In this episode, Jason Fernandes, Co-founder of AdLunam Inc. speaks with Yatharth Jain, Co-Founder and CBDO of Cluster Protocol. 🌟

Dive into how decentralized AI is revolutionizing technology with innovative solutions that enhance data privacy and accessibility. Discover the intersection of Web3 and AI and what it means for the future. 🎧

Redefining AI: Decentralized Solutions for a Smarter World

SPEAKERS

Jason Fernandes, AdLunam Inc Co-Founder

Yatharth Jain, Co-Founder and CBDO to Cluster Protocol

Hello, everyone. Just going to give it a minute for the room to fill up. If you can hear us correctly or clearly, please drop an emoji in the chat to confirm Awesome. Let's Give it a Minute.

Okay, perfect. So we have some exciting updates about upcoming events where AdLunam will be making waves. My co-founder Nadja Bester will be speaking at the hustle hour panel web 3 and media on impacts, intersection interests, hosted by thrilled labs tomorrow at 3pm CET follow AdLunam Inc to get the viewing link. I will be speaking at the wiki finance Expo in Bangkok on September 7, on tokenomics and airdrops, if you're in Thailand, this is the event to be at. Hope to see you there. Our co-founder Nadja will also be speaking at Gatherverse AI evolve summit on September 10, on a panel titled The workforce revolution AI and human collaboration and action. Follow AdLunam Inc. to get the viewing link and Altcoin Observer will be the media partner at wiki finance Expo Bangkok on the on September 7 as well, we'll be catching up with experts in the web3 world and giveaway some awesome merch. Don't miss out. So everywhere, people bring the future blockchain with you, one event at a time. So now without further ado, welcome to episode 87 of Diving into Crypto, sponsored by AdLunam Inc. I'm your host. Jason Fernandes, Co- founder of AdLunam Inc, a web3 investor, advisor, mentor, and I can't wait to get into today's discussion to help us navigate today's topic. We have a very special guest, Yatharth Jain, co-founder and CBDO of Cluster Protocol. So Cluster protocol is a decentralized infrastructure for AI that enables anyone to build, train, deploy and monetize AI models within a few clicks. Their mission is to democratize AI by making it accessible, affordable and user friendly for developers, businesses and individuals alike. Yeah, sorry, did you want to jump in and maybe introduce yourself in your company?

Yes, just doing a quick Mic check. Can you please hear me? Yes, aka Depindaddy, co founder, and CBDO to Cluster protocol. So yeah, we are Cluster protocol, an infrastructure for AI, right? So to in a nutshell, if you say what we do, right? So we are the infrastructure, right? So we are building an infrastructure for AI builders to monetize their buildings from day zero, right? So there are many individuals and projects which are working in AI, but no one is actually considering actually monetizing their buildings right, monetizing their creations, right. And ultimately we aim to templatize the space by hyper personalized AI workflows for all. So yeah, this is, in short, what we are doing at Cluster.

Okay? Awesome. You want to tell us a bit about yourself, maybe sort of what got you to excited about this problem and what’s finding it worth pursuing?

Yes, so, yeah. So basically, I, myself come from the AI and web3 background, right? So I have been in the space since a while. Have been working around a couple of PR one projects. Have been exploring things around so initially, from days, from day one of my career, I was more interested in marketing, in business development and in community building, and that is what I excel in. That is where I alongside Frati, who is my co-founder, we figured this problem out back in 2023 when everyone in the deep in space was just building another GPU marketplace. We understood the need of this. We understood the need of this to be having utility over the deep end layer, and that's where we thought of building cluster protocol, right? So we are a one stop solution for any AI builder who is looking to build solutions, and not just that, for any individual who is using AI and wants to build any AI workflow, right? So what AI workflow or templates could be made like it could be explained in simple terms, in a way that you cook food, right? There are two ways of cooking it. Either you can cook it in a traditional way, where there you would be needing everything. You would be also needing a lot of time and patience to bear to cook that off. On the other hand, you just bring the ready made foods, or the pre cooked foods, where it is just within couple of minutes you are able to do that. A popular example of that would be ramen noodles, right? So there are two ways of making it. On one hand, you can bring those cups right where you just need to pour in water, hot water, and you just keep it for two, three minutes, and it's ready. On the other hand, if you'll make it with traditional way, you would be needing a gas stove. You would need a Pan to cook, you would need water. You would need ramen as well as egg as well, right? So that's what that's what we are. That's what we are in differentiation to any other protocol in this industry. And yeah,

