{"id":"z124wIkumIA","title":"[REDACTED] 2024 | Showcase: NEAR AI x HZN Incubator Cohort","description":"Teams building the next generation of decentralized AI applications will share what they learned from the NEAR Horizon AI Incubation Program.\n\nJarrod Barnes - NEAR Foundation\nWhitney Gibbs - Ringfence AI\nDon Gossen - Nevermined\nJeremy Hazan - Hyperbolic\nJonathan Jaranilla - Exabits\nDylan Zhang - Pond\nShu Dong - MIZU Global Inc.\n\nTakeaways:\nPond’s Crypto-Native AI Models: Pond is pioneering the use of graph neural networks specifically for crypto on-chain data. This allows for advanced predictive insights, such as anticipating token movements based on blockchain behaviors. By focusing on a graph structure unique to blockchain, Pond offers tools and models that are custom-built for crypto applications rather than traditional language-based AI.\n\nMizu’s Decentralized Data Network: Mizu addresses the scarcity of high-quality data for AI applications by building a decentralized data processing network. This allows users to easily access and filter large-scale, diverse datasets, offering affordable and scalable data solutions. The result is an open-source data pipeline that enables developers to create specialized AI agents tailored to specific use cases.\n\nNevermined’s AI Payment System for Agent Commerce: Nevermined is developing a payments system designed to facilitate transactions for AI agents. Their vision is for AI agents to autonomously buy, sell, and transact in an online marketplace. By standardizing agent communication and transaction protocols, Nevermined is helping to establish the infrastructure for a new era of automated, agent-driven commerce.\n\nExhibits’ Accessible GPU Marketplace: Exhibits aims to make AI compute power more accessible and cost-effective, offering a GPU marketplace that rivals major providers at significantly reduced costs. Their focus on affordability and quality attracts both web2 and web3 projects, including those that may have previously been priced out of high-performance AI computing. This allows innovators from various fields to leverage AI without facing excessive costs.\n\nRingfence’s Data Monetization Platform: Ringfence empowers creators and consumers to monetize their data by providing tools that allow users to secure and profit from their digital assets. With partnerships in place and an easy-to-integrate API, Ringfence is set to change the way personal data is managed and valued. This approach aligns with a broader movement towards fair data practices and puts the control—and financial benefit—of personal data back in the hands of its owners.\n\nNEAR is the Blockchain for AI\nWebsite: https://near.org/\nNEAR AI: https://near.ai/\nIntents: https://near.org/intents\nChain Abstraction: https://www.proximity.dev/\nSharding: https://nearone.org/\n\nFor the devs\nGitHub: https://github.com/near \nDev Docs: https://docs.near.org/\nDevhub: https://devhub.near.page/ @NEARDevHub\n\nConnect with socials\nTwitter:   / nearprotocol  \nBlog: https://near.org/blog/\nMedium:   / nearprotocol  \n#NEAR #nearprotocol #NEARAI","channel_name":"NEAR Protocol","published_at":"2024-11-14T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/z124wIkumIA/hqdefault.jpg","transcript_text":"all right well we have everyone here now great to see you all uh folks my name is Jared Barnes I am uh the head of founder success at the New Foundation have the privilege and opportunity of working with all of our Founders uh and builders on near and uh I could not be more proud about this panel and these uh amazing Founders on stage here 6 months ago uh we really doubled down on growing and scaling our AI ecosystem and as we were in that process of growing uh we said who are the best Founders in the space that we can go find and work with uh and those are these folks sitting on stage and so we have teams building at all layers of uh this decentralized AI stack that I'm really excited to to get into and really excited for you all to hear a little bit more um about each and every single one of them so Dylan I'm going to start with you and if we can go down the line uh just sharing a little bit about not only what you're building but your origin story to how for yourself and pond came about sure uh thank you Jared and thank you nice to meet you everyone so I'm uh I'm Dyan one of the co-founders at Pond and also we are one of the incubes of near AI Horizon uh incubation Pro uh program so uh at Point what we are doing is we are developing Crypton native models trained on onchain data um so for examp I can give you some examples because AI model seems to be a little bit vague um so when we are saying saying that one token is about to pump uh based on um you know all the training results that we have on the onchain behaviors of previous tokens what might happen and based on the behavioral prediction so we can say you maybe this token has some potential in some ways so this is one of the use cases of unchain data and also unchain data models um so we started from developing user search engine because we feel like it's very difficult to find who knows who and who knows who well and it seems like it's very it's very hard to do such a thing in web to because all the data structures across different platforms are totally different like today I'm using Facebook data but your LinkedIn profile also has some of the social information of your of yourself but there's no such a third party can do this so we started from onchain data because all the financial data and social data all at one place and we developed graphical algorithm to calculate the you know distance between different nodes which are different users and uh from calculating the distance we'll be able to find who knows who and who knows who well to provide a such a user service uh user search service and later on we feel like since all the social data and financial data are at the same place why don't we just train an industrial Scale Models so that's how we started this project um we are more like a model developer developing Crypton native AI models to power all the Crypton native use cases right hey guys I'm Jonathan uh one of the four core members of exhibits and we got our origin story from building having experience in building over a 100 data sensors in the in our past lives and now we are leveraging that to produce uh compute uh for the decentralized compute Market uh and also decentralized AI companies uh web 2 and web 3 gaming companies uh and and really focusing on gen AI companies uh who want to uh just build uh with us so uh yeah so leveraging that uh data center experience we decided to provide compute uh to everyone in the market um some of the things and compan we are working with uh web 2 gen companies uh web 3 gen companies and also a lot of the cohort members inside uh near Foundation all right I guess it's me first off uh thanks for having me near I'm Dawn I'm the co-founder and CEO of nevermind so my origin story my parents met in 1968 oh no I'm just kidding um so I've been focused on the ml and AI space for going on 20 years now studied engineering at University and then went into Commodities trading um on the risk side doing back then what was called statistical modeling which then became machine learning and now has been co-opted by the AI brand um did that for a few years and then parlayed that into a career for the better part of a decade and a half in IT consulting as