Space Summary
The Twitter Space BET on Drift (w/ Solana & Multicoin) hosted by DriftProtocol. Immerse yourself in the world of on-chain trading with BET on Drift, where Solana and Multicoin strategies converge to offer a dynamic ecosystem for traders and investors. Explore the fast-paced environment of Solana, discover the benefits of Multicoin investment strategies, and delve into the evolving landscape of decentralized finance (DeFi). Engage in prediction activities, stay informed about market dynamics, and optimize your crypto portfolio with diverse strategies and platforms. Stay ahead in the trading game by learning from the insights shared in this illuminating Twitter space.
For more spaces, visit the Trading page.
Questions
Q: What makes Solana a preferred platform for on-chain trading?
A: Solana's fast transaction speeds and low fees make it attractive for traders seeking efficiency and cost-effectiveness.
Q: How does BET on Drift differentiate itself in the Solana ecosystem?
A: BET on Drift stands out by offering a platform for trading, earning rewards, and engaging in prediction activities within the Solana network.
Q: Why are Multicoin strategies beneficial for crypto investors?
A: Multicoin strategies enable investors to diversify their holdings across various crypto projects and assets, reducing exposure to individual risks.
Q: What role does DeFi play in the evolution of crypto trading?
A: DeFi is revolutionizing the financial landscape by providing decentralized platforms for trading, lending, and earning passive income.
Q: How can prediction markets enhance user participation in the crypto space?
A: Prediction markets offer users a unique opportunity to speculate on various outcomes and contribute to the decentralized governance of blockchain networks.
Q: Why is it important to stay informed about on-chain trading dynamics?
A: Understanding on-chain trading dynamics helps traders optimize their strategies, manage risks effectively, and capitalize on market opportunities.
Q: What insights can participation in Solana-based projects offer to investors?
A: Engaging with Solana projects provides investors with firsthand experience of innovative blockchain technologies, trends, and investment prospects.
Q: How does engagement in prediction activities benefit individuals in DeFi?
A: Engaging in prediction activities fosters a deeper understanding of market behavior, risk assessment, and decentralized financial mechanisms within the DeFi sector.
Q: Why is it crucial to monitor developments in the DeFi sector?
A: Tracking DeFi developments is essential for identifying emerging trends, new investment opportunities, and staying ahead in the rapidly evolving crypto landscape.
Q: How can diverse strategies and platforms enhance a crypto portfolio?
A: Utilizing diverse strategies and platforms like Solana and Multicoin can help investors create a balanced portfolio, mitigate risks, and maximize returns in the crypto market.
Highlights
Time: 00:15:42
Solana's Scalability and Speed Exploring how Solana's high throughput and low latency contribute to efficient on-chain trading experiences.
Time: 00:25:18
BET on Drift Platform Features Discovering the unique features of BET on Drift for trading, earning rewards, and engaging in prediction markets on Solana.
Time: 00:35:59
Multicoin Investment Strategies Understanding the benefits of Multicoin strategies in diversifying crypto investments and reducing risk exposure.
Time: 00:45:27
DeFi's Impact on Financial Services Examining how decentralized finance is reshaping traditional financial services through innovative trading and lending platforms.
Time: 00:55:14
Predictions in Crypto Markets Exploring the role of prediction markets in facilitating user engagement and governance in the cryptocurrency space.
Time: 01:05:37
Navigating On-Chain Trading Dynamics Insights into effectively navigating the complexities of on-chain trading for optimal trading strategies and risk management.
Time: 01:15:22
Exploring Solana Ecosystem Projects Diving into the latest projects and developments within the Solana ecosystem and their potential impact on the crypto market.
Time: 01:25:49
Predictive Insights in DeFi Leveraging prediction activities to gain valuable insights into market behavior, risk assessment, and emerging trends in decentralized finance.
Time: 01:35:16
DeFi Development Trends Tracking the latest trends and innovations in the DeFi sector to capitalize on new opportunities and stay competitive in the crypto space.
Time: 01:45:03
Optimizing Crypto Portfolio with Diverse Strategies Strategies for optimizing crypto portfolios by utilizing diverse platforms like Solana and Multicoin to achieve balance and maximize returns.
Key Takeaways
- Solana is a prominent platform for on-chain trading with fast transactions and low fees.
- BET on Drift offers opportunities for trading, earning, and making predictions in the Solana ecosystem.
- Multicoin strategies provide diversified investment options across different projects and assets.
- DeFi continues to evolve, offering new possibilities for decentralized trading and financial services.
- Exploring prediction markets can lead to innovative ways of engaging with blockchain technologies.
- Understanding the dynamics of on-chain trading is essential for navigating the crypto landscape efficiently.
- Participating in Solana-based projects can provide insights into emerging trends and potential investment opportunities.
- Engagement in prediction activities can enhance one's understanding of market dynamics and decentralized finance.
- Keeping abreast of developments in DeFi is crucial for maximizing opportunities and mitigating risks in the crypto space.
- Exploring diverse strategies and platforms like Solana and Multicoin can help in building a well-rounded crypto portfolio.
Behind the Mic
Introduction and Welcome
More than power. Never. The races on girl and ever hour after our work is ever over. Our work is never over. Never over. Just Gm. GM. Everybody, welcome to a banger show with Kip protocol web free base layer for AI. Kip Token, winner of the 2023 Chainlink hackathon backed by Animoca Ventures. We've got a great show for you today. And the title is aggressive, but I sort of feel it. Admit it, you have no idea what decentralized AI is. And look, if you are admitting this, and I will admit I definitely could have more of an idea of what decentralized AI is, then this is going to be the show for you. We're going to get the experts takes on this. We're going to really try and dive in to what decentralized AI is, why it's important. And yeah, get the expert taste from our incredible guest speakers.