so how do you let, how you sort of enable people to build their own AI workflows, and how do they monetize that?

Yes, so basically, it happens in a way that consider you need to understand what the building blocks of AI are. First things first that is compute. Second thing is data, and the third thing is AI models. AI models are nothing but a Python code, right? Combining all these three of them, you can build an agent. Of what an agent do is, Agent is a specific kind of a small niche of a model that is having some intent to do. So let's say, in future, in future, anyone can build their own clone. AI right over Twitter. So what that can do is, it can replace me. So what can be done is, it can replace me. It can replace Jason as well as AdLunam as well. And there would be Ais talking to each other. And when there, there are three different AI agents which are there, and they are interacting to each other. That is the event flow, and that's the flow that we aim to achieve, and for anyone to build it. So we have compute, we have data, we have models the and there is also open infra for anyone to build their own models as well and to put them all the platform. And on top of that, if someone wants to put their agents on that is also the platform is highly modular, and the infrastructure is highly modulus that they can put in their own agents as well, and then workflows could be built. So you could consider these workflows as how you automate things in the web2 industry, or in the web2 world, where, via right, is a similar way of doing that in that direction,

okay? And in terms of how this, like, is there a token? Or how does the token, you know, integrate with your utility?

Yes. So that's a very good question, I would say. So yeah, we are having our token so it is not listed yet. It is not live yet. But yeah, we are soon planning to have our token as well. The token ticker would be CP, dollar CP, and the utility of this, all allies within the community incentives, which would be there, within the transactional value that would be there on the platform. So as I said to you, that there are three verticals to it, data, compute and models, right? Or models or agents, you could say so when someone wants to put in the data into the big data marketplace, they can stake the stake the CP tokens, and can do that. On top of that, there's consensus and validation as well on each checks so as to maintain the integrity of the network. And that is how so like, similarly, in compute as well. If someone wants to put chip in their compute, they can stake the tokens and can do that. And similarly, if they want, if someone wants to put their AI model, which is proprietary in nature and not open source in nature, they want to put it, they can put it. And then they will also earn incentives on top of that, as someone uses and that is why we are able to claim that you can monetize your AI buildings from day zero, right.

Okay, cool. And in terms of getting into crypto yourself, like, how did that happen? How did you personally, because I know you mentioned that you know your background is in this space, so I'm curious how you got in here.

Yeah, so basically, I entered in this crypto market, I'm here since 2019, 2020 right? So, but the very first thing, or the very first interaction of me with crypto, maybe when I might have heard crypto, is back in 2016 right? So back then, when I was a kid, right? I like, there were a couple of undergrad students who were working around who were doing something around solidity. I never knew what solidity was at that time, I was learning some basic HTML code, right? And at that moment, I was sitting in my computer lab, and that guy sat near to me, and he was coding solidity, right? So I asked him, What is it all about? What language are you coding in? And he said, solidity. I was like, What is solidity? So then he explained me the concept of Bitcoin. And I was like, I told him that, no, it's kind of a, it was looking like a kind of a team that resonates to a multi level marketing scheme or something like that, and he should stay off for it. But then, after years passed, and back in 2019 I myself got interested into this technology, and then I started reading more about it. And in 2020 2021, Bull Run, I got into the crypto space professionally.

understood. And so the friend that you suggested he get away from crypto. Have you ever get back in touch with him and say, Hey, by the way, maybe crypto is, is actually something what we're doing?

