The intersection of AI and web3

Transcript

today's topic is all about the intersection of AI and web 3 Kyle what do you think about this show I was excited for this one and I walked away with it just kind of mind blown which I honestly thought was going to happen but man this was cool so we first defined web3 and AI so we're trying to make this for if you know nothing about AI or even if you don't know much about web3 we're trying to make an easy thing to understand here so that was the beginning then we dove into some use cases for AI but more interestingly we posed the question of does AI need web3 or vice versa does web3 need AI to actually work and this brought on a very very interesting conversation the summary that being well they don't necessarily need each other we talk about why they actually amplify each

other and to a level that I honestly did not think of coming into this conversation and then we just talked about the game that the guests that we had on the show built which is called AI Arena which is basically a game that teaches AI to players by training and battling these Fighters using Ai and those Fighters are actual nfts and I'm not gonna get into too much about it but I'm not a gamer don't really care too much about games I did when I was younger but this game was actually mind-blowing when we learned what it was I was like oh my God this is where AI is going in web3 I was like this is insane so this was a cool one I'm pretty stoked about it yeah stick around to the end to hear Kyle and I talk trash to each other

of how we're gonna go live on Twitch and play this game but joining us today to discuss this is Brandon and Michael from AI Arena Brandon is the CEO and co-founder of AI Arena and prior to starting AI Arena Brandon LED machine learning and research development at a multi-billion dollar fund he actually built an AI model which manages a billion dollars of assets to this day it's still going even though he's not working there anymore and so really Brandon is I think one of the top developers in the AI space globally and then Michael is part of the business team of AI Arena apparently he has one of the best gamers in the game beast mode he'll tell us a little bit about

that he also comes from a hedge fund background where he focused on asset management and trading so both these guys really deep understanding of both the AI space but also the web 3 space because AI Arena uses AI but also gives you on-chain ownership by using nfts and we're going to talk about how that works together in this episode I'm just so fired up for everybody to listen to this it was just an absolute Banger before we jump into the show though let's just take a minute to hear from our sponsors the future of social media is here and that future lives in web 3 on top of lens protocol web 2 social platforms are broken and right for disruption you see the epicenter of social media is the

creators and yet they are the most neglected web 2 platforms like Facebook Tick Tock and Instagram are all essentially robbing creators of their worth creators are a new type of entrepreneur forming new types of businesses yet with web 2 platforms creators don't own their content to the profiles and that's their product in business instead they are tied to the platforms they choose to create on well just like how crypto is freeing us from Banks web 3 is freeing us from these centralized platforms on lens protocol creators own their content own their profile and even their social graphing followers in the form of nfts this allows you to move freely from one social application to another with your content profile and followers moving along with you lens protocol enables

self sovereignty for your social graph and interoperability across the internet at Webster Academy we believe this is the future of Social and that's why we've partnered with lens to ensure that the path of social media is heading in the right direction visit lens.xyz to learn more today unlock protocol is the nft membership protocol for web3 Content subscriptions Community Access event tickets are all forms of memberships one of the most common business models for creators and entrepreneurs today building an nft membership opens many new possibilities for your members everything from tailored multi-platform experiences to the ability to sell an unfinished

subscription on secondary markets these are things just not possible in web 2. however in order for this business model to work creators need nfts which are time bound or have built-in recurring payments unlock protocol does this for your nfts better yet with a few simple steps you can create your own nft contract without code what WordPress has done for websites unlock protocol is doing for membership nfts if you're an nft Creator you can't rely on royalties as they are likely going to zero you need recurring Revenue you need unlocked at web3 Academy We Believe nft

memberships are the future of business and community and that's why we decided to build on unlock learn more at unlock Dash protocol.com today we are talking about Ai and web3 with two of the bosses in the space two of the leaders in the space Brandon and Michael from AI Arena so stoked to have you guys both on the show welcome to the show guys how you doing today great pump to be here man thanks so much for having us another fully Canadian show is it Jay that's right that's right we should have a tally on that okay let me set the the ground for this this

episode first of all Jay and I know basically nothing about AI we're kind of just catching wind of it all like everyone else's right now because it's all over Twitter but I want to just give a quick little definition of web 3 of AI and then we're gonna dive into talking to Brandon and Michael about kind of why they or how they're going to work together and why they need each other or if they need each other and just probably blow all of our minds on what the stuff is so web3 I think we all know what this is but let's do a quick review anyway from a tech side of things so web3 is a new tech infrastructure that enables digital ownership and interoperability across the internet right so it's a tech layer that makes the way we use the internet more similar to the physical world and I mean that in

terms of custodian our own Goods custom your own Goods sorry the freedom to move around with those things and those goods from one location to another now we can do that on the internet right and then we have ai artificial intelligence it's a new tech as well that brings logic to data is kind of the best way that I think of it so it's a tech layer that sort of makes robots and machine means more similar to to humans right in a sense that machines can perceive synthesize and infer information right so that's the kind of way I think about its two technologies probably two of the fastest growing Technologies maybe ever happening at like the exact same time and I think we're starting to see them integrate and combine and come together some people call this the exponential age which I absolutely love but anyway

