About – What is AI Gaming

AI Arena: https://aiarena.io

Fasten your seat belts, gamers and tech aficionados! In our latest podcast episode, we had the pleasure of chatting with Brandon Da Silva, the visionary CEO behind ArenaX Labs. What’s the buzz about? ArenaX is pioneering a new frontier in gaming with their flagship title, AI Arena—a game where you’re not just fighting alongside AI; you’re teaching it!

Brandon, a quant-turned-gaming-innovator, is no stranger to the complexities of machine learning. Unlike traditional game AIs, which operate on rule-based coding (think a series of if-then statements), AI Arena goes above and beyond. How? By incorporating machine learning algorithms that adapt based on your gameplay. In other words, your AI companion learns from you, making the experience intensely personalized and ever-evolving.

Beyond just developing a game, ArenaX Labs is on a more profound mission: democratizing AI. They’re crafting their own machine-learning frameworks tailored to gaming, defying the norms and raising the stakes in the AI gaming landscape.

For Brandon, it’s not just about leveling up in a game; it’s about leveling up the gaming experience itself. And if you ask us, that’s game-changing.

Interested? Tune into our podcast to hear Brandon delve into the nuts and bolts of ArenaX Labs and AI Arena!

Transcript

00:00 Intro Welcome to the Player Engage podcast where we dive into the biggest challenges, technologies, trends, and best practices for creating unforgettable player experiences. Player Engage is brought to you as a collaboration between Keyword Studios and HelpShift. Here is your host,

00:15 Greg Posner Greg Posner. Hi, everybody. Welcome to the Player Engage podcast. Today we’re joined by Brandon DeSilva, the CEO of ArenaX Labs. When I first got introduced to ArenaX Labs, I thought it was a really cool concept and I’m going to let Brandon talk a little more about it. But to give a high level, they’re a game development company that specializes in AI-based gaming experiences. Their flagship game that they’re creating called AI Arena is a unique game that allows players to collaborate with AI characters. You can kind of build out your AI and train them how to fight. So it’s interesting to think about how you use this when it comes to the real world. But Brandon, do you mind doing a quick introduction about yourself and talk about what you’re doing?

00:52 Brandon Da Silva Yeah, yeah, absolutely. Hey, guys, and thanks a lot for the intro, Greg. So yeah, I’m Brandon. I’m the CEO and I guess also CTO because I lead the engineering team here. Yeah, prior to this, I used to be a quant building machine learning strategies in the financial markets. But I guess always on my spare time, I would actually build AI specifically for games to test out new algorithms I had in mind. And over time, I started building my own games to create, to start testing new algorithms on. And that kind of like led me down this path to eventually start my own thing where we’re basically creating an AI-based game where AI is the focal point of the game mechanics.

01:28 Greg Posner So let’s break this down a little because it could be confusing, right? Like you say you’re sitting here building AI, right? What’s that mean? And then obviously we now have the Generative AI, which lets you build LLMs and these different models. But what does building AI mean

01:42 Brandon Da Silva in general to you? Yeah, yeah, for sure. So when I’m talking about that, I’m specifically referring to machine learning. So if you think about AI, you can think about it as this umbrella. And there’s two main things under this umbrella. There’s expert systems where you code basically a bunch of rules, right? It’s like a rules-based system, like if this, then that. And the name kind of implies why it’s called that is because you get a bunch of experts into a room and they’re like, okay, I think you should do this in this situation and this in this situation, right? So that’s expert systems. And that’s what most people think of when they think of like game AI, right? It’s a bunch of if-thens. What I’m referring to is machine learning where we don’t have an expert, rather we have people create these learning algorithms that purely learn from data. And the data in our game is basically users playing the game and we’re seeing kind of how

02:33 Greg Posner they play and the fighter kind of learns from them. So maybe this makes more sense if we can give the audience kind of some context of the game itself and what’s happening in the game. Because I did play it, or I watched a preview of it. It’s really this cool concept. So do you

