of what, you know, what's going to happen in that experience and how do they get a handle on that? I think it's a very interesting product design question. You know, how are you able to design a product where where you

as a designer want to control the actual player experience in a world that is nondeterministic like like AI is nondeterministic. So it kind of like philosophically fights against the whole core principles of you know designing these experiences and and a lot of people feel that as a as a kind of like threat but I actually see that as a marvelous opportunity and and there's a lot of debate in the industry around like okay what what happens when you know there's no control and and what happens when everybody has a personalized experience what do people talk to next day at school or whatever. I actually see that as a as a massive opportunity and and and there's already today examples of games where where the the player experience can be different. Let's just take the example of like the the Zelda games, the kind of Breath of the Wild or or Tears of the Kingdom where you know players are given a puzzle but they can solve it in very innovative new ways and still they go back to school and talk about the the games. It's not like like we can't talk about it because we solved it differently. Of course, the designers will design the frameworks and and design the overall experience and it just becomes more like like hurting

sheep the design of the the game experience rather than hardcoding things. And I think that's that's Marvel's opportunity.
Yeah. No, that's that's a very good point. In fact, just as you say, in traditional role playing games, presumably you will approach it in a slightly different way. You have a and there is some narrative beats that but you expect to solve problems in a in a unique way. Um, that's that's interesting actually from a from a narrative point of view and something that to be conscious of I I suppose for when when using AI to make games is a lot of it is how you train the model and how and how you set your your parameters and things. we talk about guardrails and things, but also I was remembering having a conversation quite early with someone about the the challenges of creating a narrative where it's nondeterministic and your you know the control goes elsewhere
and someone pointed out that things that are very obvious to us as storytelling humans like what a spoiler is for instance you have to actually tell the model that that's the case like and so for instance the answer to the question who's Darth Vader actually depends where you are in the story like how much of the story you personally seen and so actually making sure that that's a part of your your model experience is an interesting challenge for developers

that when they can't control it.
I love that example of the Darth Vader to explain that.
Um, so you know, let's let's talk a little bit about that that other blockers to using AI right now at runtime because that's the thing we're talking about. There are different issues about using uh a model in the in the office to you know to to to create things but actually using it during playtime. What are the biggest challenges to to involving that for anyone here who's working with that? I mean is it scale, money, cash, size, the model you choose? what what what kind of considerations do we have to have? I don't know. Do you have a take on that?
Yeah. Well, um I would say everything and also uh and also asking yourself the question is AI the answer to everything? Obviously not. So also I see so many companies they're using AI just to use AI. Uh you shouldn't do that. Maybe it's just analytics you're looking for. Uh you can do it in in an Excel sheet for example. uh but also using the right AI

models. We see that there are so many great models out there but you have to use a whole rainforest to make it run through uh all the computing. So uh yeah get more insights on the problem you're solving and what results you're looking for and also the data. If you have very bad data, it doesn't matter what you're doing, then you will not get the answer you're looking for.
Right. Right. Right. And also, is that is feel the same?
Yes. Yes. And and I I kind of like I see AI including machine learning and I see that as a toolbox that that we we use the right tools and sometimes AI or machine learning or LLMs or genai or whatever isn't the right tool. So you have to be like smart in choosing the right tools for the different situation. And at Supercell we also already do a lot of like uh production highscale traditional machine learning and that's that's already kind of like totally normal. But but what is what is kind of like slowing us down in adopting the

let's say bigger models the language models the image models the audio models the the video models all of that one there's like two reasons uh that are like um like slowing us down. first is really understanding what are the player experiences that make sense that actually are fun and are engaging and engaging in the long term. And that's something that we're heavily innovating and it's not a trivial thing. It's it's a very difficult game design problem. How to create fun? You know, everybody can create these cool tech demos and that the Twitters and LinkedIns are full of those ones, but something that actually keeps players engaged. So, that's number one. And the other one that is um slowing us down is naturally the cost side. uh because the moment you move to bigger models like traditional ML models are something that you can easily do the inference on the server side without totally blowing your unit economics but the moment you start to move towards text models, audio models, image models, video models, the model sizes start to blow up and then the inference cost start to blow up and then especially for mobile freetoplay it can it can seriously uh screw up your business model. I'm not personally that