Yeah? So, yeah, I had like, I might like, so basically, he's now very far from me. He's settled in the US, right? So he's working for Bigfoot now he's a consultant there, so unfortunately, we don't meet that often, but yeah, we do remember those stories of when things happen and when we get together. So we every time remember that story of how I said him that it is something which now I'm supporting to the core

understood. And can you tell us more about sort of how cluster protocol began? And you know, you mentioned you had a co-founder, what inspired the creation of cluster protocol, and how does this sort of differ for more traditional AI infrastructure?

Yes. So, as I already told you, that I was working for one of the DePIN projects out in 2023 alongside that, Pratik owned a local rig in Australia, right? So Pratik is based out of Australia. I am based out of India. So Pratik owns a local rig which used to cater the local demands of Australia and nearby region, like the place that he lived in, and in the nearby region, and that's where like and I know I knew Pratik from a personal connect of mine since a while. Then we were exploring things around. So this idea got into my mind of building this utility layer over deep in and initially, what I thought was of, Let's have someone who has knowledge at least of compute, right? Because things can be figured out at a later stage when we move ahead, but I would be needing compute any day, right? And that is where we, I and Pratik, thought of this, thought of on this idea of building a utility layer over deep in initially the idea was to have to give consumers a platform for a platform for the ease, right? The platform should be modular enough. So that is why. Then, after moving, after doing the basic R and D, then we thought of, why not to have our own infrastructure? Our own infrastructure will be leveraged, because if, in case we are dependent on anyone's else's infrastructure, or you can say so, basically, we are using arbitrance orbit for having our own chain, right? So what does that mean and why it is useful? It is because of the fact that if in case we use anyone's else else's infrastructure, or maybe a chain that there are a lot of micro transactions in AI space, right, which would be happening when we would be live, right? So at that moment, the gas fee, and there are many other factors to it, which will bottleneck the whole network. And that is why, without having our own infra, and that's how we ideate a cluster, and we started building things on.

So, you know, you hear about these AI platforms, and how many billions of dollars it takes to train them. Is that something that you're concerned about? How does something like what you're doing, how is it able to compete with these big, massive budgets?

So basically, you need to understand that these are not platforms who are training. There are LLM models, right? There's large language model, which requires hours and hours of training, if you have heard of phi pi, right? This Microsoft's newest AI model generation. That is the newest AI model that has been generated by the by Microsoft, right so it was trained continuously for 23 days on 500 H, hundreds. Right on 500 H, hundreds, that model was trained. Now what we are doing is we are not having these models as our own whole. We are working over it. We are trying to get some models which are having core utility for the crypto folks. But apart from that, we can, you can easily leverage any of the other open sourced models. And those are more than enough. And according to my opinion, what I think is a combination of open source models and a small like niche models. Which are there niche models that to train over via edge computing, or via these modern technologies where you don't need this much amount, this much amount of compute, rather, you need a significant like a significant hardware engineer, and some basic hardware which can run 24/7, over it, that would be a more optimal and feasible approach to it. And, yeah, that is how we are also having so basically, there would be some high quality open source model which would be there, which people could utilize to build agents, and then ultimately, Agent tech workflows.

Okay, so you don't really think of them as. Competitors, really, you more think of them as, sort of as being able to plug into your system.

Yes.

Okay, interesting. And so, could you tell us, sort of how you approach to AI democratization alliance with the broader web3 movement? I mean, you know, we talked a lot about being able to compensate people with tokens. How does this sort of web combination of web three and AI transform the way AI models are developed, trained and deployed?