Introducing Guest Speakers
Today we have Wham. AI decentralized gaming, 3.5 million users powered by AI, governed by Wam Token. We have game by Prism, AI driven game engine, dual token. We have Jen today, co founder, chief AI officer at Kip protocol, humble functionary. I love that bio. We also have Julian with us today from KipP, the CEO and co founder of Kip Protocol, causality. Oh my God, I can't read today. Causality devotee robust octopuse from the future. And we have my shell, the first AI consumer layer, build, share and own AI apps. Guys, I'm really excited for this. And look, I know we have a number of gigabrains with us today, and I know also we have me on stage, so let me provide that balance because we're going to need it today, I think, to really get to the, you know, really get into the weeds a little bit on this one.
Defining Decentralized AI
But also I think it's very important that we, you know, define this conversation before we get too far into the weeds here. So that being said, I would love somebody to help me break this down. How should we be defining decentralized AI in simple terms and what sets it apart from centralized AI systems? And look, Jen, Julian, I don't know which of you would like to maybe start this ball rolling, but I think it definitely needs to come to one of you two first, if that is okay with you. I'll leave it up to Jen and Julian. Where to start? Julian, I saw you come off Mike. Yes, happy to start off. Thank you everyone, for joining our space. Good to talk to you again, Jack. Thank you. Our lovely panelists game, GPT, Michelle, and yeah, I think what's very interesting is that a lot of the AI projects in web three now, I dare reckon all entered the space with kind of different genesis stories, different superhero origin stories, right?
Kip's Genesis Story
But from our perspective, Kip, or the team that made Kipp, started as an AI consulting company back before AI was called. It was just a whole mess, or mess of big data analytics, natural language processing, machine learning, even before the transformer based, I would say this current wave of disruption came from best result of transformers. So we started in consulting. We were solving a lot of data crunching problems for organizations. And what happened was we made essentially a web three analysis AI for Animoca. So Animoca was a client before they were an investor, so they asked us to make one. We did. It worked. And our structure of that was, our structure of that particular product was based on this framework called retrieval augmented generation. But long story short, it just required external data sets.
emergence of AI and Ownership Issues
And we very quickly realized that if we structure an AI product using reg frameworks and with external data sets, some of the external data sets could be very valuable. We started doing this in 2022, early 2023, and that also coincided with the time when chat GBT really, I think, took the world by storm. And also their misbehavior, shall we say, in terms of the stealing of data, became very public. So a lot of it started focusing on ownership, who owns what in AI? And what is very clear was that OpenAI and their competitors were trying to own as much of it as possible. And it kind of first started with taking a lot of data, stealing data that didn't belong to them, to train their models in which they are probably going for the possession, this nine tenths of ownership type of approach.
The Need for Decentralization
So they wanted to own all the data so that the models that is a result of that data being used to train the models is 100% owned by them. So we thought that this was a horrible future. So when we started thinking about decentralizing it was mainly from the perspective of decentralization through blockchain is the only way to get ownership, true ownership, through digital ownership, over anything. So we started to progressively think about which parts of the AI ecosystem can be or should be decentralized, but the purpose of it is so that the people creating it can actually get ownership over it. So our definition will be decentralized AI, or a decentralized AI ecosystem, would be an ecosystem where the creators of AI assets get to actually own their creation.
Understanding AI Scraping Concerns
I think I understand this, and I'm going to ask a follow up, Julian, for myself, honestly, more than anybody. But I, in my head, when I saw AI, and I don't know if we refer to it as scraping, that's definitely a term that I've had with the data side where they literally are just coming for information and they are not rewarding anybody who actually initially does the research and puts that information together. And think of it this way. So right now I can ask chat GPT, for example, give me a list of movies that I want to watch and give me all the data from these review services. But by doing that, by going to chat GPT rather than these actual review services, which literally, basically fund themselves through my actions of clicking and going through this, that's how they monetize, that's how they do what they do.
Consequences of AI Information Gathering
Because nothing in life is free to get this information and to actually cultivate that information and store it in a way that gives value to the people looking for that information concisely, that takes time, it takes effort, and it still, right now, takes a human being to do this. If AI just comes in and just scrapes all of that information without rewarding the people who are actually building these tools without reward. And this is very specific and I know gets more broader than this, but just so I can wrap my head around it with a really tangible example, there is a massive likelihood that if chat GPT continues to do that without any sort of reward system for where it's getting the information or where the initial work comes from, that worker, that person who's providing that value, getting nothing in return for AI scraping this information, those people where it gets the information from in the long term are going to disappear because it's fundamentally they're not going to get the reward from the work that they do?
The Importance of Rewards in Decentralized AI
So is it right that I'm thinking about decentralized in a fashion where if I was a reviewer, if I was somebody who was putting together these articles, and obviously then we go a little bit further into like actual products development, etcetera, decentralized AI is seeking that information, but providing the person who initially cultivates all that information with some sort of reward. I think you summarized the entire problem and also a lot of the solution perfectly. Wow. Okay. I was nervous as I was going through that, julian, so thank you so much. No, no, I think you do have an idea what decentralized AI is. So I think you can drop off the call now, Jack. Yeah, just for the rest of us who don't understand.
Exploration of Decentralized AI Concepts
Okay, cheers. Jen, can you come in on this one? Like, how did you find that take could you elaborate on that anymore? And because I guess the piece of information that I'm still missing here, julian, but I really like, I am very proud of myself right now. The piece of information I'm missing is when it gets past information, you know, like when it becomes more tangible in terms of a product and how AI is impacting that side and why decentralized AI is needed beyond just cultivation of information. Could you, could maybe elaborate on that, either of you? Yeah, I can probably go into a bit and then Julian can cover whatever I've missed, which, yeah, will probably be a lot.