uh worried about that. many people in our industry are. But um as an example, one of the reasons why Roblox wanted to acquire my previous company was that we had built an inference engine that was very uh dynamically able to do the inference on the end user devices or on the back ends or or pretty much on every any possible uh uh like piece of of you know available compute available. So so I in my personal experiences I know that it's a solvable problem but it does require a lot of research and a lot of engineering to solve. So I'm not that worried about that. That's why at superstar we're now especially focusing on the player experience innovation.
That's interesting and it's because it is a very good point about that. Um I mean many people in the room were here when Tommy Thompson gave his talk earlier on but he gave a great example of this about the predictability of cost because um if you know if you and this is an example you will have all heard earlier but in case you didn't see it. He was saying you know if you have a thousand people playing a multiplayer game and then you scale up and you get successful and it's 10,000 people you kind of know what that cost's going to be. But if you're making a a game that

uses AI at runtime, every user is using it differently uh and expecting different inferences and so actually predicting that cost is the is is the challenge and I guess it just means understanding your game and you know getting the data back and and and working on that and that that was a challenge which is an interesting observation I think
and and the whole variable cost when they are starting to be bigger it's actually in a way a bit bit annoying that your best players who spend most time with you that that you love are also creating the most cost for you and that's a little bit annoying. Yeah. Yeah. Yeah. Some of this as well is also about um again choosing the right model and optimizing the model and this kind of thing. I had a great conversation with a uh young woman um uh uh in uh from a company um called Status. I don't know this there is a game called status you can play and essentially it's like a kind of fake social media feed right and you're kind of playing social media feed and she was saying that when they launched this game they were trying to push this huge prompt into claude and it was costing them a fortune as they got more successful and actually they the reason they just they managed to reduce their cost hugely by just using a different model right and actually spending some time with the prompt and

just you know and switching I think they're on deepseat now or something but and some of that is just some trial and error but also just being sensible about you're trying to achieve right
and and coming from the traditional kind of like a ML hide. You know, the whole idea of using a general model or for a very domain specific task is such a massive waste of waste of resources. You know what why would you want to do inference on a very domain specific task on a massive model that is able to do python you know to to you know trip planning doesn't make any sense to use such a powerful model for that one. So, so that's also something where where you know I see the future uh of of being smaller um domain specific fine to distilled models and a kind of like this kind of mixture of experts type of of architecture for for smaller domain specific models.
Yeah, perfect. Um I'm going to move on to my next question but do we have any on what we've discussed so far from from the audience? It will be a great opportunity at the end but um going oh there's a question at the back here. Yeah, why not? You want to um can we get a sorry it's quite a busy room but can

we get a a mic to the chap at the back? Sorry I'm kind of half blinded but um yes having to take your question at a relevant point.
This is how vampires feel.
Yeah, I think he's in the next room. No, it's
there we go. Thank you very much.
Hello. Oh, okay. Amazing. Hi, I have a question for Otto. So, it's regarding uh what he said about costs earlier on. I think that's a very interesting line of discussion. So as we know now that um you know AI is quite heavily subsidized by you know VC funding and whatnot. So you know a lot of people suspect that as models become more efficient it will actually get cheaper. But I suspect it'll actually get more expensive as you know they start relying on more profitentric models. So my biggest concern is regarding indie developers and under developers utilizing these you know AI models especially like for image generation for text generation um you know the on the cost side because you know if they're running things um you know using you know APIs like GPT tokens or whatnot then I suspect cost might actually become like um unaffordable for

