About this talk
This talk covers the impact of artificial intelligence on productivity and innovation, particularly in the gaming industry. The speaker, Sergio from Metacore, shares insights on how AI is utilized for tasks such as real-time translations, art creation, and game testing. He highlights the significant advantages AI brings, like faster content creation and improved localization efforts, while also addressing the concerns of artists fearing obsolescence. Additionally, Sergio discusses how AI is revolutionizing game development with AI-generated levels and automated testing bots. He emphasizes the importance of adapting to ongoing advancements in AI technology to enhance operational efficiency and competitive advantage.
Full transcript
All right. Good. I think we can start. Can you hear me? Sound good? Nice. All right. Thank you everyone for joining. I come here to talk to you today about productivity innovation in AI and also to give you a few examples of what the gaming industry and a few other industries are using already with AI. Um I'll start as usual as every presentations as introducing myself a
little bit. Uh my name is Sergio. I'm originally from Portugal. I've been living in Berlin for the past seven years. Uh I've worked in several different companies before joining this last one. So I worked at 1026 Zando and Delivery Hero. And roughly a little bit more than a year ago, I joined a company called Metacore. It's a mobile gaming company, headquarters in Finland, but also offices in
Berlin. And we develop a very famous game that is called Merge Mansion. I'm wondering does anyone here knows Merge Mansion besides the two people from Metaor? I saw one hand at least. Okay. Uh it's a mobile game, a casual game. you're not really the target audience, so I'm not surprised that you don't know so much the the game, but it's a very simple Merge 2 game. We
have roughly around 50 almost 60 million downloads now. We made more than 500 million in revenue in the past four years. Um, so yeah, we're a mobile company, as I said, based in Finland. when we want to talk about AI, when people ask me about AI, I think this sentence became very famous in the past few years that uh AI won't won't take your job, but it's
somebody that is using AI that will and I think this is very global for a lot of different areas and a lot of different industries. uh so I think this is really I really uh feel that this is the right way to think about AI and when we look at tech and how it evolved this image is how I'm guessing no one here had to program in
this time but this is how people had to program in the 60s a little bit maybe with some switches and some some codes and some some circuits that you have to turn around and then it would make some calculations and output a little bit of printing saying what is the result? And if you think about the first time that you code that you program many many many
years ago or just a few years ago the the the frameworks that you use even the language that you're using they all evolved and it was the same with AI. Today uh the best AI framework or the best AI tool is one tomorrow will be another one and fortunately the way that we develop apps and games and how we develop software will continuously evolve. So we have
to adapt to that world and we have to continuously be prepared for that evolution and that's what AI has been bringing also. So I wanted to bring you a few examples already how AI is being used and especially on the gaming industry which I'll be focusing on right I mean this is some of the cases are of pretty recent companies that have been launching a couple of
things other cases are a little bit more like obvious that this has been used. We start with translations. So uh for those who work not just in gaming but in developing platforms you know how this is important. Normally translations it depends on the industry but normally it goes through a bunch of different companies or different applications that you have to run agencies translating hundreds of people sometimes
working on this uh with with AI it's very clear that you will be able to translate much faster eventually even get to a point of real-time translations when you look at games specifically you know you have localization is very critical on how you do it and the context of translations is very important recently uh we've seen that the gains that you can get out of this is
that you still need human supervision of course but it's much faster to get a content translated instead of taking weeks it takes hours the other part that also has been evolving significantly it's been in the art creation let me tell you something who works in gaming here besides the two people from no one so I discovered something very interesting that when I joined also the company was
my this is my first gaming company we I thought that to develop a game was super fast and to put a demo of the of a game, especially a mobile game, would take just a couple of weeks. Um, it was a misconception. I've learned that this is much harder, especially on the part of creating art, creating content, creating creating uh characters for the game. And this is
all manual done by people. And it takes weeks and months to actually just put something out there to start testing it out. And most of the new games that you develop, most of those uh uh experiments, they go to the garbage. They go to the bin and you start from scratch. you fail a lot. You experiment with new games all the time. We are a company that
we're experimenting with new ideas all the time and we're throwing most of them to the to the bin. So imagine having people creating art, creating new characters, creating new design for a game that after just a couple of weeks of testing, you throw it to the bin and you never use it again. This is where you know text to image that is currently being used like a
lot with Mid Journey and Deli really brings a really good innovation for you and and speeds up productivity for us to launch new games. Of course, there's always the part that a lot of the artists that create this feel like their work is threatened by AI. But when it's done with a collaboration and when it's done by you using the content that is generated by AI and
then designing on top of it, it actually brings much more uh impactful to the company. One example also that I bring here and that this is coming from my previous experience uh in food apps right when you open a food app to order some food uh if the restaurant doesn't have images of the food there's a big tendency that you will not order right it increases the
the conversion in more than 30% if you actually have images of the food but there's also an aspect that is like sometimes the quality of the picture is horrible or the quality of the picture doesn't really show what it means Right. So when you look at this, the picture on the left was taken by the phone of the owner of the restaurant and the picture from the