let's start here with Brandon what are your thoughts on those definitions would you add anything to that or does that give a good sort of explanation of the two I think that gives a good explanation on on the blockchain side definitely you guys got that on the AI side the only thing I would add is that you can think of AI as being split up into two subsets with one of them focused on Expert systems and the other one focused on machine learning and so most of the big amazing things you see in the news about AI is actually with respect to machine learning expert systems is basically like you get a bunch of experts in a room and they quote up a bunch of if statements and then it'll base basically do things according to that logic machine learning on the other hand doesn't have any a

priority knowledge right it's a blank slate and it learns through data so you feed in data and then there are various architectures and learning algorithms that you can use in order to basically learn something and perform inference later on and so most of the models that you see like these large language models or some of these systems that are playing like alphago for example these are using machine learning because you feed in a lot of data they learn some stuff and then they know how to do some stuff so I would say that's the main difference how long have we been using AI for I mean blockchain is still maybe what 12 years 13 years in total web 3 like just a few at least the term web3 how far along are we on the AI Journey any idea I think this will depend on if which

definition or which subset you're referring to so expert systems have been used for a long time deep blue that face Gary Kasparov and chess from a while back that was an expert system the more recent advances are using machine learning and so I would say machine learning isn't so far along although people have been writing about machine learning for decades but it was only recently within the past 10-ish years that we've started getting enough compute to make a lot of these systems viable right because they need a lot of data if you look at some of the more recent models there are these massive models with billions and billions of parameters and you just couldn't train that a few decades ago is that why we're seeing this AI moment right now because the hardware is catching up and then we've had basically a decade of

stuff happening in the background and now we're why is the moment happening now I think there have been a series of moments over sort of like the past 10 years I think it started with image recognition that's that was a very big moment with convolutional neural networks where I was learning to to basically tell the difference between various images to a very very high level and actually surpass the human level right that was like a big breakthrough and then after that we've just had a sequence of different breakthroughs in deep reinforcement learning and now it's large language models more recently I guess before the whole chat GPT thing it was these Dolly mid-journey these generative art models but yeah I would say a lot of it's just you have so much data now we have the compute to be able

to actually process and learn from this data and so yeah I think the next 10 years like I even tried thinking in 10 years what it's going to be like and I can't even imagine it's it's gonna be absolutely insane what's going to grow faster is it going to be AI or is it going to be web3 what's going to get an option faster that's yeah that is a great question that is honestly I'm not too sure I hope it's I I hope it's web 3 because I I really do believe in decentralization and essentially putting the power into the hands of the people and AI I feel like it can go either way right it can be really good or really bad which which one has more developers working on it

right now oh like we think that there's about 10 000 developers in web3 is a rough estimate how many developers do you think are working on AI I don't know but I if I and I and I am a betting man a lot more working on on AI I want to get into the intersection of the two but I actually feel like diving deeper into this AI conversation is going to be very useful here can you just give us some maybe some high level ideas of what AI is going to do for the average human and whether that's like the average business the average person like what other than I mean chat gbt has been super cool I'm sure most people have played around with by now I think it had a million users within five days so okay it's gonna be this smarter Google essentially you can

ask it anything and talk to it it's crazy but there's probably so much more what else does AI do that we're not really thinking about you can use it for anything we use it internally for a lot of things just to be more efficient so for example there are tedious things in developing a game and it was taking me on average like two hours to do one of these things and I had to do it over and over and over again so I'm like you know what let me make a machine learning algorithm to do it I'll show it how to do it a few times it learns how to do it now literally what I do is I just drag and drop the model it takes half a second and it's done you're not going to have a job anymore Brandon I know my model's taking over my job AI

will just make so many things so much more efficient obviously there's a big example of your background maybe you can speak just briefly to this I don't want to dive too deep into the financial side because we don't touch our ton about that on the show but can you just speak a little bit on the financial side how AI is really supporting what's happening there if you think about just the markets and trading there is so much information available to you it's almost like you don't know what to pay attention to and that's where machine learning algorithms are really great because you're able to take in all this information and find all these patterns and be able to deduce what is the important information I don't want to get too far into this but it's actually it's a lot more complicated than what I just said because there's this concept

of overfitting which means you learn basically relationships that only hold in the data that you're looking at and don't hold going forward but the bottom line is that we have all this information humans often don't know where to look and what's important and what will be relevant to your decision making and so it's great to train machine learning algorithms to be able to do this for you since they can adjust data way faster than any human can it seems to me that AI is going to make it into every single industry in every single sector and it's going to replace or amplify what we do as humans right I don't think it's going to replace everything but like you said you have tasks that you have to do all the time there's other things you have to do and you can just get AI to do that for you

