How You Can Learn AI Through Gaming

0:00Your favorite neighborhood spot grows with Square. Indeed, my favorite neighborhood spot has quickly become Todd Snyder in Williamsburg. Todd Snyder is one of my favorite menswear shops and has supplied me with all the clothes I have needed this quite hot summer. Every business has different goals, but Square is the business platform that supports them all. From opening a new location, selling something new, or just expanding their reach. Indeed, I've seen it with Todd Snyder. In Square, also, you can get real-time insights, so don't wait for end-of-day reports. Go to square.com forward slash go forward slash Real Vision to learn more about how your business can grow with Square.

0:43That's S-Q-U-A-R-E dot com slash G-O slash R-E-A-L-V-I-S-I-O-N. At Real Vision, we're closing the doors. But not how you might think. You see, this autumn, we're launching the new Real Vision, a platform built around the universal truth that knowledge times tools times network equals success, your success. It's the biggest transformation in our history and brings together everything you need in your journey from information to knowledge to wisdom, all in one place. That's incredible AI, charting tools, networking, economic data, watch lists, notes, and a whole ton more. We start rolling out to our current members at the end of August.

1:33And from August the 15th, we're closing the doors to any new members while we focus on that. But you do have one final chance to get in that door. Until August the 15th, you can level up for a whole quarter of Real Vision just for the price of$20.14. When you go to realvision.com forward slash last chance, you'll see why we chose that price in particular. It's something about Real Vision of old. You'll get to experience the new platform before the general public with no obligation to stay after that three months. And a price that works out for like$6 a month, it's what you call a no-brainer. Anyway, I hope to see you on Real Vision.

2:13It's an incredible community. And my God, this new platform is going to be extraordinary and will change as many lives as possible. That's realvision.com forward slash last chance.

2:39What's going on, guys? It's Ash Bennington. Welcome to Real Vision Crypto Daily Briefing. Before I bring in our guest today, I want to talk about something that's really cool that's happening right now at RV. It's mint season number two for the Real Vision Collective, and it's happening right now. The aim of the RV Collective is to bring all your favorite NFT communities together while giving you the knowledge you need to navigate Web3. Season 2 is the next step in this journey. To create this collection, we've pulled artwork from nine of the most significant COO projects out there. Excuse me, CCO, Creative Commons, of course.

3:14And transformed them into three unique mashups. If you want to help us keep building the super community of NFTs, head over to realvision.com forward slash collective. That's realvision.com forward slash collective to learn more and to mint your own season two NFT. With that said, let's bring in our guests. Joining me today are Brandon De Silva, co-founder and CEO of ArenaX Labs, and Wei Shia, co-founder and COO of ArenaX Labs. Brandon, Wei, welcome to the show. Thanks for having me, Ash. Guys, it's great to have you with us. I spent, as I was saying off camera, I spent the morning watching videos on YouTube, what you guys are doing on this.

3:58And it's really cool. It's really interesting. Obviously not investment advice. It's very early, very high concept, but this is really cool tech that you guys are playing around with. Guys, give us the big picture. Sure. Sure. Yeah, I can kick it off. So essentially what we're doing is we're Trojan horsing AI education through a game because games are great of abstracting away complexity. It writes to something as complex as AI. We're actually bringing forth a video game where people kind of just play and they subconsciously learn about AI. Yeah, so talk a little bit about how you guys are doing that, what the gameplay is, what the mechanics are, and what the ultimate goal is in terms of this AI education project integrated with gaming.

4:49Yeah, absolutely. So a bit more about the game. It's specifically a player versus player platform fighting game. So essentially there is a platform that you have to try and not fall off of and you have to try and knock your opponent off of. And the intention of the game is to actually teach your AI how to fight. So you're showing it what to do in various situations. And over time, it should actually learn how to fight like you. We're showing a clip, by the way, on the screen so people can see it. I mean, the most striking thing when you see that clip is that it looks like kind of any other platform fighter just by looking at it on the screen.