Yes, so if you see web3 industry or the actual decentralization of AI. There are efforts which are being taken into consideration. But there these are the baby steps that any of us are taking create any of the project that is doing anything in this industry, they are taking baby steps out, if you'll see. So, as I already told you, you divide it. You divide AI into three sectors. First one is data. Right with data? What you can do is you can decentralize it by saying that, okay, you can put the proofs of those data sets which are there, which are stored on your decentralized nodes on a blockchain. This is what you can achieve over data in Compute, as you already know, deep in itself, is a very big sector where you decentralize the compute, and you put those machines on a blockchain, and the compute is provided to you in a decentralized nature, the proofs of that is stored on a decentralized nature. Moving ahead, another use case, another good use case of blockchain in a compute would be providing the proof of compute that is verifying the compute and then, and then using that compute to provide it to the user. So there's a middleware which is there now to eliminate the discrepancies in the middle where there are advanced security technologies like fully homomorphic encryption and others where you don't even need to hamper your data upon as well. And then with the AI models, it's the training part, right, the training and the inference part. So there are the two parts which could be decentralized, and we focus more on inference as AI training is the decentralized decentralizing AI training is something which I think would be achieved, but it will take a lot of time, anywhere around 5 to 10 years. So what are feasibly? What is feasible at the moment is what we at plaster believes in venting.

So now, you know, we talk about a lot about AI models and AI development. So how do you think it can be made more accessible, affordable and user friendly to not just large enterprises, but also, you know, individual developers and small businesses? How can they leverage AI to sort of achieve goals?

I would suggest this could be done by enhancing the user experience as well as by providing more convenience to the users, right? So if, in case, you are able to replicate what web two is already providing to the users speed in terms of lending compute. Compute could be lended off in the web to theworld via AWS, via Google, via Azure, that is Microsoft, Azure. But if, in case, you are able to match that standards pipe by putting in blockchain as the back end technology and all these cryptographic and security methods into it, then I think that could be achieved. And mass adoption of AI and web3 could happen.

Okay, and in terms of, you know, from a developer perspective, how can certain developers more easily access AI and maybe to use in their in their own particular work, as opposed to, sort of what I mean when I when I'm referring to sort of more individual developers, rather than, you know, larger enterprises,

I would say for developers, there is an opportunity. So basically, you could consider this as a very niched industry, right? It is a very niche industry where things are happening very fast. The adoption is also happening. And as well as the development of tools and development of services, development of technology is also happening at a great pace. So to have collaborative approach, I think if, in case, they have a more collaborative approach. And then there are multiple solutions and there as well, it comes, it all comes to understanding the pain points of developers and providing a more convenient solution to them. Now, convenience could be anything. Convenience could be as small as providing a better. UX for something for, say, data set buying, or for, say, data handling, for say, anything for, say, fetching some data from some pipeline to data processor, dot Tech Data coprocessors to get it flush, to make data sets, anything where there is more value and convenience will win the race.

Got it. I don't know if you're familiar with the case of engineer.ai where a bunch of developers, and this happened in India, a bunch of developers said that they developed an AI model that would write code. And in actuality, what was happening was they hired a bunch of people in a room somewhere, and those guys are furiously writing code. I'm just curious if you, what do you think about something like that? If you think that that's like the danger?

Um, yeah, I would say that it was like, it all depends on see, as I already stated, that right? The adoption is taking place, and this is all part of that, right? This is all part of the adoption, that adoption curve that is happening around right? There would be multiple things which would be there and thus will be done, right? So that is why we are looking around at such things, right, where things are happening around, and then there would be specified use cases as well, which would play a crucial role. So, yeah, I think that's the start.

And so, you know, we talk a lot about tokenization, and I'm curious how you see tokenization as an incentive mechanism to promote collaboration and resource sharing when it comes to web3 AI platforms.

Yeah, tokenization plays an important role, because in this industry, every so if you'll see how AI grew. So there were these centralized providers, be it Google, be it open AI, be it Elon Musk or x or any platform, and what they have in power is data, right? And if, in case, we are able to tokenize the data and give users the access to it, and give users to a chance to monetize it. That way you would be able to promote more collaboration. And I think web three based AI platforms are the right way of doing it, because blockchain itself allows you to tokenize those assets. Blockchain itself will because it is decentralized in nature. It's transparent in nature. It is open, collaborative in nature. And that is the reason that I think that web three based AI platforms would promote collaboration and social sharing,

Guys so and, you know, we talk about, you know, data sharing, one of the things about what blockchain is supposed to be, you know, security, transparency, and then also privacy, right? So, how do you think those things, those are going to play? What factor do you think that would play in sort of the when it comes to AI development? I mean, would that is there something particularly useful about blockchain and intersection of that and an AI that will allow for, you know, increase and better, more secure data sharing or privacy preserving techniques.