02:46 Brandon Da Silva mind just kind of giving the audience that elevator pitch on what the game is? Yeah, absolutely. So at its core, the gameplay is a platform fighting game, right? For those who are not aware, we have fighters on this platform and the idea is to try and knock the other fighter off the platform. So that’s kind of the core story of like, I guess, game mechanics of like the core game. Now, in AI arena, we kind of take this a step further and say, rather than just the human controlling it, you have to teach an AI to learn how to fight, right? And so part of that is you playing the game, right? Like you’re showing it how to play. So for example, maybe I want to teach it how to edge guard. So I can actually show it how to do that and show it what to do in that specific situation. And in the background, we’re collecting a lot of data on how you play the game and we’re using that to teach the AI, right? So that’s the first kind of part of it. And then next, it’s like, okay, once I have all this data, tell the AI kind of like what you wanted to learn and how you wanted to learn from this data. And then finally, we built this thing called the inspector, where you can actually go in and see what your AI has learned, right? And you can see, okay, it learned this in this spot and this in this spot. That looks good. Let me save it. If it didn’t learn what you wanted it to, maybe you can hop back into training and show it some other stuff. You know what I mean? And that’s kind of the core loop is this data collection, which is you showing it what to do, data configuration, which is you showing it how to learn from the data you collected, and then inspection, which is just basically seeing like

04:14 Greg Posner if it learned what you wanted it to. So it’s kind of high level again, right? There’s kind of a human AI collaboration going on, right? Where the human is kind of giving their flavor of the week to the AI and the machine is going to continue to learn what you’re doing based on your nuances or your

04:30 Brandon Da Silva specialty things that you do in that game itself, right? Absolutely, absolutely. And then the AI will kind of show you what it learned and you say, okay, actually, I want you to do this instead,

04:39 Greg Posner you know, so there’s definitely this collaboration happening with you and your AI. What is training the model like? Is it going to be intimidating to someone that’s picking up a fighting game for the

04:48 Brandon Da Silva first time? Is it going to become as natural as just playing the fighting game? Yeah. So I would say in a previous version of our game, it was certainly harder because you kind of had to, first of all, you have to know how to play a platform fighting game. I don’t think we can get around that because you’re trying to impart your skills on this thing, right? So if you don’t have skills, unfortunately, the fighter’s not going to know how to fight. So that’s the first thing is you kind of have to know how to play the game. We offer sort of this playground environment without AI for you just to get used to the game, right? Get used to the game mechanics, specifically of the platform fighter. And then once you’re ready, you can start actually showing your AI what to do. Going back for a second, it was initially a little intimidating because you kind of still had to understand some things about neural networks in order to train effectively, because we often got the question from people like, why is it doing this? I showed it something else and we’d have to like kind of explain to them it’s because of this, this and this. We just actually introduced something a few days ago, which is called the simple model, where we abstracted so much of that complexity away that people can just almost just like play and it starts to learn. There’s still a few things that they have to kind of get used to, but it’s a much simpler introduction. With that said though, although it kind of like, you can think about a skill on like a gradient, although it kind of pushed up the lower bound in terms of like how good these AIs are, like at the worst skill level, it really contracted the upper bound, and so the idea is almost to give people an onboarding, get them used to the game with this new model and then once they’re ready to level up, they move to the more intense kind of neural

06:28 Greg Posner network stuff. For players, and we’ll have kind of hopefully imagery on our player-engaged site that shows what the game looks like in the videos of it, but I mean it looks like a Smash Brothers type of game, where it’s a platform fighter, where you get to learn that stuff, so do you foresee