and probably do it faster and better and you can focus on something else right which is pretty amazing but you also mentioned there's potential for it to go bad and especially if we continue down this centralized route if someone owns an algorithm that's taking over everything this could be problematic so my question to you is does AI need web3 to succeed or vice versa does web 3 need AI to succeed I'll use a different word than succeed and maybe like we have to think about different outcomes right because I'm sure some centralized Authority that manages these massive models and they basically control the world would think that they're succeeding when the rest of the world is not too happy about it right I think what web3 enables is the decentralization and democratization of

these AI models so it was actually one interesting use case I don't know if many people even internally know this but one of the things that I wanted to test out was actually an AI model as a public good that lives on the blockchain and so every time you interact with it it gets better and specifically I was gonna Implement something like a smart routing algorithm to route to different decentralized exchanges to be able to kind of minimize slippage and trading fees and stuff like that and you can imagine that every time you route a trade through there it's accumulating this data and it's basically getting smarter each and each time and it's the same thing that anyone can interact with and it doesn't belong to anybody it just lives in the smart contract and it'll kind of serve the public good and I think that's like super cool versus you

can imagine some Bank like owning this and they're going to charge a spread on this um and they're going to basically monopolize this and so I think it'd be really cool if people start developing these smart contracts with these intelligent beings like living in there to kind of like help out the public in various different like areas I I think that'd be super cool and just one extra thing the you know if you think about code it can be copy and pasted right and so if you want to transfer code to somebody else you know if I'm selling you a model that maybe helps your business run more efficiently or that you demand how do you know that a I'm not selling it to somebody else and copying pasting it how do you know that it's that I worked on this model myself in the first place and I didn't just

steal it how do you know the history of the training this can all be wrapped and containerized in the web3 layer and it enables the transaction of intellectual property which is right now technologically impossible without having a centralized intermediate or if you really trust me you know interesting so okay you're talking wrapping Ai and basically tracking who's going to use this code or who's going to use this model or this algorithm or it could also be a means of selling it so there's AI in terms of trading which I think we've all heard of that you know the Algos that are going to trade and you know there's we've that's been around for a long time but the way that you're talking about using web3 and how that's going to intersect is actually

like wrapping your code or wrapping that AI model in a what an nft or something and that's how you're going to use this as like a as a way to to track it or sell it or do whatever you want with it is that what you're saying yeah exactly and you know there's once you build out the infrastructure on that which is you know what we're trying to do it's really from there who knows what people are going to Spur up how they're going to interact how they're going to coordinate Behavior through it through the incentives at the protocol like arrives to you know it's it's just something that it literally cannot be done in web 2 yet right and so that's something that it's just that it needs if you want the permissionless and exchange of AI code and models and whatnot it just it just

makes it more efficient and and I guess it connects to Brandon's point about the public good is so many developers and like I'm not a developer so correct me if I'm wrong but so many developers seem especially in new technological Evolution phases like the beginning of the internet there's such a sharing of code right and there's a desire to build off of each other but in the past that was just shared online anybody could do whatever they want with it you can't track how it's How anybody's using it in the future you there's no there's no ability to connect it to property basically and property might be strong word but whereas now you

can use a Smart contract to essentially Rapid Control is not the right word but allow it to allow it to create further life well you can also see how it where it goes from there is that is that sort of what you're getting at absolutely and yeah and Brandon I know definitely has thoughts on this and you know personal experiences of why that's you know so important uh not just to machine learn like AI creators which we believe are creators and should be able to monetize themselves just without relying on a on a working somewhere or a centralized intermediate yet interesting Brandon what other ways does web3 and and AI

interact or come together is there other examples or other things where we use AI to better what we're doing over web3 or vice versa where we need to use web3 and the technology there to better what we're doing on AI is there any other examples that you can share I love the the aspect that you can monetize what you make so if I make a really really cool model and I think it provides value to people I can then basically either sell it to them or we can set up some agreement maybe we have an nft that has a licensing right to use this model we can do that another super cool thing it doesn't necessarily even have to live on the blockchain but it uses a lot of technology that we use in web3 which is zero knowledge proofs so for example let's say I own this model and I want

people to use it but let's say you're skeptical that I'm actually using the model correctly I can provide a proof that I'm performing inference correctly without having to reveal what the model is because let's say I think the model is valuable right I don't necessarily want to give it out to people but I can provide these zero knowledge proofs so that everyone that's using it is comfortable and this is again I don't mean to go back to the trading example but this is like insanely important for trading because there's a lot of times when investment managers will be telling you they're doing something when they're actually not and we actually back at the fund we used to actually invest in hedge funds and we use invest in Quant funds and a lot of time we left the meeting like I don't think they're doing what

they're saying they're doing and there's no way to prove it because we asked them and they're like oh sorry it's proprietary we can't share this right and so now if you say if they're like okay I have this machine learning algorithm right we'll say okay every time you make a trade I want you to send me a zero knowledge proof that the model is actually telling you to do that um and so actually when I came up with this idea that was the the first use case because obviously I was building trading algorithms using machine learning I was like oh like this would be super cool we tokenized the model we can encrypt it so no one else can see it but you have proper ownership of it and then we use ZK proofs to prove that you're actually doing things correctly it's shifted a