many developers especially once they have thousands of users. So I wonder to auto what do you think is the solution for that going forward? Do you think some something like locally run models is a solution or is there an alternative like you know another company coming about using an open source model and you know but even then that would need to be subsidized because of the compute. So I was just wondering to Otto what do you think is the solution for the far future rather than just the next 5 years where things will probably still be subsidized?
Yeah thanks thanks good question. Um I I doubt that many like major mainstream games will actually integrate like third party like like GPT or claw APIs directly in the game loop. I I I I I my prediction would be that that for you to be able to run that at major scale uh you probably want to do something different where you actually control some of the inference and in in that case you would probably look at the open source models and when it comes to the kind of cost base and the VC funding and and that one with open source models you aren't relying on the inference cost of the the the cloud providers. So from

that perspective that side of the the cost equation is solved. And when it comes to the kind of R&D side of uh developing the new foundation models you know you can always uh use those bigger models to to train your smaller models like like for example DeepSeek we used use the US models to actually train the train the models. So I can see a lot of opportunity for gaming companies to take open source models and then uh use the big models to actually train them and then fine-tune the models for their specific use case and then distilling them to smaller ones that where you can do the inference inference uh on on significantly lower cost basis and going back to what I was explaining on the mixture of experts models like said that when you're solving a domain specific problem you don't need to have a model that actually can do anything and generalize you just need to be have a very good model that is really good in that. So, so I see the future where where many of the use cases have been are being powered instead of these massive foundation models uh with some with some like prompt there and some

instructions rather with smaller uh specialty purpose models that are really built for that specific use case and in that case the inference costs and the kind of compute needed is going to be significantly lower. So that's that's that's how I think it's going to be done. Plus when that is then moved to the end user device there is no variable cost for for the developer.
That's a very interesting point. I don't know whether you wanted to add anything to to that.
No, I completely agree and also one of our most efficient models is the one we developed back in uh 2019. Uh so we were also doing AI before when it was just just a nerd. So
yeah. So uh I absolutely follow you on that. But it is an interesting broader question though and thank you for that that was very specific that that um there are kind of two schools of thought I guess that that AI will empower indies in a way that it won't for the big AAA companies because India can maybe take more of a risk and the bigger companies you know will be less inclined on the other hand to to the question there the

flip side of that might be that only the bigger companies can afford to to use the the AI at runtime because of the costs or for the development involved and so on and so India's might be a disadvantage I mean do you see it swinging either way are indas in good place.
It depends. Uh
yeah, I I I I think that the technology will be democratized uh with with all of the open source uh models being made available and with new kinds of fine-tuning and distillation techniques out there and then leveraging those big foundation models to create synthetic data to actually train your domain specific models. So I just all of the tools are already there. So, so I I uh of course the challenge is for smaller indie studios that you do require some kind of like AI competence in the team, but I also believe that in the future a core game team needs to include somebody that is a specialist in AI because it's just otherwise you won't be able to, you know, create the games and add any kind

of AI features into the games. So,
interesting.
Yeah. And humans are still needed. I think that's also our experience. uh you can use the AI for the very
boring tasks like training the models but not for everything not for all the thinking and because then it can go very wrong.
Yeah.
I tell you what before I move on I've got I want to come back to some of the challenges around a but actually this is a good I'd like to ask you a little bit more about um about your company and your product. Um you mentioned this at the start. Tell us a little bit more about the the the kind of user experience of of of AI in moderation about what you do. what what what if someone hasn't used your product before or or or you know worked with your business before how what how is that process how does it work?
Yeah. Okay. So what we do is we analyze uh all content in real time and we do it by if you have a chat or some other content on your platform you send it to us and we send it through

a huge number of different AI models that's working together. Um and we not just only uh analyze the content but also the player behavior. Uh for example, how uh uh the players are behaving in that specific game. Uh and we have trained our AI uh on a very good level. So for example uh we are able to predict uh cyber grooming uh spam bots uh super early so they can be stopped before any harm happens. That's uh that's the way we use our tool.
That's great. And of course there are particular challenges to different community environments. Uh the language that the AI is is is good at predicting that that that simple algorithmic rules won't. For instance, when we're playing games, we might say, "I'm going to kill you." And actually, that looks really bad on Twitter, but less so on a forum for Call of Duty. Right.
Exactly.
Yeah. And that's maybe the best thing you can tell your friend. It's like,