06:45 Brandon Da Silva the addition of more characters, kind of the expansion of that game itself and trying to build that out into its own ecosystem, for lack of better words? That is a great question. I’ll answer it in two parts. One is maybe people have to stick around to see what kind of happens as we kind of continue to grow the game. We certainly are planning to expand a lot of aspects about the game. I don’t want to spoil exactly what they are, but I will say the reason why we designed sort of like one kind of marquee character, at least to start, and basically the way it works is that the character can vary in terms of its weight, and depending on its weight, whether it’s lightweight or heavyweight, we will derive a bunch of battle attributes, such as like power, speed, all that type of stuff, and what we’re trying to do, we’re testing something out, is we really care a lot about balancing the game, to the extent that we want every single weight to be perfectly viable, because ultimately what we want is the game to be 100% about skill. We don’t want someone to get sort of like an OP character and just naturally dominate. We want the core gameplay to be you training the AI, not a combination of I kind of partially train the AI, but I got this OP character, therefore it’s able to dominate the leaderboard. I want it to be 100% reflective of essentially like your skill. Now the question is then why do we have different like weight classes, right? And the idea there is because we want it to be almost like less boring, if you know what I mean. We don’t want everyone to converge on the same play style, right? Naturally, when we have sort of like a heavyweight versus a lightweight, the fighting styles will start to vary, right? And that’s kind of what we want to see. We want to see varied fighting styles, but everything being viable such that no one has a clear advantage. I hope that makes sense.

08:41 Greg Posner It does. I mean, right? You think about a boxing game or a fighting game, when you pick a heavyweight or a lightweight, different people are going to have different skill sets associated to that. What makes this interesting concept to me is, I guess, let’s talk about the fact that are these models technically NFTs that are being created about a per player basis, right? Is this basically like your trading card, an NFT that just basically has all your fight stats on there? Is that a fair way to think about this or not necessarily?

09:10 Brandon Da Silva I think certainly in the Web3 version of the game, that is definitely true. I will say though that we have kind of two versions of the game. We have kind of the more traditional version where people definitely don’t need to interact with Web3 at all, right? They can just play the game because they enjoy playing the game. We will offer competitions as well. It will be structured differently because it’s a bit harder in the traditional sense. But yeah, we do offer sort of a Web3 version of the game where the characters are essentially NFTs. So over time, as you make them better and better and better, you can start kind of like trading these NFTs. Maybe someone wants to buy your NFT off of you if you put in hours and hours of work and it’s crushing it on

09:52 Greg Posner the leaderboard, you know? Yeah, I think that concept of NFTs in my mind is one that makes sense. I know a lot of people talk about the ability to have a skin and bring a skin from one game to another, but that means the styles need to be somewhat similar. But if you just have attributes of a fighter, right? And you say, hey, this is the weight of the fighter. This is the punch power. This is the speed. This is the dexterity. Those are almost the way you’re playing a Super Smash type of game or Mortal Kombat or Street Fighter, right? Those same stats and attributes still exist in each one of those games. So it’s almost, I don’t know, obviously, it’s jumping ahead of the gun here, but it’s kind of cool to think about how you can kind of use this data once you start having it. And is it plug and play capable by bringing it to other platforms and being able to kind of just go with it? And I don’t know, right? That’s a tough question to

10:40 Brandon Da Silva answer. Other people would need to adopt it, but I think that’s a cool kind of thought behind it. Yeah, definitely. I mean, we built it with that in mind. I mean, like you mentioned with the skins thing, and even before that, when we first started building this, we kind of looked at the state of NFTs and were like, okay, it’s all just pointers to an image, you know what I mean? And I’m like, we can definitely do a lot better than that and provide actual utility to these things. So that’s why we thought about kind of tokenizing these AIs, which, yeah, it’s not only just the AIs, but like you said, it has all these attributes associated with it that we use in game. Yeah, that was certainly kind of like one of the motivations. And we are building something out so that you can port, actually, yeah, I’ll say it. Like you can basically port things into different types of games. I’ll just say that, of course, there are obviously challenges. I think initially, it’ll apply to games that we build. So for example, if we build like five different games, we can port this core model NFT into different ones. I won’t say exactly how we do it, but yeah,

11:47 Greg Posner we’re certainly thinking through that. Yeah, it’s a cool concept, right? And then when you’re training a game or a new model, you at least have starter data, which you can continue to bring and kind of learn from there, right? Absolutely. So you go from like a fighting game to a kart race, and you have same maybe weight, whatever. Obviously, you got ideas that are cooking in there. So I guess you kind of started explaining this before, right? But what are, in your mind, what are the differences between Arena X Labs and the game AI Arena? Where does