little since then because we care a lot about Mass adoption and democratizing AI but but yeah that's something that's always passionate to me interesting I want to add one more example that I I was actually reading a thread on that was just mind-blowing to me and if we think of a future where you know right now we have humans working with other humans well we're about to have what we already do if you're using chat gbt where you have basically a machine or a robot kind of working with a human well eventually we're gonna have Machines working with machines and humans are actually just out of the picture I think of like Tesla's moving around and maybe going and getting like their their power and it's done without humans and they need to pay for that power well machines are gonna need to pay for each other for the tasks the

things that they're doing and maybe they can do it with Fiat rails probably not I don't think a machine's gonna hand cash over another machine but it can do it on crypto rails right because it's programmable there is no intermediary there's no you don't have to trust anything and so it seems to me that it makes sense that when machines start working together and there's value that needs to be exchanged crypto rails seem to make the most sense for that as well so that's and also another really cool example I don't know if you have any extra thoughts on that but I really enjoyed that thread as well no definitely I think that that's absolutely correct I think honestly my default whenever value has to be transferred whether it's locally or globally but especially globally I don't trust the traditional Financial rails anymore it's crypto rails is the way to

go all right let's let's transfer into the conversation around the story of AI Arena Kai unless you wanted to anything else on High level Ai No no I think it's perfect to go into it because really what you guys have done is taken web3 taking Ai and packaging it into a game and so it's the perfect combination so let's let's go down that hole for sure yes so Brandon you know you you obviously have had a very successful career you've probably could have done a lot of things why did you decide to combine both Ai and web3 and how did you come up with the idea for AI Arena yeah for sure so back back at the fund we also used to do some research into

any of the digital assets so the way one of my co-founders he used to lead the public markets investing and one of his mandates was in to invest in digital assets and so he had me look into nfts this was like mid-2020 and I had looked into nfts in like the the 2017 era but I hadn't really gotten into them and so whenever I get into something I like basically looking at the source code and really coding it up from scratch for myself and so I started looking to enter these and I realized I was like holy like these things are just containers for anything that's unique and has value and I I remember I went away and I was like yo the thing that I make that's unique and has value is my machine learning models what if I tokenize them and then I was like that

can solve so many problems it can solve this ownership problem where I can finally stop people stealing my and it can stop I can solve the other problem that I constantly saw in the industry I saw this theme emerging in the machine learning space where there was a heavy Reliance on credentialization and I saw very talented people have to settle for jobs that they were overqualified for in my opinion because they had taken the time to self-learn machine learning and they were really good actually and they can actually apply machine learning very well but they didn't have a PhD or a master's in machine learning and so they had to settle for a job that really they didn't want because no one else would take their resume essentially and so I realized that they can actually make a

model they can tokenize it and if it truly is valuable I.E they are actually good at what they're doing then they can actually make money and this should be a viable path for them and then if you layer on top of that the fact that you can actually compete with these models and this was actually before the whole play to earn like thing went wild I actually took inspiration from a project called numeri which is a machine learning competition to predict like financial markets and you can actually earn their NMR tokens and I was like oh that would be also really cool is that these tokenized things that are living on the blockchain not only can they sell them but they can train them and then compete against other people that are training these AIS as well and then we can basically reward them in tokens

which you definitely couldn't do in in web 2 especially if we wanted a global competition which our thesis is that Talent is distributed globally and especially there's a lot of pockets in the world where the inequality is even larger where these people are like so good at this stuff but there is no jobs there and as we started building out AI Arena we actually realized that this definitely is the case and we've got confirmation in this because people are joining our community from all over the world from Africa southeast Asia and they're telling us they're like one guy he was even like I have a p HD machine learning but there's no jobs where I live and it's just like blowing our mind and so yeah we we wanted to build this to give an opportunity to people to monetize doing what they actually love

that's amazing the the reason that you started this comes from such wholesome Roots it's no wonder you want to spend your time making public goods you're a good human thank you for that we need more people like you let's let's explain AI Arena to the audience so they understand because we're talking about it now so can can one of you break down the game and how it works the game itself so visually very similar to Smash Bros right and that's probably where the difference ends so that's it yeah it's a it's a player versus player battle fighting game right with the exception that each character that's battling is it's intelligent it's an nft right and it evolves and it and it trains and it you train it and then

it fights autonomously so it's super unique first of all the experience of training and raising an AI once you get it out to the box it doesn't know how to do anything it's it's literally a dummy right and you're sort of injecting basic training into it and the training part is fun you actually have to play the game to train it right and it picks up on your Tendencies it learns the skills you put into it right eventually then you can go into What's called the AI inspector which Brandon will definitely speak to which is like looking into the brain of the AI right and now you're looking at how does it react in certain situations that's where you can fine-tune it and add Implement strategies into it right so the reason why we're extremely different first of