I think that, yes, these techniques play a crucial role in that, because, in sense, what happened in this, in these centralized servers, or maybe anywhere that is centralized in nature, there are less possibilities of error like errors could happen. But if, in case, we are utilizing blockchain, the very advantage of utilizing this as a tick is that you have modularity in terms of selecting the security layer as well. You can use a basic SHA 256, hash as well to make your data cryptic. And you can even use the most complex mechanisms, like tees, like zk. And there are many other which would be, which could be utilized for these security, these privacy, securing AI techniques. So you could say these security methods, but the very three popular of them are TV, is fhe and vkml

guide. So what would you say in terms of real world applications where decentralized AI could make a significant impact. So especially like sectors such as healthcare, finance, supply chain management, what are some sort of real world applications where you think decentralized AI could make a dent?

I think in healthcare, it's very important to understand that Uh, the bias should be eliminated if you have heard of a story, right, if you have there, there was an incident in the US where there was a kid, right? There was a kid and his mother was having some undetected, undiagnosed, uh, disease since 13,14, years, right? And what that kid did was nothing, but just gave all the reports and all the symptoms that were there to achieve ChatGPT 3.5 and it was not the case that the mother was also not going to doctor or something. She used to visit the doctors frequently, but still she was undiagnosed, and chatGPT did that within two minutes, right? So that's the beauty of health. That's the beauty of AI in healthcare, and I think it will increase. There are a lot of my friends who are walking around with like with more diseases like cancer and many other diseases which are working around in building AI models and AI agents around that coming to finance, yes, it is like a first or firstly. When, when we say AI and finance, there's the predictive nature of AI comes into picture where there could be something it can predict, right? Something it can predict, maybe the stock markets, maybe the crypto markets, maybe something more than that as well. And in finance like cluster protocol is also taking a leap and building an AI suit around everything defi. So basically, we will be launching this soon. So it's an AI suit for your day to day defi usage. It is made for degens, and it is made for anyone who is into crypto. So you could do everything you can do sense, web on chain, off chain, bridge you could do when you can so there are many advanced AI features, which includes advanced portfolio recommendation tokens, AI power token sniping, and many more features are coming onto that AI suit. So which we, which we will be announcing really soon. So stick tight and like, keep an eye on our socials as well. So coming to coming next to supply chain management, yes, supply chain management could even be made better via AI in terms of saying that, okay, let's consider there. There is like here, also the predictive nature of AI would play a crucial role where you have a supply chain, and based on so it could be used in food category, where it AI could predict that, okay, the freshness of the fruits food. Or maybe there would be multiple ways on how supply chain management would be improved by

cool and tell us more about like, where what stage your project is in. I mean, do you have users? Are you still beta testing, or in what sort of development phase are you guys at?

Yes, that's a very interesting question, and a good one. So basically, what we did was we did our test net back in a couple of months. It was a data annotation test net that had almost 50,000 plus users who onboarded and contributed there. So they validated and annotated data daily. On top of it, that test net lasted for almost one month. Now we are moving towards launching these AI agents. And after this, we are planning our token as well. And then we will be moving towards having the platforms beta live onto mainnet.

And when you apply to launch, launch a token.

So basically, we are looking to launch the token tentatively in quarter four of 24

cool and, as I've been mostly around, sort of some development priorities, and you guys need to put in place just for your peer to integrate the token into your system.