Decentralized AI Definitions and Different Perspectives
But yeah, I guess there are, you know, the reason we chose this space title is because there are different takes on what a decentralized AI is like. Is it, is it when you have multiple computers doing the inference, is it a mixture of experts model where different people own different weights, perhaps? Is it, is it something like cobbled AI where you have multiple people donating inferences? It civet AI where people are making their models available via a platform. Yeah, everyone kind of has their own take on it. And I would say what we're trying to kind of work our way around right now is how to turn all these ideas into products. So Tao, you know, the ogs in the space, they have some subnets that are doing inference right now, they're providing it.
Current Challenges in Decentralized AI
The problem is how to get that to the customers and get it monetized via some other means than line go up. So right now, I think the subnet one is offering inference via cobbled horde, which I mentioned just now. And that is they're doing it for free. They're doing it as sort of a donation, which is great, but it's not necessarily a long term prospect, especially when the line starts going down. So yeah, that's what we're all kind of working around right now, I guess. Yeah, that makes sense. Go on Julian, please. I'll just add last bit to that.
Components of Decentralized AI Products
It's an easy way of thinking about it is that in order for a decentralized, if we think about what the decentralized AI product is actually it is made up of at least two or three components, depending on whether it's reg involved. And these three components would be a model providing the inference, an application providing the user experience, and payment gateways, all that and a dataset. If you are using a rack framework, then you need an external data set. These three components make up a complete AI product. And this distinction between these three components is often missed, especially in the context of big tech because in big tech like OpenAI, for example, they own everything.
Big Tech Control in AI
It's their model interacting with their app and the data sets which they are convincing. They're trying to convince everyone to upload into their systems and therefore they own it. So they own all three, but there are three very distinct components. And these three distinct components, it becomes very clear, especially to anyone who's been doing commercial work in AI, because often the people who are working on the models are very different from the people who are working on the applications. And the distinction is also that it's also important, because most of us ordinary people, Jen is not an ordinary person, but I'm an ordinary person, so I can't interact with the model directly.
Connecting Different Groups in Decentralized AI
I need an app to give instructions to the model and the people who are working on a model who are uploading a hugging face. They generally come from a researcher background, very highly cerebral, highly research and technical focused background, whereas the people who are making the applications that is capable of serving users, often from a product background or a marketing background. Even so, in what decentralized AI needs to solve, I think the point that Jen was alluding to is that how do different groups of people who are each good at their own thing, like training models, making models, or making a front end, like a chatbot that is tuned to respond in a certain way, or people who have valuable datasets, how do these three different groups, or three different industries of people get together in web three, are able to link up their products and then still make money?
Economic Incentives and Challenges
Because one of the points that you mentioned, Jack, which I really agreed with just now, was that if you remove the commercial incentives for doing any of this stuff, then they would just stop being done. So I think I'll wrap up this little bit on Kip's part and invite our panelists to share their thoughts. But it's Kip just solves very basic bread and butter issues of make sure these decentralized components, when they're deploying the web three data can flow between them. And then after that, make sure that money can flow in a trustless fashion between three different groups of people, three different companies, three different individuals, or three different industries of people doing different important pieces of work in AI, it will still be able to operate in web three or still be able to get paid.
Conclusions on Decentralized AI
That's all kip does. We think it's very important, but that's what we're focused on. That's all we are focused on. Thanks. Truly fascinating, because it is definitely something that has gone through my mind, utilizing GPT as a creator, definitely. Sometimes you just need to source information quickly, and the best way to do that right now is to utilize these AI tools. I do wonder what that means in respect of the future, though, as you've mentioned, if the people who right now are finding ways to monetize that and that's withdrew because these engines just aren't reciprocating the value that they take away by utilizing all this information in one place.
Future Considerations in AI
Yeah, that's, it's very surreal, dystopian sort of scenario that I actually am really struggling to wrap my head around how that would work without decentralized AI, without finding a way of using almost the mass appeal and the fact that more people can find the right information, because that's one thing that AI does very well. So, yes, maybe on an individual basis, you wouldn't have your own website that's drawing in as much money in your niche, but it would be exposed to so many more people that there's still like a real ability to monetize there. And also because of the utilization of AI, there is that reciprocal nature of it will save you time.
Inviting Guest Speakers for Their Opinions
So this is really fascinating to me, but I would definitely open up the mic and the stage now to some of our guest speakers and get your take, because you're all building incredible tools. I would love your take on where you fall on decentralized AI, centralized AI, and any take really at all, based on the conversation so far. Devon, I'm throwing the mic over to you first because I got an emoji from you. So game. GPT, Devin, Mike, over to you.
Participant Perspectives
Yeah, what's going on, guys? Happy to be here. Thanks everybody at Kipp for the invitation. Always love joining you guys. I think we've been on a few spaces already. It's always a pleasure adding to the conversation here. I'm always obsessed with making it really simple for everyone so that they can follow along. And then once we kind of build a foundation, being able to make it to build up from there.
Decentralization Analogy
So the mental model, the analogy that I'll propose to all of you guys is, let's say, in this fantasy world that we're building, we have all of the money in the entire world. Would it be safer, you think, to have all of the money in the entire world in one safe? Or would it be more safe to have small increments of that money in like a network of distributed banks all across the world? Right. That effect, I hope, if you came to a conclusion in terms of safety, this network of decentralized banks would be a safer way if you have all of the money in the world than one centralized bank that owns anything.
The Risks of Centralization
Because if some person, no matter how secure it was, found a way to get into that centralized bank, it would wreak havoc for everyone. This cause is one of the positives of the blockchain story. It lends itself to the phrase, if it's not your keys, it's not your crypto, and it's that everyone, as long as you have your private keys, can own the value that's in your wallet, and no one can really take that from you effectively. The way that I see decentralized AI is that same concept, but bridged to data.