12:12 Brandon Da Silva one begin and one end? Yeah, yeah. So I would say Arena X Labs is the tech company, right? AI Arena is one of the products we’re building. And so for example, one other product Arena X Labs will be building is a researcher competition. We actually haven’t really talked about it much publicly. I think we talked about it a little bit, but essentially you can think of this as a centralized hub to host really fun machine learning competitions, right? This will be more code-based, right? If you think about AI Arena, AI Arena is a no-code game, right? Where you’re just training in AI. You can think about this research competition that we’re building up right now as for developers, engineers, and researchers to kind of build different types of models and submit it to compete against each other. That’s kind of the idea at a high level. So over time, Arena X Labs can be building a lot

13:05 Greg Posner of different types of really cool ways for people to interact and use machine learning. So with, excuse my naiveness here, right? But when you’re starting to do machine learning and training of models like that, is this a model you’re building on top of, or is this a brand

13:19 Brandon Da Silva new model you’re building? What’s the technology that’s powering it? Yeah. So we’re pretty agnostic in terms of like, I’m specifically talking about the research competition. On the AI Arena side, we built our whole machine learning library from scratch. So we’re not building on top of anything. We’re using concepts that people have kind of like published papers about, but we also introduced novel concepts specifically to overcome some of the challenges about applying machine learning to like this specific game. But yeah, on the researcher side, what we want to do is we want people to be able to build on top of whatever frameworks they’re comfortable with. So for example, two very popular deep learning frameworks are TensorFlow and PyTorch. We’re building things out in a way where they can basically train in whatever framework they’re comfortable with and they can upload their model onto our platform and then

14:09 Greg Posner it starts doing its thing. Yes. I was taking some notes here as well. You said TensorFlow, or what was the other one, just for my own knowledge? The other one was PyTorch. PyTorch. Thank you. How do you want to see the game continue to grow? Because I do think there’s different places you can take from that or different directions you want to go. And I think it’s all exciting, right? And as I am asking you this question, I’m also thinking the thing that we often talk about a lot on this podcast is kind of how you manage your community. And I think your community is going to be important to you, right? Because A, they’re giving you feedback in the early days of your game, which is important. And they’re also the ones that are helping build this model. So are you talking to your community? What’s the platform of choice when you do that? Like, how do you collect their feedback? So two very wildly different questions, right? Maybe how do you listen to your community or how do you foresee the future of the game continue to grow?

14:59 Brandon Da Silva Yeah, yeah. Let’s start with the community because ultimately they almost help guide where the game goes. We’ve been doing this for, I would say, two years now where we’ve been actively engaging with the community. And really, it was through engaging with the community that we’re able to actually have a lot of these features that we have now because they’re kind of like, hey, I’m playing this game and this doesn’t really make sense. And I was like, oh, we should add this thing in there to help it make sense. And they’d be like, oh, I want this functionality. And literally half the functionalities we have is because of our community and because of the feedback they give us. And so the way in which we do this is we primarily manage our community through Discord, right? And right now what we have is this thing called exclusive access where we have a group of tight-knit community members that are playing the game constantly that are always giving us feedback. And then if we ever want to ramp up the feedback, we’ll have a competition, right? Like a five-day competition. Engagement definitely skyrockets during that. And then we’re able to get the best feedback, especially since people are very competitive. And so if anything’s going wrong, you can count on them to complain about it, right? So we get usually the best feedback when we have these competitions. And yeah, I think that’s it. Our community, at least from what we see, very much appreciates the fact that we are so hyper responsive where if there’s a bug, we will fix it in an hour, right? Or at least we try to. Or if they recommend something, we’ll try and incorporate it within a few days to a week, if it makes sense, of course. So they really appreciate that. And it kind of keeps them around for longer. Because it feels like they’re actually helping build the product. You know what I mean? And that’s something I really love kind of building out in the open is like, we constantly get their feedback since we started

16:53 Greg Posner with our alpha, constantly iterating, developing and improving with the community. Yeah, it looks like there was, I think, three of you that worked on starting the company, right? Was there kind of a moment in your past when you took a look at the two other guys and be like, hey, we got to start this something? What was the driving motion or the driving force to kind of get this started?