Yeah. So, as I've already told you, about the token utility, it resonates to that only where there's a robust mechanism of operators, well data, than much more, which is there all the things are available on a white paper, you guys can have a look over it so and also, we would be soon starting our own builders program as well. So if there are any AI builders who are listening to this, then you guys could check our website or our Twitter handles to explore more about that AI program .

understood and what challenges do you face in developing a decentralized AI ecosystem, and how have you sort of managed to overcome those challenges,

I would say, the quality data and Like data is something that is really difficult to find, because either ways, you will find, like all the data that is having a digital footprint is already with these big corps, right? And to have quality data that is there, and secondly, that is private data. So these are the two things that I think is the hardest to find. But luckily, there are some ways we have figured something out over those lines as well, coming to the technical development, yeah, there are a lot of challenges when you say, when you are actually decentralizing the approach of say. So you need to understand that decentralizing the compute is not an easy task, right? So that's where I think there were some initial challenges that we faced. But apart from that, yeah, the development have been smooth, and the quality data and the private data, the two parts where we think that there were major challenges that we faced, right?

And how could, how can democratization of AI lead to more inclusive innovation, and what are the sort of steps necessary to achieve that as a goal?

Yes, so basically, AI will lead to inclusion and democratization will help users to make things like the steps are there? There's no playbook or something or there. There's no predefined steps. Right? Are necessary for achieving the goal? What I think is, if you want to democratize the AI, then make something that makes users life easy, right? If, in case you are able to make that happen, if, in case we are able to achieve that, where there is something which is making people's life easy, then it's a win, win for you. You could understand in a way that and you need to understand the audience size as well initially. Right now, if you'll see the layman, even the layman who is using AI is somewhat a builder, or you could say somewhat techie in nature, at least a little techie in nature. I don't know. Like there would be many folks as well who are not that techy, but at least the one who is using chat GPT, or has figured out to use how to use chat GPT, is somewhat interested in tech right now, right so right now to say that, like AI, development is in its nascent stage, it needs To go to those folks as well, those who are living in remote areas and stuff like that. And AI doesn't always mean chat bots, right? AI, actually, AI, is something you could consider, where things are happening on its own, and that is AGI or maybe intent based AIS or maybe AI workflows, right? There are multiple things to consider it in that direction. And that's where if, in case, you are able to provide utility with ease, with convenience, and that is the thing in which you could say that democratization of AI can lead to more inclusive innovation. So curious. I Guess we Lost Jason.

I had some technical issues, but yeah, just to jump right back, back where we were, to pick up where we left off. So I mean, I was just. Curious. You know, I know this is a this is an area, an industry, that you follow very, very, very closely. So I'm curious, if you have to, like sort of project out, let's say, five or 10 years. What? What are some of the things that you think we're likely to see from both the industry in general, and maybe your problem statement specifically? Where do you think both of those things are likely to lead up? Let's say five years over 10 years, because we're in an exponential, a growth, exponential growth industry. So I'm just curious what you where you think it'll play out?

I would say in five to 10 years. I don't think that AGI would be achieved. So if, in case, someone wants to understand what AGI actually is, that is automated, generative intelligence, basically, AI is nothing, but you could consider that that an LLN model or AI thinking of its own. So it's somewhat similar to saying that, okay, you have told AI, or you have told someone, which is confuse frig in nature, right? Which is computer nature, to cook food for you, and that is it you, you have said that cook ramen for me. That is it that AI will figure out, okay, okay, so someone has said that, okay, cook ramen for me. Then I'll go to the kitchen. I'll figure out the utensils, I'll figure out everything else. If, in case, I don't have ramen, then I'll go to the grocery store to buy some ramen off. I'll come back, and then I'll cook that for you. So that wouldn't be achieved, I don't think so. Yeah, web AGI could be something that could be achieved, but that too, with limited agentic workflows, is something that people can look for incoming 10 years, where you just need to understand how to give agents the instructions, and after that is being done, the agent is doing good for you, and your workload is reduced a lot. But then you also need to understand the fact that you need to understand how to do, how to like, how to put those instructions over and intense, over to the agents, and with limited capacity, either ways, the agents are doing pretty well. The agentic market is growing. And there are multiple projects which are working around AI agents, AI agent tech workflows and many more. Some are decentralized. Some are some are centralized in nature as well, which are doing very good. So I think, yeah, like, these are my thought processes and thought viewpoints over this. And yeah, like, either ways, I think that these centralized authority authorities and these AI by centralized authorities might grow, but yeah, some of the projects which are there in the decentralized space can outshine as well.