Ownership in Decentralized AI
If you have your own data set, if you have your own way of computing AI, that is a much better way of what the AI future can be than if it is centralized. Taking owner, I guess taking homage to what Julian is talking about, having a decentralized AI future where we have data that is distributed across everyone else's computer, much like I said, these banks are everywhere across the world, is pretty much the only way to have ownership of certain things, depending on what you're building in AI. Additionally, it is probably the only way to have privacy.
The Importance of Privacy
And privacy is important based on what it is you're exactly doing with it. If it's something like your pet's name, or if it's something like what your favorite video games are, I think privacy probably doesn't mean too much to you. But if we expand out where science is going and what industries are being digitalized, what if it's your entire genome? What if it's your DNA and what your a, cs and t's and g's are, and like the actual genome that makes up your DNA, right? Probably a little bit more private.
Wrapping Up the Discussion
Probably something that you'd want to hold on to, just like you want to hold on to the private key for where you hold your bitcoin, right? So that is like one of the many ways that I would kind of think about this, is decentralized AI another thing that we can, and there are many kind of aspects and dimensions that we can angle on the benefits of decentralized AI. One thing that hasn't been talked about is like, bias.
The Role of Decentralization in AI
We've seen a ton of other AI models come out that have made the news, like Google's, Gemini, and a couple of others that have talked about the bias that comes with certain AI models and the reasoning that it comes to based on how it's trained. Decentral AI is the only way to kind of permit bias because if there was only one or two AI models that everyone uses to do certain functions, if that model becomes biased in some way, it could just like in the same way that a thief could go into the one safe, really cause everyone to interpret information, come to a resolution that actually isn't close to truth. So that is, that's my take.
General Themes in Today's Discussion
I want to. Today's call, especially with the topic, is more general. So all of my responses are probably going to be more philosophical, more high level, versus like, what we're specifically doing at game GBT. But we'd love to get in specifics whenever the conversation goes that way as well. Yeah, I love that, Devin, and thank you for that. That definitely helps me wrap my brain around this a little bit more, like, yeah, decentralized. Feel like this is almost the perfect space for this conversation as well.
The Importance of Decentralization in Financial Context
Like, I think there's going to be a lot of web three years here who understand the importance of decentralization. When we draw this back to, you know, financial side, when you look at what happened in 2008 and, you know, a lot of people at this point from the, you know, the credit crunch, the financial crisis, like, when too much of the decision making goes to centralized entities, even when there's multiple of them, that should be acting, you know, in a specific way from experience, like, we have seen how that goes wrong. So the idea that can't happen on a much bigger basis if there's like singular entities, not even multiple entities, where there's level of competition that, you know, in cares, fairness, I think that's, again, really dystopian.
Concerns About Centralized AI Futures
I do. I don't see that future ending so well. Like, without some real huge caveats. Cluster, your hand was raised. I'd love your take on the conversation so far. So, yeah, so basically I joined in late, but, what I have understood that we guys are, debating around decentralized AI. So what I understand from decentralized AI is and what were talking around, what Jillian was talking around, what Jen was talking around, and what even game GPD was talking around is that in the centralized ecosystems, right, be it built by Google, be it built by OpenAI or be it built by anyone, the major thing that is there is lack of trust.
Trust in AI and the Role of Web Three
And what I think is web three is all about trust, right? Being it transparent or being it whatsoever. The core properties of blockchain are. So what I think is AI lacks trust. And web three gives that scope of trust in that AI. That is why we would move forward for decentralized AI, right? And for. For builders, right? For builders, right? Initially, it is not possible to get something of a whole decentralized ecosystem where everything is decentralized, right? The current state of decentralized AI is somewhat like most of the brands and companies and protocols which are working around are almost web 2.5, right?
Challenges in Achieving Full Decentralization
Most of them might be utilizing those open source models. Most of them are trying to build those models over customized data sets, maybe over private data sets and stuff like that, right? Or even like they are trying to have a collaborative platform where people are coming and doing crowdsourcing over data and stuff like that, right? But in the end, the ultimate goal for decentralized AI is to bring trust in the AI space. That's what I feel, and that's what we at cluster are also aimed for, right? So, yeah, like, happy to chat around more over this topic.
Efficiency in the AI Debate
I do love this idea of trust and the value that provides a lot of people and a lot of faith as well. Right now, in terms of why you would, you know, why you'd learn to one end of the spectrum in terms of decentralized and centralized, I think that's definitely going to be a key element of it. I think faith, trust, whatever we want to call it. I do think this is becoming more and more of a paradox with how quickly AI is developing. You know, like I, you know, this idea that previously the term or the, even the phrase, like I believe it when I see it, like I do not believe.
The Perception of Evidence and Reality
Now you can say that word with like a straight face. Like, I do not believe that seeing necessarily will believing at this point unless it's in the real world. But when we talk about video footage, etcetera, there's so much more going on behind the scenes here. And when we think of centralized AI and back to Devon's point on game GPT, this idea of like actually an entity that has a bias and therefore wants to provide you with a specific narrative to essentially, you know, control your own opinion, I think that's again a reason why people are going to naturally lean once they understand that.
Understanding Consumer Choices
And I guess the understanding is the big issue here to a more decentralized path. One question that I think we have to pose during this show, not to be too pro or against, just pure neutrality and for information and for education on this subject is the efficiency debate. And I think that's definitely one thing that I consider in terms of utilizing this technology as a creator, as you know, wherever I get value. So I'm going to come from it as a creator here, because that's realistically where I derive most value from, is how can I create the best content or incentivize the best conversation and education around these topics with as small of a time as possible?
Balancing Efficiency and Value Creation
These are the two things that you're sort of trying to balance, and how do I provide the most value with all of those things in mind? Efficiency then becomes obviously hugely valuable and a huge piece to this equation. So some, I think definitely, I have seen, will argue that the centralized AI is less efficient compared to centralized models. What's the take on this and how do we address these concerns where realistically, I think it's going to be a huge uphill battle if we have a centralized model which can get you this information immediately or almost immediately, and support you in terms of the value that you create, in whichever area you derive value from, and a decentralized model that can do the same thing, but it doesn't do it as efficiently and it does it slower.