"Oh, go kill them." Uh, so context is everything.
Exactly. Context.
And uh if you're just using a word filter, that could be uh then you're ruining uh the player experience.
Yeah.
So context is everything.
Exactly. Good. Good. Well, look, I'm going to move on to my my next question there, and it was something that we have spoken about in this room a little bit earlier on today, but also comes up a lot in conversation, and that's about how the sort of adoption of AI both by developers and players. There's there is some skepticism out there, and a lot of skepticism out there. Um, we're a self- selecting group. People are here because they're interesting and and and that you're interested in it and and and so on and often people are working with it. But actually when you, you know, talk to, uh, talk to players, as Tommy was showing this morning with, you know, Steam feedback and so on, it puts people off even the fact that there's AI involved in the game game process. And it really depends. There's been a number of surveys out there, you know, I think Google Cloud did a survey where it said nine out of 10 developers are using AI. But then you can go somewhere else and there was another survey that I that I

that I saw that said that actually there's a huge amount of skepticism and like 70% of developers don't want to get integrated because it's difficult. So it really depends on who you speak to. So how do we move forward with the industry? Right. There's there's obviously some people are very driven by what they see as an ethical implication there.
How do we use technology to its fullest by accepting that there's there's there is it about education? What what do we do and how what's your experience with this? I don't know, do you want to take us into that conversation?
Yeah. Yeah, sure. So, um for example, in our case, um we have we are pretty aware of like what AI can do and what it can't do. And it's pretty obvious for us that that AI will not replace humans, but AI will actually give superpowers to to to the folks working working on. And I think that that's the big problem is that there's a massive misconception and that like all of the kind of social media, LinkedIn, Twitter are just boosting that that kind of misconception of of you know AI replacing. That's absolutely not what's happening. And at Supercell we have made a very strong stance that that AI is used to actually provide our excellent experts with the

tools that they can be be more effective. Uh but it will change work and that is scary. It it is a massive change in how work will be done and um it actually abstracts the work uh to a higher abstraction level uh by actually um automating a lot of the repetitive task that is done and this is not anything new. Similar things happened in the industrial revolution for uh repetitive manual work. Now the same thing is happening for repetitive knowledge work. So there's nothing new in it. And back back then you know 150 years ago back then there was a lot of you know skepticism and problems and challenges but but we as a society made it and and that's something that that that that we are talking a lot about like what does it mean for people working in the gaming industry and there's like this evolution happening in these crafts but it's not like replacement but it's an evolution in the craft. So you can think of it like like we we can like um segment our people into to developers, creatives and others all doing knowledge work. for developers

um it kind of like moves uh it is an evolution from like just you know writing software into becoming more like an architect
and and for creatives it's it it it's more moving from actually you know doing the pixels uh into actually an ad type of a role and and for kind of knowledge workers it's more like evolution from a doer into a manager where where all of these craftsmans now have you know massive amount of AI AI interns or AI team members to to help them in their work. But AI does not have taste and and that's something that is very important. AI will not be able to create the games of the future because we need game makers to actually, you know, figure out what is fun and what what's what's cool
and and you were also asking about like what how will this impact our industry? I think that that here I travel a lot lot around the world. Um I spend a lot of time in San Francisco where we have an AI innovation lab and um I just you know see this massive difference in how AI adoption in gaming is seen across uh