all there's this aspect of it's almost like you're raising a child right the closest one yeah you have an identity you feel like you're attached to it because it's picking up on your Tendencies it's fighting how you want it to fight then it's on chain you own it so that's one way it's different second of all there's the aspect that you actually need a play to fight so a lot of criticism on simulation games is that okay it's like football manager whatnot it's fun but I don't get to play it right and a certain point you want to play something else you play no to train this you actually you have the better you are at fighting games you're involved right so that part if you like playing games you'll be satisfied so that's that's a second game Loop and then the third and final game Loop

is when once you've trained it up once you feel that it's competitive right once it's beating you up you're trying to beat and it's beating you up you're like all right can't do much more you let it fight autonomously Evergreen competition it's ranked right it's a ranked leaderboard that that element of it fighting autonomously creates what we call we we have a saying the dopamine hits different right because it all game Loops are it's a dopamine Rush it's a dopamine hit and Neuroscience papers have proved this when you are not in control of a certain event right but you have some input into it when you reward or fail it's a way bigger swing in the experience that you feel the the analogy can be like

professional athletes or the parents of professional athletes or the parents of somebody yeah that's that's writing an exam a kid that's writing exactly that the parents sometimes are more nervous than the actual people because the pupils in control the pupils thinking about other things they're maybe involved there might be the shot but the parent that's not in control is almost like waiting and that's the feeling of watching your AI because you have a hand in it right and but also at certain point you let it deploy and simulate when we know that this game Loop is exciting based off our Alpha tests like we've had people saying they were literally screaming like at their computer because hey I did something that they did or did not expect it to do those like alphago moments that like

move 42 when when the world's best Alpha go player was just completely Bamboozled by a crazy move from the AI like you see a lot of that right and most games make their entire living off getting one of those game Loops really right we're elegantly combining all three and we're not forcing it just happens to come about elegantly combining all three three like insane game Loop brushes you know that it's never been out there before you've literally leveraged the game Loops of parenting right it's actually mind-blowing right now that's exactly what it is except for this is in fighting of course but I'm wondering is there is there a V2 coming out where you actually raise a kid and you practice being a parent and you end up seeing the

results of the kid oh man my mind every every parent should have to take that course before right that would be that would be so good for the world I like that we promise there's no diapers in this game but yeah if we go even further you know why why is this more than just a game it we spoke about the beginning at the end of the day we're really tokenizing AI right and this is my and and we're creating the rails for this to happen and just how we're abstracting AI into a visual fight we take the same infrastructure and Abstract basically whatever else there's demand for this type of you know interaction right and and given the role given how big of a role AI already plays in our life and

arguably will you know this is this is something that we could see penetrating like all corners of the market or the economy yeah so what's the what's the level of of skill in terms of I don't mean gaming skill I mean AI skill necessary in order to get involved like could Kyle and I who have absolutely no coding background yeah come in and make Fighters and they probably would suck and get murdered by absolutely everybody because we're not developers but yeah I know that your focus is education which I love I love that you're coming at this from a you're you're basically using gaming which is fun to educate people on

at complex topics such as AI I'm just curious at what level is the most your users at so a lot of our users actually don't have ai experience and what we found is that having AI experience it does help in certain aspects but it doesn't give you an outright advantage in fact our top five finishers of the last few beta tests that we've ran have not been AI people and that's one of the amazing things that we found and it was probably one of the most proud moments of my life was in the first beta someone that had no AI experience was playing our game and on day two of the beta they were educating other people about Ai and they were like it's sophisticated topics too

about like how to properly collect data to ensure that your data set is balanced so you don't have a skewed policy and we were just looking at each other like whoa this because people were self like subconsciously learning about machine learning by playing a fun game a fun and intuitive game and they didn't have any prior experience and what we found is that when people are engaged when playing this game they will just dedicate more time and they'll ramp up on the learning curve so fast right and yeah you don't have to have any experience to play just one thing to add and I probably should have started with this there's really two games and two competitions the one I just described is the game just the regular game the no code version right and this is the the games that Kyle Jay you guys can just pick up

play with a joystick keyboard train your AI literally need to know nothing about AI then there's the researcher competition the way we the way where we compare this to is like Formula One there's a driver competition and then there's the constructors the cars right so the gamers are like the drivers they're given an nft and then they have to maximize the you know the crap out of it just like a driver maximizes the crap out of his Mercedes or his Ferrari in F1 eventually right we're gonna introduce this researcher competition where this one is going to be coded and this is we're going to attract AI Talent AI researchers to code the car or the nft

and this is going to be like the constructors right now let's wrap it all together we've just created almost a Marketplace for AI where you have demand from The Gamers The Gamers demand the the vehicles the nfts to fight and then you have natural Supply coming in from the researcher they'll tune the parameters the code version and try and make the best fighter and Supply it onto the market for gamers to play right and and that really happened really organically and so yeah there's the code version and the no code version because that's really important so in the no code version if I'm just sitting there button mashing because that's my entire gaming ability then then I can still actually make a good