Diversity of AI Models and Consumer Preferences
So I'd love to throw the mic over to you on this one and then I'd love to get back to Julian or Jen to maybe add some colour to this. Hi everybody, Danny here, co founder and CEO of I'm excited to join the stage and be part of this conversation. I come from a gaming background more, so I'm definitely enjoying the conversation. To learn more. One thing before I get into the question that you just asked, is the thing cluster protocol just mentioned about the trust and the transparency, I'm all for that, but I'm gonna give a bit of a contrarian, let's say, take on this.
Understanding Consumer Perspectives on AI
Most people, when they heard about the advances of AI, I believe that the first thing they thought was, I'm gonna make my own AI, have three employees and make a billion dollars. Right? Just the same thing with admitting it, that we have no idea what decentralized AI is. And that implies it is like a centralized version of a thing where you have a. Basically a script, a robot, a server that makes you money. This is the simplest take to describe what most people may have thought at a point.
The Question of Efficiency
Coming back to the question, and Jack, could you please repeat it? Because I'm not AI and I have a 16 megabyte. So very much around the efficiency side. So, you know, like, is it likely that efficiency is going to lean towards centralized models? And if so, how are we supposed to dissuade users of these models to go to the less efficient option? Well, I think I. If it comes to a consumer grade app and access the market, and the users will just pick the one that is best for them, like it has always been like this.
Consumer Choices in Technology
Of course not. The best technology wins, the best adoption wins, and the best sale of it wins. But I do believe we will have a split world between decentralized models or decentralized computing power for AI versus centralized ones that are very specific to a business. I have very, a very hard time believing that Google, Apple or OpenAI will all of a sudden have a change of heart and say, yeah, we're gonna just open this for everybody. And everybody can benefit at a certain rate, because at the end of the day, behind the screens, we are humans up to this point.
Market Dynamics and Future Prospects
So mark my words, this is 2024. Maybe in a year we'll have an AI joining the space and counter arguing this. And it's just a matter of what people use to achieve the objectives they desire. This is how the balance is given by the market based on the needs of it has, and the person, the service, the company that offers the best solution to a problem wins in a certain way, the customer's commitment. Yeah, I got. That's exactly the thing.
The Efficiency Debate in AI
So is this logical? Is this shrouded, in fact, right now, this idea that the centralized AIH will be less efficient? I think that's probably a question that we should lean into a little bit more. And then the question is there a world where people still understanding the long term cost of going down these centralized routes would still choose to go down a different avenue? Because maybe not necessarily. Efficiency is a bad term, and I'll elaborate on that in a second.
Quality versus Speed in AI
But I do think there is something to this idea of efficiency over quality, and that is probably where I feel like decentralized will actually win out. And that is also where I think the utilization or the use of these things that gets a little bit grayer. Do you want answer fast or do you want a quality answer? And I think that is somewhere where decentralized AI definitely could win out.
Trust and Data Security in AI
So first things first. So basically, what I told about trust is all about data, right? So there are many factors where though OpenAI claims that chat GPT doesn't record their inference actions, right. Inference actions or inference queries. But the main part of the major part is when we are extending the context window. Yeah. When we are extending the context window, when we are adding more cash to it, and we are making centralized systems or centralized AI more convenient in terms of having a better UI on a better UX, basically a better user experience in order for AI to be more human friendly.
The Importance of Data Security
Right. Then there are chances of data leaks. And we have already seen that there was mistral, right? Mistral. There was a model of mistral where 800gb of data was already leaked from there. Right. So there are many possibilities which decentralized AI could cope up with. Right? There are many possibilities which decentralized AI could cope up with. But yeah, we need to make sure that the thing that we're using for privacy, be it fhe, be till or whatever we are using, right.
Ensuring Privacy and Security in AI Solutions
We need to make sure that those solutions of wrapping those data into private layers are secure enough. Right. Initially there was a wave where people were using ZK co processors. Then there was a wave where people were using fhe. Now that has already been fair and people are working around on something better or newer. But I think that, yeah, a combination of ZK and fhe could be a future. So that is where I think things should align around. That is where I think people should work around, where there could be a collaborative approach towards ZK and fhe.
Efficiency and Speed vs. Quality
I do also take your point about efficiency and speed. Like, I think my biggest thing is that they aren't the same, are they? Like, if I want the best answer, and I ask chat GPT for answer around a subject, and I say, look, I don't care about time, I care about quality, to this response. This thing takes days. I don't know if anyone's trialed this before. Like, it literally takes days. So right now it is 100% like setting its parameters around speed of information provided, not quality of information provided. So when I consider decentralized and how this debate around efficiency comes into play, I still feel around quality and ownership. These are two really fundamental pieces that when people are asked these questions about what they're looking for as a consumer or a user, they, those things aren't separate.
Consumer Expectations
You know, like, it's not, oh, I want it fast, or I don't care about how, like, the quality. That's never the answer people want, unfortunately. Like, I think it's just human nature, but they want it fast. They want it the best. They want it, you know, it's all of these things combined. And I think, again, this is where decentralized and really starts to find its feet because I do think through policy. I want to add a small anecdote because Jen is still in Thailand. We were just on the road show, pitching to university some of our AI solutions. There are many use cases in which a smaller model in which the model doesn't, it's not a one size fits all monster beast, like King Kong model, like, and the like, anthropic or OpenAI are trying to build.