right is then AI generated. H and this is from the previous company that I worked and I mean the image here you can't see it all because I had to fit for the presentation look a little bit better but basically you have a much better perception what the food is and you can and the conversion rate of this is much higher. All right. And this is done
also they have a pipeline basically this company that runs all the images that are generated by the users and runs and suggests new versions of what that image could convert much better. Another part that AI has been used in the entertainment industry has been with voice generation with doubling. Um, and this is very interesting because it's uh it allows you also to not just because of the
localization but also in terms of accents or sorry or in terms of uh the different type of characters and different types of voices that you want to generate. It's very hard to do doubling also it takes a lot of time but if you actually do with with AI it improves the way and the speed that you do it. One thing that I've saw also recently and that
works very well AI testing bots. So AAA games for instance are very hard to have testing bots. But recently with the improvement of LLMs, it's been something that you I've seen I unfortunately I asked to to show that video but I was not allowed. I saw a AAA game that is about to be launched and how the testing bot was completely killing in the game and it
was really good and really advanced and they could test things that in the past it would take them weeks. probably when you finish a game, when you're playing games and you finish the game and if you look at the credits, did you see the amount of people that had to test the game, it's probably sometimes bigger than the people that had to develop the game because there's
a lot of manual efforts and the automation is still not in in in the way that a lot of people would like it to be. Testing bots with AI is actually going to improve that significantly. One thing, who plays Candy Crush? Not a lot, but okay, that's fine. But you heard it properly, right? So uh Candy Crush actually presented last year that most of their levels are
already AI generated. What this means is that they have also a model that generates the levels and the experiments. It is still visualized by by humans. So people still go and and and validate if the levels are okay to be launched. But most likely AI generated. Um and since last year they already putting more more focus into this. In the future they expect even that uh it
will be very easy to define new levels by a simple text input that the designer says how he wants the level to be created or generated. Then a couple of the things that you can and you can test this and create it really easily is that you can create icons or memes based on trends. Right? This was AI generated with with the with midjourney where I said
I'll do a presentation that I want to talk about memes and icons you can create what do you suggest and these are the image that create and why is this useful is that sometimes there's trends in the market that you have to uh and especially in gaming that happens a lot you might have to create certain content for that certain season certain season or certain moment uh
that it's going through or a meme that became quite viral and you want to use that in your own app in your own game and instead of spending hours and sometimes weeks creating that you can actually just AI generate most of them. Yeah. Uh another example that we've seen a lot recently and that is you you probably see this a lot when you go to Instagram when
you go to YouTube and you see a lot of these reels most of them now are autogenerated with AI. There's a lot of different tools doing this that basically pick up the most important points that a game has of recording has from a different session and that generates the highlights of that game. Another example. Let's see if this is going to work. So, this is just a
simple example of uh of art creation with with a little bit of video and sound and also the video editing. And let's see if this works. There's some sound on this. Trust me, we cannot hear but it's fine. So, let me tell you that nothing major, right? But it is for those who had to develop this, imagine develop this by hand with the normal tools that you
have of 3D modeling and image creation and sound creation, you would take you days, weeks to do this. This is an example that was created in 45 minutes by a colleague of mine just to see and to test how uh a new art creation would work and the new elements would be working. So this is how it much it can sorry and then of course and for
programming and for engineering a lot of these tools and and there was there there's been a lot of conversations in this in this in this u in this conference about many of these different tools who uses any of this one of them at least. Okay. a lot, right? And you probably seen what an improvement in the way that you program, right? We've seen a lot at Metacore
an increase of unit testing for instance because people always leave tests for the end and not creating. And this these type of tools will actually like make you much more efficient and in the era in an era that productivity is everything that companies are measuring productivity all over the place trying to optimize more and more these tools actually bring a great return on investment. uh we actually
validated this by seeing an efficiency around 5% of the worth of the time that people will spend developing which is completely worth it. uh we we've been using copilot a few people have been trying out corourser and I think it's it's a great investment and it's a great era of continuously learning out of these tools. Of course they are also like still young and still learning and
they will evolve with time. Um so is the road AI or die? And this is a very personal opinion that I have around it. Um I don't think that every every company needs to work with AI. I don't think that companies that don't work with AI will actually disappear. I think they will most likely lose their uh strategic advantage if they don't work with AI. And I
think the majority of companies that are early adopters of AI have a potential to transform themselves to become more and more efficient and to become stronger business. it recently an example that I would like to bring you. It's it can change the way you read and let me tell you h how it's I was I was doing a review for a friend of mine that wrote a
book and and there was parts of the book that I wanted to dig deeper into what it was meaning into sections that he also had uh uh extracts from other books and I couldn't really understand or I wanted to dig deeper into the explanation that it was mentioned there. So and by using I was using deepsea back then to try it out and ask actually ask actually
ask actually ask actually ask actually ask making questions like what do you think that this means and how can this be an real example in real time and it starts instead of you just reading a book and not having a two-ways conversation you can actually use all of these tools to actually start having a more interactive session and understanding deeper what is happening there. So it really