fighter maybe maybe yeah oh or you can make a fighter that that the the matching of buttons I don't know if that's a good fighter but yeah it'll probably beat yours Brandon there you go okay so we're using web3 and nfts to tokenize AI models and this is outside of the game but I guess this is sort of what you're doing in the game you're teaching people that they can do this now do you see other examples of this happening within AI or other people using smart contracts to

tokenize their AI models or how do we see this progressing I think there there is a few projects that are are trying to apply AI on blockchain now I I I don't want to talk too much about them because I truthfully don't know too much about them because they don't put much out in the public but but what I will say is I would say not all AI projects are made equally so remember in the beginning of this I talked about this segregation in AI of expert systems and machine learning what I'll tell you like I'm pretty confident in this is that most of these people are implementing expert systems and they're disguising it as something more sophisticated and the

reason why I know this one is like a lot of the documentation is very vague but but also remember that I used to invest in quantage funds and so as soon as machine learning started getting popular in coin it everyone would Rebrand themselves as an AI fund literally everyone they were we know they weren't doing AI but they would say that they're doing AI just in order to get more more money through the door and so I would say to the audience that expect that this is going to come that everyone's going to be rebranding as an AI crypto company um and just do a little bit of due diligence to make sure that they're actually doing what they say that they're doing with that said even with that even if they were doing like the more sophisticated stuff I would say

that no one's coming even close to doing what we're doing and it was going back to what what Michael said in terms of we focus a lot on the human aspect and integrating human in the loop and so it's not really as much fun to kind of click a button and it just does everything for you right and you're not kind of in the loop and participating and giving feedback to your AI so I would say no one's really at that level I would love to see more competition I'm an insanely competitive guy maybe it'll make us like go go even further but at the moment there isn't really anyone else doing that in the space yeah and just to maybe to expand on that the you know the closed systems a lot of stuff you see branded on AI is just if statements and it's very closed like

every game you've seen up until now it's the developer has has given you the ceiling they've programmed what's possible and what's not right it's just very parameterized if statement it's important to know when you have true AI like we do in our Tech stack it's really Evergreen it's infinite there's no bounds the AAC can learn from itself and it just reaches a certain point where in traditional games the closed system as brand you have to code every possible field right and you know and just be mindful of that when you see you know people branded as AI whether it's just linear regression or it's true AI to kind of piggyback on Jay's question there as well where do you guys see this going you guys have created this awesome

game and I'm sure you're going to focus on that for a while I'm sure it's going to take a long time to build this game out to wherever it's going to go and it's focused on fighting right and kind of teaching AI as well but I feel like just I mean we already came up with the parenting idea if there's got to be so many applications here with this idea to train Ai and then set it out into the world and see what it does there's more than just competition I feel like I actually was watching a thing on AI last week and they were talking about AI revolutionizing education like school and I think they were using these AI education lessons I guess or modules in I forget which countries it was but to teach English and it could teach English faster than when we learn English in the

first however many grades it was cutting it in half or something something ridiculous it was really really cool so there's so many applications to where AI can go I'm curious if you guys in your team meetings or whatever have thought of like what's next after AI Arena outside of just fighting is there more more things that you guys are thinking about that's an excellent question I mean my answer is going to be biased to what I care about so a lot of my research is in something that's called reinforcement learning so basically learning how to act in an environment through some reward mechanism and so a lot of people that train sort of like it's called continuous control in the fighting game for example you can use reinforcement learning to that and so you can kind of reframe almost any problem as a reinforcement learning

problem because everything you can think of as having a reward I'll take a step back here for a second so like when you're a kid growing up if you touch like a stove top that's hot that's and you get burned that's like a big negative reward and so you know not to do that again right or if you do something and you get like a KitKat for example you're like oh that's awesome I'm gonna I'm gonna do that again to get it right and so that's the whole reinforcement learning and so I think there's going to be a lot of really cool I mean there already is a lot of insanely cool research in reinforcement learning that I have done phenomenal things that playing games but I think this is going to extend to the the real world and just being able to train robots basically how to interact and so

I I think I think Tesla's working on this that this like humanoid robot to basically act as your assistant right and so I I think I think that's coming okay but it sounds like so right now we're in the phase of we are training the AI to to learn okay I got a feeling we're gonna get to a point where the AI and the robots are going to be training the humans to learn right yeah it's yeah is that where we're gonna go with this because that's well that's great I'm I'm almost I'm almost certain we are because like just chat GPT is like an excellent example of it it's true like we go to it for help honestly God we're we're stuck on a coding question we're like yo chat Chief and T like how do you solve this and it starts telling us and so I

definitely see people going to these systems to teach them for example like you mentioned with the English example let's say I want to learn Mandarin and I want an efficient way to learn Mandarin give me syllabus you know what I mean and then it will print out this thing and say okay start by learning this this and this I'm sure with all the information available to these models on the internet that it can condense it down and find the most efficient way to learn something so I'm super pumped about that because you can ingest more data more efficiently you know absolutely makes sense and I think what's so cool is if you use the education example there is a great example of how Ai and web3 can work together AI can provide the ability to teach to give you the data you need to