Big Model vs. Smaller Models
They are trying to build the big model that serves 80% of use cases. For example, 80% of the, of any stupid question each of us might think of asking a search engine. They want to be able to serve that. But there are tons. Okay, when you talk about the long tail, there are tons of things. Let's say, for example, you wanted analysis on a particular topic. You wanted deep analysis, okay? And the knowledge base from that came in books. So in many instances, AI, if you use the monster models are actually designed to be a good conversationalist. Okay? Which is why they are prone to hallucinations. They are prone to hallucinations because they want to keep the conversation going.
Hallucinations in AI Responses
They are, they are not trained. You have to prompt them very specifically to say that. If you don't know, say you don't know. That's like the hardest thing to say, that, you know, stop making shit up. Right? So we're talking about quality. Sometimes quality is just an AI that says I don't know because I don't know. Instead of, you know, I'm a conversation artist. I can calculate the, I can calculate the probability of the next token, the next word in this sentence that might make sense or whatever, because that's my KPI, that's my brief, you know? So the quality, when you talk about quality sometimes is, Yeah. It's about not having all our needs being served by these monster models.
Monster Models Limitations
Okay. I think this goes to a point that was raised just now by Devin, I think, yeah, not have these monster models who are designed to serve 80% of our dumb questions. Yeah. Like all our needs. Yeah. What this answer reminds me of. Julian is actually, I'm a big cuban fan, like Andrew Huberman, huge fan of like his content. He really serves like just the most fascinating, like PhD worthy information to people for completely for free through his YouTube and podcasts and his shows. And one thing he recognized really early on because he saw, you know, GPT and some of these other tools as a way that he could provide more information to his users with that speed in mind and with quality in mind.
Challenges in AI-Generated Research
And he started asking GPT, I think, specifically, can you service me the case studies, the, you know, the research, the papers on these things? As I try and go in, it was giving him fake papers. It was literally, it was like, yes, here's answer to why you are correct about how sleep deprivation causes lower IQ output over time. And here's a paper on this. And he went into the paper and it was a completely fictionalized paper because it just wanted to keep the conversation going and save Andrew the answer he wanted, not the answer that he needed because he's a researcher. I think that is terrifying when you think about how these things currently work. And also very significant to this point of efficiency, quality and ownership as well.
Accountability in AI Ownership
Because another thing about ownership is accountability. I think people don't always think about those two things in tandem, but they very much are. So if you own something, for example, information that you're providing or content that you're growing, whatever that might be, you're also accountable for that information. With an AI tool, the accountability is really not there. And it's serviced every time you use one of these tools. And at the bottom, in really small writing, it says, this could all be made up in some way, shape or form. Like the small prince. Like when you ask it a question, it literally says at the bottom, yeah, if this is important to you, maybe just don't actually like, take this verbatim.
Impact of AI Limitations
Maybe this is actually not a fact and we're just making this up because we want to continue a conversation. Wham, your hand is raised. I'd love your take on this conversation so far. The first thing, I think we should rewrite the title. We have no idea what AI is and then just put in some extra stuff there. And partially this is true when it comes to decentralized AI. If I were to picture in my mind something I would see like a giga brain, right? So if, for instance, you take Ethereum which is the, which has the ethereum virtual machine that right now processes the smart contracts and holds the blockchain. I would see the same thing for a big giga brain that is spread out.
Purpose of Decentralized AI
But when it comes to why or what is the purpose of a decentralized AI, I think this is where it would be set apart from the big companies, because most of the, let's say, okay, wham, wants to have a decentralized AI for our gaming ecosystem. Already we've narrowed it down for a specific purpose in our vertical, doing things that we want. And right now I would argue that it's not necessarily AI, it's just a lot of code, probabilistics, machine learning, not necessarily, not intelligence like you just mentioned with the footnotes. And in terms of distributing it in this context, we would actually distribute the truth of things.
Decentralizing Information
Exactly like with the blockchain, you distribute the truth of transaction that the truth is everybody has the truth. Okay. When it comes to decentralized AI, what do we decentralize? And in this case, this is just a personal take and a bit of a philosophy like game. JPT said we're decentralizing what we consider to be the truth for this specific model. And I would say AI. Right, so then comes with what is the purpose of this? Of course, to help the community, the group, the business in some way to get an edge. It all comes down, actually, when you guys were talking in my mind when a few movie shots from Terminator, because at the end of the day, somebody will poke around and want all the resources.
Future of Decentralized AI
I'm just. So when distributing things, the purpose would be to get some sort of edge that needs to be protected. Okay? You're decentralizing it with the help of blockchain, but the blockchain is just the money for AI, right? You don't have KYC for models and you don't have government issued id. So they cannot trade money, but they can trade crypto. So the speed in that case is just a matter of transferring value. And this is just. I would stop here to get the conversation going and not bottling it up. No, I definitely see your take here and yeah, I think honestly, look, we are getting very close to the end of the show.
Concluding Thoughts
This one's been fascinating to me and really appreciate everybody who tuned in. Thank you for all the speakers who've came up as well and just really added to this conversation today. It's been fascinating for the listeners. Let's get the likes and retweets off the room out for this final ten minute segment of the show where I'm definitely going to try and get into the specifics of these projects and how the centralized AI factors into what they are building, because I think that's definitely Devon, you mentioned it earlier on, so I will provide the mic over to you in a second, and then we will keep this one going.
Decentralized AI Utilization
Devin, your take on this, like, when you consider decentralized AI and game GPT specifically, what does that mean to you? Like, where is the utilization for you as a game, as a project when it comes to decentralized AI? Yeah, and I'm definitely going to answer that and also talk a little bit about Julian's last response in terms of, like, the foundation models and like, narrow models. I have an article pulled up, and this is Elon's Xai company. And depending on if you have premium x, you may be able to use Grok, which is Xai's model. He recently built a facility for Xai in Memphis, Tennessee.