17:13 Brandon Da Silva So a bit of background on the two guys I started this with. One of them was the art director. His name is Dylan. He’s actually my cousin. The other one is Wei. He used to be my boss. All right. So I have very close connections with both of them. We all have very complementary skill sets. So basically, when I came up with the idea, this was like late 2020, I knew kind of I wanted to build this game with this AI mechanic. And so I reached out to my cousin Dylan because prior to this, he was in the animation industry. So I knew he had sort of like, well, I just know for a long time, I know he’s an incredible artist, right? So I was like, okay, I need to reach out to Dylan because obviously it’s a game. It needs to look great. And so I knew he’d be able to deliver on that. So we started getting together, thinking through things, how we’re going to do it. At the same time, I reached out to Wei. I remember I just like, I asked him one day, I was like, yo, I have this crazy idea. Just hear me out. And then I kind of like walked him through it. And he’s like, yeah, I want to be a co-founder, basically. And so that’s how we got kind of the three minds together. And yeah, we just started this. We all kind of saw the massive potential in this. And yeah,

18:31 Greg Posner we were just keen to get going. So when you got started, you said it was about three-ish years ago? Is that right? Or like, this was before the big rise of the trash GPTs and these other, right? LLMs, right? And you mentioned earlier, right, that you talked about machine learning, which I think AI is just a branch off of machine learning, right? That makes it easier for people, right? But now that we’ve seen these new technologies in our life, are people coming to you and asking you like, hey, when are you going to put this technology in your game? When are you going to put this technology in your tool set? Like, is it like, all the time?

19:03 Brandon Da Silva All the time. I have to tell people all the time, like, guys, like that’s not what we’re doing. And partly it’s because, so if you look at something like chat GPT, it’s so easily accessible through an API that it’s not clear to me that anyone that uses it is able to differentiate on a product easily, right? Because everyone’s going to be using it. Same thing with a lot of the generative art models. It’s like, how differentiated can you really be? I think one thing we really pride ourselves on is that we are doing something that we believe that we haven’t seen anyone else do, right? Which is this like deep personal connection with you and an AI and growing with this AI. And so to do that, like, we almost have to, I wouldn’t say it’s noise, but like, ignore a lot of this other stuff that like people are doing. And people constantly tell us, oh, you should incorporate like intelligent NPCs, like, you know what I mean? With chat GPT. And we’re like, no, guys, like, this is what we’re doing. We’re very focused on this. And yeah.

20:02 Greg Posner I think that makes sense, right? I mean, we get that a lot too, like, oh, how are you going to integrate this technology? And we’re like, it doesn’t really make sense, right? Like, exactly. And that maybe sounds good, but in the concept of what’s happening in the big picture, it doesn’t make sense to put that stuff in there other than to say, hey, we now have this tool in here as well. Absolutely. Absolutely. When you were with your two other co-founders, what came first? The game or the kind of the AI or arena X labs?

20:28 Brandon Da Silva The game actually, then the game came first, because this was the thing that we were going to build. And it’s kind of like, okay, we have to actually make a company now. And that we got like a solid prototype. And we saw some initial traction because we started testing it really early with family and friends. And we’re like, it looks like people kind of like this thing, we should probably make a company. And that’s when arena X labs came. And we had all these ambitions, like I talked about the researcher platform as well. But at that time, it wasn’t clear to me that it was going to be under like, called like a parent company. But yeah, that’s just how the cards fell. And we think

21:02 Greg Posner it’s probably cleaner that way to separate it. And you’ve also spoke about a little bit throughout this podcast, it’s kind of those next steps in those evolution points of the game. But A, do you have a roadmap that you communicate with your players? Or is it kind of one of those when you’re this size, you can be agile and kind of switch it up. But do you have kind of these big