allow you to improve to give you reinforcement learning and then web 3 can give you the verifiable credential that says that you did achieve this type of learning and it's trustlessly on chain and now you can get a better job because of that or you can prove to somebody that you actually have this skill because of that all around the world not just by calling up your University and asking the print your diploma right the other thing I think about is like Supply Chain management right like obviously AI has huge potential in Supply Chain management but then web 3 also a key part of that ensures that whatever is happening on the supply chain is verifiable and can

prove ownership of these Goods as they move and then can be interoperable if they're moving across platforms I think we could we could almost break down every single industry and show how Ai and web3 work together and I think that that's why really understanding that the different layers that they provide are so key to to like the Innovation that's coming right and that as I think you said I don't Brandon I don't know if you said this on the show if you said this before we started recording but they don't necessarily need each other but when they do work together they really do complement each other yeah absolutely and like when I said that I I'm coming at it from the

perspective you don't need this for example You could argue like back in the day you don't need life you know what I mean um yeah yeah you can get by but it's definitely much much better so yeah like you don't necessarily need web3 for a lot of this stuff but it makes it way more efficient and so yeah why why not right and and I think the oh sorry Kyle sorry to call you off the I think the possibilities are just as endless as you know they were thinking about the internet like in the 90s right there's this there's this technology maybe the hardware hasn't caught up but how is somebody in the 90s gonna predict you know if we spoke about this apps like uber Airbnb which are enabled by the

internet enable completely different business models the transfer of services and it's just not possible without nobody in the 90s could foresee that right that what that technology could enable down the line and I think very similarly web3 and Ai and they're they're merging are going to enable business models that in hindsight might look obvious but for now seem impossible and it all starts with firstly creating the infrastructure for it helps if it's an open system and they can feed off each other web3 definitely like Brandon said it doesn't need AI but certainly helps helps that cause absolutely absolutely let's go back to AI Rita for just for a second here I'm curious if you can just kind of give us a lay the

ground of like how far along are you guys in terms of is this thing Fully live do we have a bunch of users is it growing fast is it like where are we at there and then what's next for you guys what's what's phase I don't know what phase you're on right now but what's the next phase I guess yeah for sure I can talk a little bit about this and then Michael feel free to supplement this but we're actually launching our beta on Monday or I don't know when this is gonna go live but this goes live on Wednesday so okay you will have just launched it'll just launched so yeah we're we're doing that so initially we ran sort of betas on the previous version of the game which is a much more simplified version mostly a proof of concept and then we realized we

had to kind of go dark we went dark world seven months or something like that and just completely rebuild it step it up a notch and yeah this is the first time that the public is going to experience the the full loop of the new and improved aarena game in terms of like track let me just ask quick is that the no code version or the code version the no code version the no code version we we tested both of them in the previous version of the game we're now just launching the no code version we're going to launch the coded version probably I don't want to commit to anything but sometime around Q3 of next year just because we we want to build it out first with the no code version get a lot more traction get People Pumped about this so to the point where

researchers recognize it and they're like okay this is like something cool to work on uh you know what I mean it's a lot easier to get interest from Gamers than it is for researchers because they have so many interesting stuff they can dive into especially now with like generative art these large language models we want to really get them excited about applying reinforcement learning to a really fun game um yeah I'll play you one-on-one what do you think maybe give like a week for us to train our our Fighters or something and then we go live on the show let's do it on the wall up next week streaming on Twitch I feel like we need to stream it on Twitch is that what Gamers do these days no I'm not a gamer but I think so I love it I love it oh yeah well one thing I'll

just mention about this beta is it's a it's a closed beta we're running like a mini tournament that there's gonna be 10 teams kind of competing against each other it should be a lot of fun and then yeah and then after that going into next year we're gonna focus a lot on rolling out a free-to-play version of the game because we're essentially going to have an nft gated version where they can actually earn tokens and then a free to play version where they can basically aspire to get there but eventually work their way up until the into the nft Gated version and the free to play version is really important because we don't want people that are interested in learning about Ai and playing the game to not have the opportunity solely because they don't have for example enough East to buy like an nf2 and so we

want to really make it available to everyone because we we say it and we truly mean it that we want to democratize AI we want literally everybody on the earth to be able to have an under standing of this technology that's going to change the world it's amazing where you said it's an in eth so I assume you're building uh in the ethereum ecosystem can you talk about where you're building and why you've chosen that as well yeah sure well we we recognize the importance of network effects and at the time of us kind of building it ethereum definitely had the largest Network and it's still today it does of course the scaling issue is a bit of a problem and so what we're doing is we're launching on arbitrim that's the L2 of our choice we tested out a few scaling Solutions and arbitrum happen to be it's sort of the