The Cost of Models
And this supercomputer, or supercluster, has 100,000 Nvidia H 100 GPU's. And it's estimated it costs around three to $4 billion. Right. So whenever we think about like, the cost that it takes to make a model efficient, that really does price out, like a decentralized way. So if you want to be a foundational model, like Julian talked about, you want to be able to answer 80% of any type of question, you really are going to have to spend a lot of money. And that is where the openais and the groks, or Xai's and we have Claude, are really kind of effectively making their corner of the market.
Open Source Decentralization
In terms of the decentralized way, I think there is a couple of ways that people could actually do it. And the good thing about everyone who wants to perform a decentralized way of aih is that open source, much like in crypto, is very much a factor in AI as well. So not Claude, not chat GPT, not grok. But there are other models like Mistral, there are other models that actually open source everything, and you can actually fork their model and then use that for your own convenience to do whatever you want in your own decentralized way.
Narrow Models for Specific Applications
In terms of what we're doing, we are creating our own model to be really good at writing video game code for a very narrow subset of video games. At the moment, game GBTV, one is really focused on hyper casual games. So when that comes to games like fruit ninja games, like subway surfer games, like Candy Crushed games, like Snake IO and all of these kind of viral mobile apps. We found that creating a narrow model to train datasets to write codes like these games, and then to effectively reskin the visual assets that are required, whether they be 2d or 3d visual assets to make games like this.
Future of Decentralized AI in Gaming
That's our own kind of approach on how to use AI models decentrally, to have a very narrow focus, at least at the moment, the decentralized AI path of really trying to take what's already been given by these foundational models and then to fork it using open source variants, and then to narrow it to do something that your model is going to be very specific at doing is, I think, at least at the moment, the future and where most people are going to go. But I would love to see a decentral model kind of build kind of like these foundational models that compete with opening OpenAI and everything else.
Challenges Ahead
It is kind of a cost problem at the moment, but I'm super positive, super bullish on AI crypto, as usual. Thank you so much, Devin. And again, thank you to all the speakers who've contributed today. I think we all get very close to the end of the show. So, Julian, I know we have a couple things to go over from KiPP protocol side, one of them being any news announcements for our audience today around KiPP that we'd like to, you know, confirm before the end of the show. And also we have a couple of questions that have come in that have won the AMA questions.
AMA Questions and Discussion
So I definitely want to make sure that we highlight those as well. Would you like to round off, initially, any thoughts from kid protocol side, based on the favoring of this conversation today and any announcements or any news before we run out the show? No, I think let's answer the eme questions. Okay. We got a lot of great responses, so let's do that. Yeah. Awesome. So what better way is your team looking to sustain AI decentralization in the next ten years, considering future challenges like exponential growth in terms of user size? Well, I'm not sure the questionnaire is from web three, because he said ten years.
Steps Towards Decentralized AI
Yeah, I'm kidding. The short answer to this is that we see the steps towards decentralized AI. Actually, right now, the steps needed to get there, to us, are crystal clear. We need to solve the bread and butter issues that's preventing people from creating more products that will attract users, more decentralized AI products that will use this if we solve the bread and butter issues. And the great thing about sometimes the crypto bull market is a lot of incentives come in automatically because of that.
Revenue Driving Objectives
It creates the attention, it draws the eyeballs once every couple of years. But the core revenue driving objectives of product makers need to be met. So we need to solve those first. In our case, we had focused on very large builds. For example, one was with open campus. So we are trying to build a massive online course provider together with open campus. That was OCP. Twelve in which 10 million worth of was approved for the building and managed by KipP for the building of OCU.
Large Builds and Hackathons
So that's a large build. And we are doing another large build with our partner Moit in the entertainment space. After that, we are just going to use these two examples to organize hackathons so that solopreneurs can take reference from all the stuff that we are doing with the larger projects. So one step at a time. My opinion, or our opinion, is that right now we are kind of already being monopolized.
Consumers and the Fight Against AI Dominance
We are already consumers, okay? In a world that is already dominated. Dominated by the AI bigwigs, the big tech companies, open AI, they are already winning. So what we can do is we have to fight. What we must do is we must fight. Otherwise we're just going to end up as consumers. Not much needs to be thought of, in my opinion, ten years from now, because the fight is right now. It's not like we are planning for. It's already happening to us. We are already being robbed. Okay, I was going to use another word, but yeah, starting with f, but yeah, we already being robbed. I don't know if you need to elaborate on that bit. So right now we just fight. Right now we just built and we just fight. We just fight back, you know? So that's our take anyway.
Scholarly Concerns on AI and Copyright
Okay, let's get to this next one. This one's a difficult one. I, as a scholar, build an AI model and this isn't me. I am no scholar. I, as a scholar, build an AI model and train it. Use our. Sorry. And train it's using one of your AI resources. I publish a paper and or register a patent. How the copyright of the patent is defined in the paper. Should I cite your product? Do you have any portion in the patenthood? This is all one word. Patents, patent. Okay, I'm not sure that makes complete sense, but. Julian, did you get the gist of that one? I think I got it. Yeah, you got it? Yeah, yeah. No, you don't need to cite us. You're good.
Integrating AI on Blockchain
Yeah, we're not publishing any model that you're editing or anything. Yeah, no, we're not publishing base weights or anything that you would need to cite if you were going to be doing research. If you sell your model via our site, that's, we're perfectly happy with that. Amazing, amazing stuff. And then the final question, what is the centralized AI? Also, how are you going to implement or integrate AI on blockchain? Let's say you do. Then how are you planning to create sustainable incentive structures? So I think I would summarize that on the, because we've obviously elaborated hugely on what is decentralized AI in the show today. I think the final part of the question is the one that we should focus on, which is sustainable incentive structures and how you guys plan to implement those.