5:26But obviously, it's not. Exactly. And it looks like that because what's happening is you as a human are imparting your skills into this almost like a digital representation of yourself. Right. And I'm sure we'll get there later. But it's really cool that you can actually have these things on the blockchain because now you have a digital representation of your skill in a game that can actually play on behalf of you represented on the blockchain as an NFT. But I'm sure we'll dive into that a bit later. Yeah, Wade, jump in. Anything to add? Yeah, I think what we're really excited about is the fact that not only have we created a kind of a genre leading game, it's something that's never been done before.

6:11When we use AI models and embed it into the focal character in a game, it really starts to change the experience that the human players have with the in-game characters that they're able to train and build up and enhance over time. And we hear a lot of the feedback from our early community players where they really grow this kind of intense emotional bond with the characters because they really had to train them up every step of the way into a very proficient and exciting fighter in the arena. I think beyond that, I think one of the missions that we wanted to kind of create and achieve with AI Arena is to really democratize AI, making the knowledge base and understanding of something that's perhaps a little bit threatening and technical and it may feel out of reach for some people.

6:59We want to make sure that we're using a medium of gaming to kind of bridge that gap a little bit better and more efficiently and bring people into the world of an AI in a very friendly way and use the engagement loop of the gaming to really encourage them and motivate them to explore more. So it's like a stepping stone to learn more about AI through the medium of gaming. Gosh, it's so interesting that Rush to mind. The first that, you know, gaming interfaces seem to be, at least in my view, and I've got this thinking from a gentleman named Alex Fernandez, who's been on Real Vision a number of times, really are kind of the most elegant portal into the digital world.

7:38It's something that people, particularly young people, are very familiar with. It's very natural and very intuitive. The other thing that struck me when you were talking about this idea of being with your AI avatar every step of the way, it reminded me a little bit of the late 1990s phenomenon of Tamataguchi's. This idea that you have to constantly keep an emotional tether to a physical device. It's just really interesting when you think about what that interaction looks like. yeah i think the the example that you gave there um is very appropriate except i would say um it's like the actual it would be like the natural end state of what tamagotchi could have became right which is in this case um because they're ais they're very flexible they can they can theoretically learn an arbitrary uh permutation of like a combination of skills so what you see in the competition as a result of different people's perspectives on training and different skills that they have imparted on their AIs is very expressive AIs playing the game in different ways.

8:45And that's really exciting. And it leads to a type of game experience where it's really an infinite game, an infinite competition, because there's really no skill cap or ceiling. And we've lifted that constraint away from gaming. And all you're left with is really the observation of like the ingenuity and creativity of individual players and what they were able to teach their AI to perform in the arena. So what's the gameplay experience? Is the idea that basically the actual live gameplay takes place during the training phase and then it gets automated from there and the two AI bots essentially compete with each other?

9:24Yes. So the way that it happens, you're right. So the actual kind of the traditional kind of gameplay experience is on the training side and there are three steps is really actually showing your AI what to do first. And then there's the configuration and inspection, which is really kind of contextualizing how your AI learns the information that you just showed it. So when you're collecting data, you're literally showing your AI, I want you to punch or block or jump. And then you take that information and then you contextualize it and say, I want you to learn these skills in this way. That's kind of the configuration step.

10:02And then the final step is the inspection, which is actually the feedback loop where you're able to diagnose at any moment in time what your AI has learned, what its current policy is, which is its decision-making framework, and understand how its current behavior is going to be in a game. And you can use that as a feedback to loop back into the beginning of the loop again and try to figure out how you can enhance and improve its performance in the future. So it's a very iterative kind of training process. and then once you're comfortable with your AI's performance, you can actually submit it into the arena where it autonomously battles other AI characters trained by people.