can change the way that you read. It allows you to do more. For those who are using AI, for sure that you notice that now some things became much easier to do. You can do them much faster, right? And that is exactly what I think it brings to you. It brings you the ability to instead of spending hours doing a task, you will spend only a couple
of minutes doing it. It allows you to become much more efficient. And once again as I mentioned this is a area area of time that measuring all the time and productivity of what you're doing is crucial. So that's what this to type of tools bring you and they can enable you as a company and as a person much better. One thing a lot of people have been
talking is also about AI agents. There's a talk in this stage today about how to create an AI agent which I think is going to be very interesting. But there was also some comments that you know maybe this is the year of agents. Other people say that this is not going to happen. Um but I think AI agents is like AI models right now and and they're
very good on executing one task and as as as mentioned also in in this conference you know AI agents are basically LLMs in a loop and maybe they're very good in the first task but they might not be very good in the in the other tasks after that but I think it's a it's really evolving technology. I think there will be a lot of agents that will
allow you to do much better work and there's a lot of need of handholding these agents still right because they still need they are still babies they don't know how to walk and you still have to learn and teach them um one thing that I think that it's crucial and that it's important for you to probably even see where you the place that you're working how things
are is what we call the state of AI right so uh very much a lot of a lot of the things that you have to go through is just think this from one to five in which stage are you when you start with the stage number one are you aware is probably a stage that I would say pretty much everyone is aware now that AI exists um
and and just that people understand in the company that this is a technology that is out there then are you exploring is your company already using any of the tools or exploring or or or at least uh uh planning to use any of these tools tools. That's something crucial and you can just use things out of the box. Uh basically tools that are already out there, services
that allows you to start experimenting. Third, the proof of concept. Are you developing anything that is already actually being possibly used inside of the company? Are you already developing something that you're implementing or you're using open source? Some of the LLMs that are out there, are you already implementing them? Or are you using any kind of the tools like for instance, we use scenario AI to create
icons and small art. These are the proofs of concept that we can do and then production use. Um so this is a phase that not a lot of companies are already in there but basically where they start putting what they have developed with AI with AI are they putting this into production as is this been shown to their customers. Is this in their app or in their
game and being used massively? And the last the last state is when you basically create a platform out of it. So in this moment and you see a lot of companies here already talking about that they managed to get to that state where they have a platform that is already using AI and they can make it uh accessible for externals right this can also be that you
develop in a way that you make it accessible internally right you can really do this and we did this also in our company to understand in which stage we are and everybody aims to of course become especially the the stage number four production news that you actually put this into practice and you see how this impacts your business. The future is going to be complicated and predicting
the future is always unpredictable. But as we move on, especially when I look at games and and especially when I look at at at the ability that of of running smaller models of of being much more efficient on running those models, how you can then implement this into different games. I I do see that for instance the interaction you have with with non-playable characters in the game
is going to evolve in a way that is going to be very personal and a couple of things that I also think about that is that in terms of um how I see it in engineering is that you know platform engineering is going to be extremely simplified I don't know if you seen the presentation from cloudfare yesterday but actually the what they are talking about I do
see this happening very very fast right that I work I used to work in a company that our platform team was more than 200 people and we all were talking how eventually in the future we don't need that many people to run a platform. Um we also think that observability bots and agents are going to be like quite remarkable. What I uh what happens nowadays is that
you know you have to develop your own observability most of the times you have to create your own dashboards. This will evolve eventually a lot and you also got to get to a point that your first line of defense for incident management for instance is going to be a bot and not exactly a person. and maybe when it escalates then it escalates to a human. So your
own call might might might reduce the cost significantly. Automation testing also to 100%. And this of course is a prediction might be wrong but I think that with the with the advance of AI and the advance of the models we're going to be able to really have much more unit testing end to end testing and and and follow this pyramid of testing in a much better and
much more efficient way. And fourth but not least and and and before the last one sorry personalization will be bettered in all games and platforms. The efficiency of uh of of of our CPU usage and and graphic cards usage with small LLMs with open source LLM will actually be improved significantly that you will be able to run this and you are already able to run this in
your phone but it's just going to improve much more and more and more. Uh and this will allow platforms to be very personalized and games to be personalized. The experience will get to a point that you really feel completely immersed into the experience that you're having. Um and the last point that I have here is that this also brings a lot of responsibility on most of us
that are leaders or or managers in companies because uh it will allow us to to to become better in what we do, faster in what we do and much more efficient in what we do. And this is something that I think that we have access now to a database of knowledge that we didn't have for years and we have it now. And it can really bring uh
build a strong in the company and make you much more more efficient in what you do. And that was pretty much it. I I spoke a little bit too fast, but that's nervous. Uh when it kicks in, you get a little bit nervous, so you have to speak faster. Uh but that's pretty much it. Thank you so much. Feel free to reach out. Happy to be here.