Economic Incentives for Ecosystem Growth
This director, that cape breed for us, the first thing is there are two, I guess there are at least two layers of economic incentives needed to bring about this ecosystem. The first is the basic p and l, or revenue incentives needed for products to be made. And then we are actually working on a new platform that on top of this introduces maybe slightly more speculative nature, on top of the nature of the asset in the first place. Because imagine this, you can buy and sell an asset and the person who created the asset could sell it, or parts of it anyway, could fractionalize it and sell it, which we use the ERC 3525 token standard to do. But the value of the asset, obviously itself might go up, might have a market in itself.
Navigating Speculative Markets
So I mentioned a little bit just now that every couple of years when the crypto bull market comes, certain attention and eyeballs and resources and capital is drawn in. That's good for the very. I would say this is always a short term boost in all the attention and the capital that follows. But that has to be built on top of the base layer, which is the p and l layer, the revenues layer, because otherwise, then all you have is a boom and bust cycle. So the sustainability would come in. And in doing this, we didn't plan to do this in the beginning, but it was more of a pragmatic approach to say that, look, it's going to happen. The speculative nature of us being web three is going to happen, but we have to plan our project based on something that will work in any cycle.
Planning for Sustainability
But, you know, once every couple of years, there will be a speculative part of the cycle that comes and that relates more to the asset that is being speculated on. Right? So we are doing that. So the sustainable part of it comes from the p and l layer and not the speculative layer. That's how, think about it. That makes so much sense to me. And look, guys, I have learned so much. Like, my. My head hurts. Like, I don't get educated this heavily in an hour's. Like, that's just so much information that I'm going to have to wrap my brain around on this one.
Final Insights and Announcements
But I just, again, really want to thank all of the speakers. I would love to let the speakers share more about any of their project announcements if they want to. And I really want to thank them for joining all space as well. Amazing. Okay, Juliane. Wham. GPT Cluster, if you guys want to share any information about your projects to this amazing audience, which I was just about to thank. So thank you, audience who've tuned in so far as well, and then by all means, go ahead. Now, this is your opportunity to share a little bit about yourselves before we wrap up the show.
Exciting Project Developments
Yeah, I can break the ice here. Okay. Yeah. No, no. Yeah, so, yeah, like, for announcements. Yeah, we recently hit 100k on Twitter. That was one of the major announcements around also. Yeah, we are super excited for our mvp, which is going live next month. So stick tune, stay tight for it. So, yeah, these are the two main announcements. Apart from that, yeah, there are a couple of other alphas which I give either ways in many of the spaces around. So, yeah, thank you guys for hosting me, and it was a pleasure to be here.
Wham's Engagement and Innovations
Thank you so much. You've been awesome with your contributions today. Cluster, wham. You have to. So I'll throw the mic over to you before we finish off with Devin on the game DPT side. Awesome. So wham has been around 2019. We're building, if you want a mega app with games where people compete in tournaments to win digital assets, we've grown quite a bit, especially in a bear market in 2023. And we've been toying around with AI before. It was cool because we had a problem, an actual problem, and that was we didn't have the funds to actually license all the games that we wanted.
Building a Decentralized Infrastructure
So we said, okay, we have to figure out a way to have content, and that was our way into the AI world. And Wham. Right now is building this mega app. We've had quite a few success, and we are the number one DAp on BNB chain in 2022. This year, went to Solana as well, and we're actually building a decentralized node infrastructure for AI security for our platform because this was another problem that we faced, hacking and cheating. So, yeah, this is Wham. Thank you for having me. Excited to meet you all you guys and see each other next time.
Final Thoughts and Wrap Up
Thank you. So yeah, again, really appreciate the contributions. Devin, over to you to round us out before I throw the mic. Quickly back to Julian and Jen for any final thoughts. Over to you. Of course keeping it quick because I have talked a little bit about what game GPT is doing and what our product is. We have two main products at the moment. We have challenges and arcade challenges is actually in the immediate future going live and we have a skilled based wager for some of the biggest web two games in the world that are integrated with our platform. We have Dota, two counter Strike, two, PUBG Mobile, PUBG Desktop, Destiny, two Clash Royale, Brawl Stars and Chess.com all available and live on our website to create skilled based peer to peer wagering and to create a challenge.
Launching New Features
And see everyone on our platform join if they want to see if they can get more kills than you in a week or more headshots or more wins. However you want to slice up the challenge using the games API that is going live as early as this week and tournaments are going to be announced in the next few days. Our second product is arcade and that is the actual AI generated games. Our app is coming September of this year, so just in the next 30 days we will already be live with our mobile app. All 15 games are going to be incorporated and shortly after an NFT collection is going to launch as well. So really looking forward to that.
Crossovers and Collaboration
And if any of that sounds exciting, just follow our profile and check for updates. Love it. Let's do a crossover game. DBt oh that's what we love to hear on this sort of show. Like just great minds coming together building cool shit. So look guys, again, really happy with how this one's laid out. Feel like I now don't feel so seen by the title of this show. Actually, I have some idea of what decentralized AI is listeners. I hope you guys do too. I really I thoroughly enjoyed this one. Would love to see in the comments how you guys have experienced the show.
Audience Engagement
And look, hit me in the DM's as well. Like we're all in and we'll iterate in here. This is new. We want to provide value to you. That's why we're here. That's what we're doing. That's what we're doing here. So hit me in the DM's or hit us in the comment section for how you enjoyed the show. Any feedback, any improvements, do let us know. Julian, Jen, how did you guys find the show? Today. Any final thoughts before we close out?
Closing Remarks
Excellent show. Thanks. Thanks, everyone for joining our audience and also our panelists. So we are Kip protocol. We are dead focused on pushing out actual use cases in decentralized AI. We want to be the leader in that. And we have our note sales going on. We have our uprising campaign that is continuing. And anything follow our Twitter. Anything more, follow Twitter, join our discord, all that. And Jen, please close out the show for us.