10:42So what's the platform that this runs on right now? Right now, it's entirely run on the browser. We've actually built our game engine from scratch to support this. And the reason why we wanted to launch it on browsers is because we want it to be as accessible as possible to as many people. and so there's kind of nothing more accessible than just going to a website and start playing immediately, right? So no downloads required? No downloads. No, just going to the browser and start playing. Yeah, that's really cool for two reasons. One, for accessibility and two, obviously lots of people in crypto are concerned about security and have hesitation at least about installing things on their computer.

11:28Absolutely and rightfully, so yeah so tell us talking of which tell us about the blockchain tie-ins yeah for sure um so there there are a few but i can tell you some more on the immediate horizon so um essentially like like i mentioned um you are training these characters right and so they are these digital representations of your skill um and then what we can do is we can embed that into nfts and and the reason why we actually started going down this path we started it like three years ago is because um we saw the technology of nfts and we thought it could be so much more than what it was just being applied to which is just a pointer to uh some some image at that time and at that time um i was a quant at an investment firm and so i was building these machine learning strategies um and me and way uh we're doing a lot of the crypto research back in the day for the firm.

12:29And we started looking at NFTs and understanding the tech. And we're like, holy smokes, these things are just basically containers for anything that's unique and has value. And so the light bulb kind of went off like, oh, like we can embed like some true utility in these containers, right? Which is these machine learning models that actually do stuff. And that's kind of how it started. And so, and so, yeah, So our characters can be represented as NFTs that sort of are tradable. And the idea there is that if you are truly good at playing the game, you can actually improve basically the value of this NFT through training it to basically play better.

13:10Right. And so it becomes this meritocratic thing where if you're better at playing the game, you should kind of like accrue more value, if you know what I mean, because now this thing is able to beat the majority of other NFTs. does that make sense hey everyone we're going to take a quick pause and hear a word from our partners we'll be right back yeah it does in terms of the the way that you guys think about it what data is actually stored on chain in terms of the utility functionality and integration with the game itself yeah yeah that's that's a great question so um initially it was actually the entire artificial neural network was stored on chain and we took it a step further and actually encrypted the neural network and stored that on chain however we had to kind of take a step back from that because it's going to help if I contextualize the progression of our game we started off with a much simpler proof of concept game that could run on a smaller neural network And as you increase the complexity of a game, the larger and larger neural network is required to do the task, essentially.

14:22And so we basically upped our game to the current platform that it is because of the fact that we found that previously a lot of the strategies were converging onto the similar thing. And it was getting a little kind of repetitive to watch. And so we're like, okay, we need to expand the search space of possible strategies a lot, which meant we needed to increase the size of the neural network. And unfortunately, we started hitting constraints in terms of like when we wanted to save the model on chain, it actually required like 10 different transactions just to push all the data on chain per model update.

14:59And so it just wasn't feasible. So what we ended up doing is just storing that on IPFS and just storing kind of the pointer to the model on IPFS. And then when it comes time for runtime, it basically checks the blockchain, pulls from IPFS, and then basically like restores that model and starts using it. So it's essentially a pointer that maintains a kind of security state. And then you go against IPFS where the actual model is run. Exactly. It's the usual constraints, right? It's bandwidth, speed, latency, all that stuff. 100%. Yeah, yeah, 100%. Yeah. This is interesting to me because from a gameplay perspective, it's kind of a new model.

15:40I'm not aware, at least, of any system that uses essentially live play for training and then automated execution for competition. That's an interesting idea. What's the feedback you've gotten? Are people getting their heads around exactly what that means? yeah i mean uh we've been fortunate enough to um accrue a very um fervent uh small community that have been building with us since uh i think it's like 2021 it's been a couple years now and they've been giving us feedback every step of the way um reception have been great i think the common uh common feedback that we get is that people consistently say that this is a game experience that they have never come across and it's still to this date something that's truly unique, whether it's Web 2 or Web 3.