best one they have the most responsive devs and so yeah so we're going to be deploying an arboretum when we initially did the tests on arbitrum our community really liked it it was like really quick transactions and yeah so we're and low gas fees and so yeah that's where we're going to be building interesting and so beta comes out next week what's the what's the phase after that when will it be more open yeah so we're going to be doing a few closed betas so we we have a bunch of people that sort of applied it was very oversubscribed but we had to kind of narrow it down to 100 to like each of these each of these datas we might expand in the future betas to be like slightly larger larger and larger

but essentially after we finish these betas we're going to try and roll out a free-to-play version so that literally everybody can try it out awesome so before we get into to our speed rounds as we as we always do I guess people can't play right now even though it is it is live when they're listening to this but if they can't play yet what can they do can they follow you on Twitter go to your website is there anything they can do to try to get in on the next beta when it opens like where where should people go better yes yeah definitely I would say head on over to our Twitter our Discord our discord's where all the action happens that's where you can earn sort of like access into the following betas so yeah I would say join that those two and then in our

Discord we have a bunch of official links that you can click through our documentation to kind of read all about our project we are going to be doing a few sort of like live streams of the competition as it's happening next week so check us out there and then yeah we're also going to be rolling out some tutorials so you can kind of see a walkthrough of the game and how it works yeah I think everybody should go at least go to your Twitter and watch the beta that you guys have on your Twitter because it's it looks the game looks amazing it's incredible it looks I can't wait to to battle Kyle and take him down with shooting lightning bolts and all these different things that I saw in the game it looks so yeah

okay let's speed round here before we finish the show I'll go first question Brandon will go to you first what's an nft you'll never sell my AI Arena nft that was easy yeah because you've trained it so much like telling your child yeah you know what that actually is the case and and remember our previous beta like people were pissed that we had to kind of like redeploy the contracts they're like what the hell love this thing yeah there's a different emotional connection there no doubt interesting Michael what about you hey Irina do you know you know that that that's that's our baby I I was the one

that was pissed when I had to get rid of my one of the one of the people that yeah I trained up a guy named We called it beast mode and and sort of uh it just just watching it fight and I know there's gonna be someone out there that beats it there's no doubt about it but I got bragging rights for a week so that's fine okay favorite person to follow on Twitter and if you can maybe direct it towards somebody who's a leader in the AI space hmm oh yeah for me it's this guy named Christopher Ola he used to work at

Google brain and open AI he's absolutely phenomenal he also used to lead This Journal called the distill the distilled Journal which basically what they do is they create interactive sort of like machine learning articles and that's actually what got me into web development is because I I was doing machine learning stuff and I was like how do I like display this for the world to see like in a really cool way and I saw his articles and it blew my mind where you can just toggle a bunch of stuff and you could kind of see like how the different algorithms react and it just yeah I and that's part of the inspiration actually for the AI inspector where we allow you to see into the mind of the AI so definitely go check out Christopher Ola he has

extremely high quality content which has some courses on deep learning neural Nets presented in a very approachable way I know that one of Brandon's core tenants is that if you can explain something if you can't explain something basically and to a five-year-old then you don't get it yourself and I think that AI contrary to popular belief is something that can be presented and taught in a digestible way and yeah it just takes the people to to present it that way yeah it's awesome two great suggestions definitely gonna share those and encourage our followers we'll link to in the show notes to both of those gentlemen that you guys mentioned okay last question if you had a billboard that one billion

people were going to see what would you write on it you know I I'll put my motto on there my motto is my machine doesn't learn unless I do oh I like that nice yeah that's really good appreciate it yeah that's just just because I'm all about sort of continuously learning and in order to develop these better and better machine learning algorithms you yourself have to understand the technology better in order to whether it's like create a new loss function or stuff like that but yeah that's what I will put you know what I I asked chat what is it chat GPT or whatever I just asked it as well and its answer was kind of a cop-out it goes if I had a billboard

that one billion people would see I would use it to promote a message of kindness compassion and empathy I would encourage people to be understanding and respectful of one another and to strive to make the world a better place for everyone I would want to inspire people to be their best selves and work together to create a future that is bright and full of hope so very nice but yeah I want to know what it was going to say don't tell me about it I want to give you the words you have an answer after that yeah yeah GPD doesn't know how big the billboard is either this is whatever they like in 10 words do it in time gentlemen this has been an absolute pleasure thanks for coming on the show and breaking down AI in such a simple to

understand way and also thanks for working on something that is so focused on onboarding more people to Ai and web 3 and educating the masses really excited to play the game really excited to be Kyle and really excited for everyone out there to try it I really appreciate your time it's been a great it's been really enjoyable it's been a pleasure guys thank you so much for having us it's our pleasure thanks so much for your time and all that you guys do for the space like I've been following you guys for a while so appreciate it thank you for listening to web3 Academy we hope this helps you along your web 3 journey and if it does please share this episode And subscribe so you don't miss the next one nothing in this podcast was Financial advice crypto and web 3 can be risky you can

literally lose it all in fact if you invest on account of what we say you probably will lose it all so don't do that in all honesty the point of this podcast is to remove the noise of markets and price and focus on utility implementation anyway so you should not take any of this as Financial advice thank you friends and see you in the next one