21:21 Brandon Da Silva pillar marks you want to hit within the next 30, 60, 90 days? We do. We don’t communicate them all to our community. The reason for that is because well, there’s a lot of reasons, but I don’t want us to come across as like, I’ve seen it a lot where people like, kind of like over promise these things, people get really excited, and then they don’t deliver. Right. And then they’re like, Oh, we’re going to do it later. We’re going to do it later. We internally have very tight deadlines for ourselves. And we want to hit these certain pillars, as you mentioned. But we won’t kind of tell our community that we’ll tell them kind of like, high level, like we’re planning on launching around this time, but we won’t say we’re going to release this feature here, this feature there,

22:04 Greg Posner this feature there, you know what I mean? Yeah, if you want to make sure you set the expectations and exactly things on time. Exactly. If you were to go back to yourself 15 years ago, Brandon, you were still in school here. Hey, what were your what did you dream to be of when you were growing up? Did

22:19 Brandon Da Silva you have kind of your expectations on building a game when you were younger? What was your what was your dream? That is a well, that’s a good 15 years ago is a deep cut, but 15 years ago, I wanted to completely just go into martial arts. But it’s, it’s a tough business to make money in. So apart from that, once I realized I wanted to do something to kind of make money, I was, I’ve always been very fascinated with mathematics. And I remember, at one point, it was in like a calculus class, I like, I learned about this concept of a derivative. And then I was like, Oh, it’d be cool to like, try and calculate the slope of like a like a price chart and try and like, see if you can predict it, where the price is going to go. It was very naive at the point at that moment in time. But I was like, Oh, what if you can approximate this with some, some like function. And that’s when I started thinking about, Oh, it’d be cool to become like a trader, because I’m really interested in mathematics. And I ended up doing that. I ended up being a quant, building ML trading strategies. But yeah, it’s just, I really like math. And combining math with technology, which is basically machine

23:36 Greg Posner learning, it’s always just been like a passion of mine. Yeah. And then add in the martial arts aspect and creating a fighting game. And you got actions of when you were younger. Exactly. I like, I like the idea of going into trading. That was always my dream to learn how to time the market, come up with a weird Python script that can understand it seems so obvious, but

23:57 Brandon Da Silva clearly, clearly it’s not there. Yeah. What stage for the AI arena? What stage are you at in kind of the launch process and what type of help are you currently looking for? Yeah. So right now, we’re trying to gear up for launch. There’s still development work that we need to do. And we’re doing that. But right now, what we’re trying to do is foster like a really good play testing community. So we’re actually, we just kicked this off not too long ago where we’re trying to get people from sort of like other gaming communities into our play testing server to start play testing the game. We care a lot. I think I mentioned before care a lot about user feedback. Ultimately, we want this to be a very enjoyable, fun game for people to play. So right now, we’re just in the process of like getting not as many testers as possible, because after a certain point, it’s hard to keep up with all the feedback, but like a critical mass of play testers to just keep

24:54 Greg Posner giving us feedback, let us polish refine the game and get it ready for launch. So you kind of talked about the feedback. I’m curious, because we do talk about feedback a lot. Where does it go? Right? So when you’re in discord, you’re getting feedback, do you keep this stuff in gear? Do you keep it somewhere else? Like, how do you, I guess there’s two ways to look at it. How do you capture that feedback? And how do you prioritize what’s going to be? Is it going to be from someone that’s been

25:16 Brandon Da Silva playing for a hundred hours? Is it going to be someone that spent money on it? Really, in terms of feedback, we don’t, we don’t put certain people, like certain priority over other people. We take everyone’s feedback equally. They can be, it can be literally, they spent five minutes with the game. And if they have a great idea, they have a great idea, you know? And so actually one of our new play testers, like joined yesterday, he gave an idea and I was like, oh, that’s, that’s a pretty good idea. We should do this. If the idea is simple enough, we can probably implement it within super quickly, right? Like maybe five, 10 minutes, test it out for a little bit and then ship it. Right? If it’s something that’s a bit more difficult, or if there’s a big laundry list of ideas, we use, we don’t use JIRA, we use another, another system called linear. But yeah, we just keep track of all the issues in there. And then, and then yeah, and then we’ll assign it to certain people that we think are like good at developing that stuff. And then, but I will know we don’t take all of the, all of the recommendations. We like, we’ll look through it and then we’ll see, okay, this one’s good, this one’s good, this one’s good. This one is obviously cause like, they just don’t have like much experience. Maybe they just kind of like didn’t actually play the game, you know? Yeah. All feedback is good though, because if they’re saying something, it’s cause they had a certain experience playing the game. And so how can we change their perception