16:26There are aspects of the game where you segregate kind of the training, which is live, with the battling, which is like autonomous, and it's actually like infinitely scalable, right? Because if you think about it, the saved version of your model is 100 % available at all times. So you can actually, you know, theoretically, you can like simultaneously use your asset to battle across many different competitions concurrently um and and these these kind of features make the gameplay uh very unique and differentiated and it's it's very exciting to watch because as you can imagine uh when you go through the process of training it's kind of like we say it's kind of like raising a child and you're so vested emotionally in how they um kind of perform later on because of all of the hardships that you went through to get them to perform at a certain level, it's immensely rewarding to watch them succeed.

17:22So that detachment point of watching your AIs perform actually heightens the experience of watching these competitions. And there's actually really good kind of theory-based reasons why that is. But we see that the players actually really do enjoy this. Yeah, and just to double click on this to reinforce the idea, which I think I understand, the better player you are, the stronger you'll be in training. Therefore, the stronger those characteristics your character will inherit. So essentially, when you're watching it play, you're watching, in effect, the cumulative sum total of your own gaming skill and experience play out, though in an automated format, based on the AI learning algorithms that you've embedded in the system.

18:08Absolutely. That's exactly it. And then you're also watching what someone else was able to do with theirs. Sometimes even when you lose, you're learning a lot by observing how other people's AIs perform, which you take back into the training of your AI to try to one-up your opponents and continue to collect that ladder. So in other words, if you see a move or a particular combination of moves that's defeating your character, what you can do is then basically include that in your own training so that the model teaches your character that so the next time that you fight or you can counter that move i mean that's just really interesting conceptually yes and that that leads to the infinite nature of this competition because you know it just constantly morphs as a function of who's playing the game and the best players will continue to push the meta forward um so it's fascinating to observe and and where is the community right now is it on discord yeah our our core community lives on discord um we have a server and players from all around the world some of our most active communities are from like southeast asia china um the kind of eastern europe africa and obviously a pretty strong presence in like english-speaking western geographies as well but it's it's quite diversified globally so it's very exciting to see people from all backgrounds all walks of life and geographic kind of locations come in and compete within the arena in this like very novel game mechanic.

19:37So when is it playable for everyone on the mainnet? We're currently targeting a mainnet launch either end of this year or early next year. So we're making good progress towards that. The game is built, is functioning, is playable. Obviously, still some refinements that we have to make along the way. So right now where we are at is we have beta testing available. People just have to come into our community, sign up, and you can get access to the demo. If you are actually quite good at the demo or if you're demonstrating that you're taking to the game quite well, we invite those individuals to what's called exclusive access, which is a smaller group of people that are playing the game at a very high skill level.

20:22And that's a core base of players that we want to cultivate because when we launch on mainnet we want a very strong set of competitors to lead the way in showcasing the game through really exciting um gameplay content so we're slowly building up that core group of competitive players um and we run constant play tests over time for for that cohort and we actually that's actually how we also distribute our nfts is that people have to come in and earn them. So only for the people who are placing in the top of these playtests will they basically get the ability to mint one of our NFTs. Hey everyone, we're going to take another quick break and hear a word from our partners.

21:04We'll be right back to the Real Vision Crypto Daily Briefing.

21:11Just to take it a little bit more broadly, what does this tell you about the current state of AI and where we're at with it? I think a lot of people think about large language models as being kind of the predominant paradigm, obviously a text-based interface where you're typing and the system is learning this very different in terms of the interface, in terms of the application, in terms of the ultimate framework that you're using it in. What's your general take on where AI is right now and how what you're doing in the game relates to that? Randy, do you want to start on that? I can probably add my perspective on it as well.

21:46Oh, I think, yeah, what if I think... I'm cutting off a little bit. It's okay, you're back. You're back now. Okay, okay, I'm back. So I heard the part where you said the large language models, text interface, and then was the question, sorry, I cut off for this part, but was it kind of like what we're doing is different? The question was just basically where do you see the current state of AI right now and how does this game interface with that? I see, I see. So one thing I'll note is that we don't directly kind of like interface and incorporate some of those other elements. So the two biggest things that we're kind of seeing is generative models, specifically on the art front and on the language front.