Europe, US and Asia. And if you look at the Asian companies today, they're fully embracing AI. Of course, there's some, you know, like they have less regulation and and maybe even less ethical considerations or I don't know if that's the case, but at least they are very bullish and very aggressive on in doing that and they are moving really fast. And then I'm I'm like I'm seeing these western companies and the western kind of like industry being skeptical and not moving quickly. And I see an analogy here to the consumer electronics industry in the US in the 80s back in the days. You know the Asian companies started to copy the innovations coming from the US in the consumer electronics space especially in Japan. And you know the US uh consumer electronics industry folks were very arrogantly saying that that made in USA will always be the stamp of quality. you know this this the consumers will never choose this inferior product quality. Fast forward 10 years, what happened? Consumers chose the cheaper cheaper alternatives, lower quality h and the market shifted there

to Japan and or Asia. And especially Japan, they upped their their quality game. They became the innovators and consumer electronics industry for decades pretty much like crushed in the US. And I'm seeing a very similar pattern now happening in gaming with with Western gaming companies having a very conservative and sometimes even a little bit arrogant um response. So that's that's what I see like like something that needs to happen in our industry.
Interesting. Interesting. Yeah, I get you agree.
Yeah. And absolutely and skepticism is good. But also what worries me is all the uh all the regulations. We especially in Europe, we have decided to regulate everything before we know what it can be used for. And but what we know is that the bad actors out there on the internet or in the game store who wants to make the bad stuff they will use the AI they don't care about regulations right
uh so we also uh I completely follow you

on that we and if you don't look into the tools and understand how it can make you better uh then I think we will yeah step back and give the chance to someone Yeah. Yeah. Yeah. Absolutely. Isn't it that that saying that you know the the the the antidote for for bad AI games is make a good AI game. But but um the um uh but it's a really good point. your point about automation and this is I was working quite a lot on this recently about how AI is the the latest in a long line of us automating processes right and and I um you know within living memory Photoshop was frowned on for in a lot of you know media companies to start with and right now this week in the movies Tron Aries has just come out and again many youngsters in the room may forget that the original Tron was denied an Academy Award because using computer graphics was cheating right and and and you know of course that that sentiment didn't last long through that generation but that's where we are now and um and so on. Look, we've only got a few minutes left. I'll try and get some time for questions. Um um I want to just dive

into a final couple of practical comments then.
Developers here who are maybe here because they're just starting out on their AI journey. They're interested in in you know in starting work with AI in games. What what practical advice do you have for them? Like one concrete thing they should try. They've come away from Pocket Gamer Connects Helsinki and this AI summit uh game changer summit. What's the next thing they should do to act on the things we've been talking about on stage? I don't know. What's your first piece of advice for everyone?
Well, uh I'm from Norway and uh it could be the same uh advice we give for people when they go up in the B mountains. You haven't been there before. Talk to the people who has been there and don't start out. It's good to start experimenting, but uh do some research first. That's my that's my advice.
Perfect. Yeah.
Awesome. Um mine mine would probably be which is heavily influenced my uh by my founder background is that you know first focus on the player experience. Nail that first and really find the best

experience for your players. Worry about scaling later. Uh do not start to prematurely scale. Yes, there will be challenges in the cost and the scaling and many technical challenges for you to solve when you need to scale it. But most likely most of the the projects will not even go past the kind of product market fit of those features. So focus really on the on the player experience first and then the scaling problems can be solved later.
That's great advice. That very very good advice and I think that you know one of the things I I would say is look at the problems that you're trying to solve and not not like AI is cool so what can we do with that but more what what are we trying to achieve for our players and is it AI
the process that will make that work? It may not be incidentally you know but but maybe it will be. Um we're going to take questions just a couple of minutes just on that then we've talked a lot about different kind of player experiences the use of AI in different parts of the of the experience from um from onboarding to the game itself to moderation and so on. What AI enabled player experience do you think will be common place the next