26:37 Greg Posner of it? You know? Yeah. It’s something we kind of, I work at HelpShift. It’s something we stress a lot is collect feedback from your players, right? I mean, you can be designing the best game you think is possible, but if it’s not something your players are playing or they’re resonating with, like sometimes you have to do this slight detour, despite you may not wanting to, but if it’s what

26:55 Brandon Da Silva your players are saying, like they speak volumes, right? Do you track anything like play time by player? Are you looking at player specific data or not quite yet? Yeah. Yeah, we are. We are. We started tracking that, I think sometime in the spring. It’s really fascinating to see player data. I don’t know if you want me to get into it at all, but yeah, we started certainly tracking that. Obviously there’s a big spike in player activity when we launched these tournaments,

27:23 Greg Posner but yeah, that is something that we’re tracking. So I would love to go into it. I’d also want to make sure we don’t go too far into this privacy side of things, but from a high level, like where are you pulling this data into? Do you pull it into some BI tool or are you pulling it into

27:37 Brandon Da Silva something else to monitor it? Yeah, so we pull it into our own system. So back when we started collecting the data, we created our own dashboard. And so we just store all the data in our database, just pull it into the dashboard, and then we just kind of visualize it, right? We have a bunch of different visualizers for time spent playing different game modes, hours per day, all that

28:01 Greg Posner type of stuff, retention, all that type of stuff. Yeah, once you get into that data, like you’re saying, right, there’s a lot of cool stuff there. And there is a lot of data typically you can be looking at, but being able to find the appropriate insights and kind of things, there’s a lot of stuff you can do there. It becomes a slippery slope of how much data you have. Yeah, for sure. Do you have a favorite story you’ve heard about kind of people playing the game so far? I mean, there’s different things I could think of, like I was thinking about if it’s something I could do with my son and kind of each train a model or something educational wise, I’ve been learning how to do machine. Have you gotten any feedback, kind of gave you chills, like, oh, man, that’s

28:38 Brandon Da Silva perfect. Yes, I remember exactly when it happened. It happened December 2021, when we were doing our alpha test. And I remember the specific community member that said this. I don’t know, it was like, I remember it was that moment where I went to my co-founders, I was like, this is why we’re doing this, you know? And basically, so the whole idea behind AI Arena, or one of the big reasons why we’re doing this is to make AI education more accessible, right? People are playing this game and they’re almost like subconsciously learning about AI, more specifically machine learning. And so what happened was that was the first time we did kind of, it was a closed alpha test, right? It was kind of publicly open where we invited certain people from outside, like not the core team, to test it out in this tournament. And it’s important to note, nobody from this, like, player group had machine learning experience prior to this. And so what we found was after the first day, someone else was struggling and we had this one community member start basically teaching them about a certain thing in the game. But when I was reading it, it was like they were teaching them about machine learning. So the specific thing they were teaching them was the importance of balancing the data set, right? Because the way you play the game, like you mentioned before, is you have to collect data by playing the game, right? So for example, if you want to show your AI how to punch, you actually have to go up to your opponent and punch. And so what they were teaching them was the fact that if you only show them how to do something on the left side of the stage, it’s not going to know how to do it on the right side of the stage, right? And maybe you show it how to do it 10 times on the left side and one time on the right, the machine’s going to interpret that as the left side’s more important. And so it’s going to skew the way the machine learning algorithm actually learns. And they were explaining all this without even realizing they’re basically talking about machine learning and really good training practices and data collecting practices. And I remember I was just like, holy smokes, this is sick. And then over time, we just saw that they started learning about different things like this concept of a learning rate, batch size, epochs, like these very machine learning specific terms. And they’re just