so you think, what do you think about, you know, we see robotics as improving on almost, you know, exponential pays as well. What do you think the possibilities are somebody putting, you know, an AI interaction, UI, sort of into something that's like, humanoid as a robot, and how long, how far away do you think we are from something that's like, let's say, humanoid, but is able to talk and understand based off of something like ChatGPT?

I think that could be achieved. I think people are working over it, but to have actual AGI where there is an actual neural network, that is, that comes into the part of decentralized science, right? And, by the way, yeah, like after this space, within next 15, 20 minutes, I would be going in a decentralized science space itself, right? So there are a couple of my friends who are in their decentralized science and append hours there. There are a couple of others as well which are working really great there, and that's the field of their expertise, I guess. But what I have understood from them is that Desai as well is on the pace where it is growing around, but it's not at that pace you could say where there could be actual humanoids, which would be made over blockchains, but yeah, in the centralized world, you can imitate a human. You can clone a human. You can clone humans actions, right? So consider an AI which is tracking all of your needs or all of what you are doing. Let's say that there's an AI that tracks this sort of space, that tracks all of my mails, that tracks what I talk, that tracks on all of my higher conversations. And then a common conversational AI could be built, which can. And talk like me. And on top of that, you can even have multi utility or multi model stuff, where it's not just limited to chat bot, but maybe voice, but maybe that that would be as a that would be considered as a base player, do more intense and action that could be feasible. And I think that could be done within next five to 10 years as well.

That's pretty quick. That's pretty quick. What advice would you give you know, young developers that are looking at getting into the field? Where do you think a good place to start would be for somebody who's just sort of curious about AI and the opportunities in the AI industry?

I would say the industry is really rewarding, if you might have seen when the ChatGPT was on its road, right. Still, right now as well, the ChatGPT is on its road, but when it was getting popularized, right, there was a trend of people opting in from prom for prompt engineering. Prompt engineering is a very basic rule, which is nothing fancy. You just need to understand English, and it's just the case of trial and error, right? You need to understand English, and it's just trial and error that you need to do. And based on that, you earn money out of it, right? And the salaries at that time were exorbitant, like people were making 150 to 200k per annum just by doing pump engineering. But then, when people understood that, okay, one level granular to that could also be built then, then that type or that by got away. And I think, yeah, like, one advice I would love to give to the developers is keep building and also look for convenience users. Convenience and users ease. Always make sure that you are building something that is for the community, and the community can benefit it up and always think of revenue first approach. Instead of having an approach where, okay, you could, you could be taking your token off, and then you can do multiple other stuff around that, right? I think that if, in case, developers make sure to have a users first approach, as well as the developers are making sure that there is a proper revenue stream defined to it, then I think they will succeed anyway in life, maybe earlier, maybe later.

Got it so what about the large cost of large language models. You know, seven, 50 million per day for ChatGPT Lawrence asked, How is this sustainable? Consider the market bubble for AI investments. Is it sustainable?

So you need to understand how why it is right. You need to understand why it is because there are two things to it, as I already told you, first one is training, and another one is inference. So what is happening with ChatGPT and all these models which are there, which are public in nature, is that there is inference, causes created to it. There are server causation to it. There are GPUs which are running for inference that they need to cater rest on top of that they are. They are regularly doing R and D, right? So charging stop at 3.5 they built four. They built four. Mini they built they built four. Oh, now there are rumors that they are building five. So on one hand, they are. They need to cater all the inference that is happening around over their existing models. On top of that, they are training on other models as well. So this is the major cost of training and inference that caused them this exorbitant amounts of burn of capital

understood. And you think, do you think the current trend towards investment in AI will kill you. And how do you think? How that do you think that that was sustained, because it's just so many, so much capital moving into AI these days?