time we're here in a few years time maybe on stage that isn't now that is maybe just starting out now what do you what do you think the the we could predict the future we'd all be billionaires right but but um what do you think is the way it's going what experience do you think is going to be common place next time we have this conversation
yeah well uh gaming has always been a big part of people's lives so I think like playing a game today in the future you can uh have a talk, have a continuous experience with the characters in that game in different part of your uh digital life or maybe also in your physical life. So extending the the characters into different um
yeah into your life it it yeah
yeah interesting absolutely I think you know well one of the other hot topics at the moment is like sort of trans media and the extension of games into other things and the role that AI can play in that is very interesting
um alto
yeah I would say that that that that uh might sound like a hot take but uh 100%

of the games who are not specifically wanting to be 100% human handcrafted will h have some elements of the game built by AI or using AI as a tool by humans. Uh so that's that's one. Second of all, I would say that that um most of the games published in a couple of years time will have some kind of AI features at least on the meta side and maybe they're starting to enter the kind of core core gameplay too. And uh third one is that we are going to see novel kinds of entertainment experiences uh where where AI powers new kinds of ways how players actually spend time interacting with the IP that gaming companies have have built. So it's going to be a very different world in in two years time. That's my prediction.
Absolutely. That and in that you agree. The other the other thing of course is there's so many other experiences that we can't cover in like a a 40-minute session, but yesterday we were speaking about um kind of monetization and user acquisition and about the as you mentioned the meta there and about what you see in store and the degree of personalization we're going to see as well through through AI even if it's

just analyzing the the data and serving you a different offer from someone else. That's a given I think probably
today already.
Yeah. Today already. Exactly. Exactly. Exactly. Um right I think we've got just a um like a a minute or two overall. Right. Oh, I was just going to say one of the things I've noticed, we keep talk we sorry, one of the things I've noticed, we keep talking about AI as one thing,
but there are so many different areas here. You didn't even talk about answer generated search. I think search is going out the window and answer generated experiences are going to change uh discovery completely.
Oh, it's it's so hard on companies like us to publish websites and they're used to SEO.
It's just going to completely change everything. And there's so many different areas about the way we function, the way we build, the way we analyze, what we should build, the way AI is going to have a massive impact. Uh we probably have got quick one quick question. We got a question over here.
I see one here in the middle if there. Oh yeah, perfect. You got it. Thank you.
Thank you. Uh well uh hello. Uh at Cortion, our company, we we try to help games to integrate AI, but usually when we talk with these uh gaming studios,

they are a bit lost that exactly where they need these AIs to be used. Sometimes they are so excited and they want to use it everywhere and sometimes they are like have a very I would say dreamy plan for what they want to do. We we talked about how important it is to understand where they need it but how they can go through this process or understand that where can have the most uh like they they could have the uh like effective way to use the AI integrated in these systems.
Maybe I can share how we look at it at Supercell. So, so when we pri prioritize the the AI opportunities across the group, we have many games, we have many functions. It's it's already a pretty big company, but we we prioritize those opportunities um from three perspective. One is obviously the business value. Does it make sense? Um the second one is the the feasibility because there's a lot of hype around AI and especially if you read the news, there's so much stuff coming every day in the news of

technology that is not yet production grade ready. So, it's kind of like a funnel where that the technology matures. So you have to be very um in uh to be able to actually you know find the signal from the noise because there is so much noise. So that's the kind of second dimension you know where the AI technology is production ready and then the third one in our case is you know where the people are ready you know that means that the teams are receptive and willing to do that and and you know ideally of course we would want to prioritize everything from the business value perspective that would be the logical case but for example in our case actually the technical feasibility and the people receptiveness are actually very important parts of driving the prioritization decisions. But hasn't that always been the case for any adoption of technology anyway? I mean, there's a reason why we call it sabotage is because the workers chucked their shoes into the machinery because they weren't getting paid properly.
Exactly.
And that's the fundamental thing that we forget is we think they were anti-technology. They weren't.
They were anti not getting paid.
I'm afraid.

On that note, I'm going to have to unfortunately say thank you very much and and sort of end it now. Thank you very much, guys.
Well, thank you and thank you very much. Thank you
for me. What this [Music]