30:53 Greg Posner learning by playing our game. And it kind of blew our minds. That’s awesome, right? I mean, I think that’s the best way to learn is just by not understanding or not realizing, hey, I’m sitting here reading about machine learning and AI and all these things, right? It’s just you’re looking at it in terms you can understand the way you can digest it. And I mean, once the community starts teaching the community, it’s kind of like, oh man, we’re onto something here. We

31:14 Brandon Da Silva just need to be able to keep that fire going. Absolutely. Honestly, our community is like, I love our community. And they’re so nice and inviting. And when new people come in,

31:24 Greg Posner they’re always teaching them about things. And it’s awesome. So you’ve been at this for a few years now. And I’m sure those days you wake up like, I don’t want to do this right now. I don’t want to go through this. But how do you get through those days where you know you’re going to be a struggle or maybe it’s something you’re just not looking forward to seeing or hearing, but

31:45 Brandon Da Silva how do you mentally prep for yourself for those days? This might sound really weird, but I drink tea. I’ve like, especially more recently, I got into this practice of like just making tea, drinking it for like half an hour in the morning. And yeah, it’s surprisingly like helps a lot. Before that, I used to just listen to like relaxing music as I’m just like coding and just going through the day and just prep myself. Because yeah, it’s very easy at times to get overwhelmed, especially when you have a laundry list of 50 things you need to implement in the next month. It’s like, oh my God, like, how on earth am I going to do this? And especially when we’re starting, because we didn’t have a big team, right? I was the only engineer for a while. And so I was the one that had to implement everything. And so yeah, certainly at that time, listen to a lot of relaxing music, sometimes listen to motivational speeches, you know, do what I can

32:45 Greg Posner just to like get in the zone and start going, you know? I love it. I usually drink tea at night, helps me calm down that or a bottle of wine usually help me get my heart rate a little lower and just relax. But I do like that. It’s kind of the trials to do meditation type things where that doesn’t really work for me. But yeah, that’s a great way to go into it. Find a calm yourself down and go

33:07 Brandon Da Silva into it. Absolutely. Absolutely. I think that’s all questions wise I had for you today, Brandon. Is there anything specific you want to talk about or share with our audience? Nothing other than just, I hope everyone tries the game. I think hearing about it is one thing, but I think you won’t really understand it until you try it out. We actually got that feedback a lot that some people even said, I didn’t find the appeal when I was hearing about training in AI, but I got it as soon as I tested it out. You know, like they started playing, they’re like, Oh man, like it’s actually learning from me. And it’s doing what I showed it almost like you’re the parent teaching this child. You know, it’s something you just have to experience. I would say try it out once. If you don’t like it, toss it away, you know, but I think most people will end up finding it very unique. And I think

33:57 Greg Posner most people will like it. Yeah. And I agree for people who are interested in AI or even just fighting games, right? Being able to see how it can learn from you, kind of build your own model, not realizing maybe that that’s what you are doing. I think it opens up a world of possibilities of what you can do from within gaming, how you can start training, how you can start doing all this stuff. I think, I mean, if you have endless amounts of funds, I feel like it can be endless

34:18 Brandon Da Silva directions you can go in, but where can people find information about AI arena? Yeah, absolutely. For more information, we have a documentation site, but you can probably head over to airena.io. We’re actually doing a revamp soon to give you like better access to everything. But yeah, you can head over there for some information. We have kind of like a learn more button where you can learn a bit more about the game. Then if you want to get involved in our community,

34:44 Greg Posner you can join our Discord. Yeah. And we’ll have all the information to Brandon as well as AI arena and X Labs on our player engaged website. I really appreciate you coming out today, Brandon. It was great to kind of hear about both companies, what you’re doing. I’m excited because I think this is a great use of AI and a great way that you could potentially scale multiple types of games. So I’m excited to see what you’re doing. And again, thank you for coming on today. Awesome. Thank you so much, Greg. It’s a pleasure being on and yeah, thank you for having me. Yeah. Have a great day.