22:36Obviously, for good reason, right? It's very easy to interact with these things. Something like a chat GPT, you interact as you normally would with another human being, right? and then you get some output. Similarly with the generative art stuff is like you interact like a normal human being would say, I want this and then it gives you something output, right? So it's very easy for these people to kind of like interact with these things and get like really good results. Kind of where we fit in with this is basically we actually help fill the gap of AI knowledge, right? Because while these things are coming out, there's a lot of people that are looking at this and they're like, damn, this is kind of scary.

23:19You know, I don't know what's going to happen to me, what's going to happen to my job, what's going to happen to the future with AI. And so what we're providing is basically an opportunity for people to kind of learn more about this and realize, hey, it's not so scary. Let's kind of take it once a better time. Let's learn about this thing. Because often what I find is people are scared of what they don't know. You know what I mean? And so where we fit in is we kind of provide a new interface to interact with AI in a way that you actually kind of learn about the AI. You know what I mean? When you're interacting with ChatGPT, you're not learning about the AI behind ChatGPT.

23:55You're just getting a direct output. With us, you're actually starting to learn about it and learn the nuances about actually teaching an AI, right? You're like, I show it this specific thing, and it learned something different. It's like, why did it learn something different? And you start by playing your game subconsciously, just like having this knowledge kind of ingrained in your head. And then over time, you're like, okay, I kind of understand this AI thing, how this thing's working. It's not something to be kind of like so scared about. You know what I mean? Right. Yeah, that's really interesting.

24:26And I think that's spot on when people interact with ChatGPT. It seems like a black box. They focus on the output, the functionality, but not on the process of how AI gets there and what the potential uses, I think, could be in the future. Guys, really, really interesting conversation. It's really cool stuff that you guys are working on. Final thoughts, key takeaways, about 30 seconds each, if you wouldn't mind. Let's start out. Well, let's start with you, Wei, since we started with Brandon first. Yeah, final thoughts. I think, like I said before, we're really excited about what we're building.

24:58There's really nothing out there that's comparable. I think we're at a moment in time where we're starting to, I think, the general population is starting to realize that AI is going to permeate and affect every aspect of our lives going forward. So from a mission perspective, it's kind of imperative for us to really democratize AI from a way that we believe is very effective, which is through the medium of gaming. Like I said before, it invites people to explore AI a bit more. What we're not promising is that this is not a replacement for like a, you know, undergraduate degree in like computer science or AI.

25:37This is kind of the stepping stone. This is the gateway to launch people into an exploration of the space and doing so in a way where it's actually very fun and engaging. So we're quite excited about what we're doing. Yeah, well said. Brandon, final thoughts, key takeaways. Yeah, absolutely. I'd like to just build on that a little bit. I think, like we said, this is not a replacement for education, But something that got me really interested in AI was when I, this might sound silly, but when I saw someone do a simple example of making a cart automatically go to a top of a hill, right? And something so simple was like, oh, this is really cool.

26:13Let me explore it further. And we hope that people by playing our game will kind of spark an interest and want to pursue kind of like AI further, at the very least that they get interested. And before we leave, I just want to end it off with saying it's not as hard as you think it is so long as you reframe it differently. A lot of people play games and they learn these very complex game mechanics because they're really interested in playing the game well. Just think about this and AI and machine learning, these terminologies as new game mechanics. And it certainly opens up the doors for you to really absorb this stuff.

26:50Really interesting stuff, guys. Thanks for joining us. Pleasure. Thanks for having us. That's it for today. Make sure to check out our website, realvision.com forward slash crypto. That's realvision.com forward slash crypto. It's free to sign up for our crypto content. Raoul Pal will be on the show tomorrow. You definitely won't want to miss that one. See you at 9 a.m. Pacific, noon Eastern, 5 p.m. London time with Raoul tomorrow. Thanks again for watching and listening, everybody. Have a great day.

27:34What's up, revolutionaries? Thanks for tuning in. For more content like this, head over to realvision.com and get unfiltered access to the very best, brightest, and biggest names in finance.