I think, yeah, like, AI is VCs favorites, right? AI is VC favorites these days. So be it web two or web three, I think there would be more inflow of funds from VCs in AI sector in general, and I think that in the coming months, right, or in the coming years, where the economy, so it all depends on economy, if, if, in case, there's a economy Which where people are happy parking their funds in their banks right then you would see a more of web two. Ai, ai, economic. Web two. Ai, funding happening more often, and if, in case, there is a distrust in the current existing financial system, and people look forward for, say, a. Towards Bitcoin or these other alternative illusions of finance, then I think that web three AI could get, like could get some good some more inflows of fun. But in general, I think yes, AI will be getting inflows from all around the space, and it is not just a hype or a bubble that that is there for a couple of years.

Awesome. So now you know we're getting towards the end of this episode, so we're going to sort of invite questions from the audience. If you have any questions, just hit that request to speak button and we'll bring you up to share your thoughts. Feel free to use the emojis to let us know when you're ready. In the meantime, Yatharth, why don't you tell us a bit about, you know, your personal philosophy and what keeps you going through the bear markets and through, you know, the times when funding is scarce,

yeah, so it's so basically, what we thought of initially was, so my philosophy is very simple. I like to build, right? I like so be it the bear market of the bull market. My philosophies are remains the same, where I trust the system. First of all, I trust the decentralized systems, be it BTC at $5,000 or $50,000 or at 500k I trust the system. Alongside that, what I believe is in my work, the work that I'm providing for the society, for the community, or the product that I'm building for the folks is actually creating value in the ecosystem, right? And that's where we are more dedicated towards it. Yes, we are. We are also not bootstrapped at the moment, like we have VC fundings as well. But the matter of the fact is that that doesn't define or that doesn't affect the work we do. We do the work every day the same the same amount of time, and we would same dedication as we were doing before that as well. And just that the funds were utilized even for more R and D, and there are other stuff as well where the funds are used. But I we need to understand the core philosophy for any founder is to make sure that the team is stable, the team is working around and the leader. What's the role of the leader is similar to what the role of founder is right? You need to just put in the right team at the right place, and you need to get the right work at the right thing. That's my philosophy.

So really good. This is really good philosophy. And just our curiosity, you know where towards the end of this, this program, so do you have any sort of parting thoughts you'd like to share with the audience? Maybe, what's coming up next for cluster protocols, specifically, what some of the things that you're excited about, just sort of final thoughts,

yeah. So stick tight. As I already told you, there are a couple of AI agent tech workflows that we are working upon and as we are working upon so we will be putting them live really soon on beta, so you guys can test it off. Also, alongside that, yeah, we are planning a couple of other giveaway campaigns as well, so our socials to have a look over it. So yeah, a couple of immediate updates that are there for the community. Apart from that, yeah, it was a pleasure. It was a lovely time with you, Jason and with the team AdLunam. And thank you the team. Thank you to the team, AdLunam, twitter space, and, yeah, something that I really enjoyed, and the questions that you asked Jason were awesome, and it was great to have there. And thank you guys for having me in this panel.

No, I mean, thank you so much. It's been such an incredibly insightful session. I'm sure you know our audience learn a lot about AI, sort of some of the prospects that the intersection of AI and blockchain have. Yeah, so I want to extend a big thank you to your tar Jain here for sharing such valuable knowledge with us today. A huge thanks to all of you tuned in, ask questions, engage with us. We excited to keep bringing you more content to stay connected with us. Thanks again, everyone. Take care. We'll see you at the next session. Thanks. Yatharth, have a great day, everybody.

Thank you guys. Signing off. Bye, bye.