a subject matter expert in data and analytics um that had me TR travel all over the world building large scale data States for ML purposes for some of the biggest companies on the planet HSBC L'Oreal AXA mizuo yada yada um caught the Crypt bug in 2016 uh went down that rabbit hole um was enamored with the um I think sophistication um and elegance of using a blockchain for providential integrity purposes um so in 2017 I co-founded a project called ocean protocol which was one of the first attempts to bring Ai and blockchain together so I've been doing this for the better part of a decade now before it was ever cool um a lot of people used to look at me side eyed like what the are you doing um but yeah I've been in this space for a long time and where we've landed now is with it we've really distilled basically a decade's worth of into a hyperfocus on AI payments um so never mind we have a very specific view of the world we believe what we're witnessing right now is the rise of a new consumer that's going to manifest as trillions of AI agents and that these agents are going to change Commerce forever um we believe that this is a category defining moment um the new category being AI Commerce and in response we're building the PayPal of AI to be the engine behind AI Commerce and get these agents paid boom Jared thanks for uh for moderating and it's been an amazing experience working with near Horizon uh in this previous incubator my name's Whitney I'm the co-founder and CEO of ring fence AI uh I've been in the space since 2017 and you know as a an observer at first and uh then a Community member for a lot of now defunct projects um but I you know I really found my passion in mining at first and over the course of a few years I built a a pretty successful mining podcast called the hash rate podcast which is now the mining podcast um still in operation today and uh built a company called Compass mining from you know zero to about 100 employees Compass is now one of one of Bitcoin uh bitcoin's largest hash rate managers and we serve thousands and thousands of retail Bitcoin miners um in 2022 I really caught the AI bug we were getting a lot of inbound from people who wanted us to stand up GPU facilities for them and you know eth mining was was dying so got really curious as to why so I started looking into where all of this uh compute power was was being directed and really caught the AI bug stepped down as CEO of compass and focused all efforts on uh learning more about the AI space became a student of uh of the process of how to build and and scale one of these businesses uh developed a couple of different uh AI products that are now in operation mostly in the UAE um and then in 23 started ring fence with a couple of uh a couple of friends and our goal with ring fence was to help people monetize their data simply put we found that you know right now there's a tremendous push for AI but the barrier is going to be the the quality data that people are able to to gather and much like the thesis of never mind that Don has uh we believe that there's going to be trillions and trillions of these agents that are out there in operation and each one of them are going to need different types of data and most of that data is going to be unlocked by tapping into consumers uh which presents an opportunity for these consumers to get paid which is really where we focus at ring fence so uh the first product that we built was geared towards creators and that is eer the eer app uh went live in September of this year we had over 140,000 creators that were excited to start using the platform so that's been in operation for a couple of months now and today actually if you go to ring fence. we've just released our API so that anyone any platform any company has the ability to integrate ring fences data governance and monetization into their own platform so you can collect protect and monetize your own users data and we believe that this is going to be the real unlock for Consumer data to help us break through the current data wall that a lot of companies are facing um while paying people for the data that they provide right because we've all been giving our data away for you know as long as the internet has existed and now I think it's only right that people actually get paid for what they contribute so that's really where our Focus lies and you know it's been my journey up to this point thanks yeah hi everyone uh my name is Michu uh you can see that my name is not listed here because I almost cannot make it but like uh finally like I made it and then thanks shared to bring me back into the into the stage yeah I'm super glad to to join the Horizon program and share my story so Mizu was originated by uh it's actually founded like last year and uh uh like that very first we were thinking like what's the most problem like what's most fundamental like problem want to solve for the space and uh the problem we see is that there's very few AI applications coming out we see like a lot of progress in the found found Foundation model side the application ecosystem is still like not boosted and we look into that and see why and uh uh we look back into this kind of uh Web Two like normal application side we see that the open source code really boosted the application like web applications or whatever this kind of applications for this kind of space and for AI side we see the op this story most most of the data is actually multiplied valuable data I say by Enterprises by this kind of foundation model companies that's why we decided to go to the direction that we want to do some open data stuff not like private data and we believe that private data can also be open sourced if we can introduce some synthetics like technology here to anonymize it or by extract this kind of pattern and so we can share the knowledge and share the structure out for other application developers to to use so uh then like Mizu started to uh investigate in this kind of uh synthetic data Direction and also like the op General open source data Direction and uh then uh we realize that to make that happen like we really need a scalable Network and uh it naturally leads us to this kind of deepin solution because uh I really like the saying that cloud service is for scale and uh deeping is for hypers scale if we can Engage The efforts not only like the data but also computation resources across this kind of multiple uh like basically the user in the community from all over the world I think we can make like something really cool and very different we can accumulate the data accumulate this kind of uh uh computation resources together to build a very hyper scale but like efficient and transparent and open source open source data and share that to the like you know to to the to the uh bigger community and for the greater good that's like what we're building so the very first product we're building is basically a DP Network to uh basically like we be a Min like tegram mini app and a Chrome extension so people can run the code across mobile devices and laptops to help us to process data it's a decentralize the data search and data processing Network to help people to fill out this kind of valuable data across a major like more giant data set and then like give people like valuable insights that's what we're building yeah amazing so I'm excited to dig in a little bit more with kind of pointed questions for each of you and then we'll zoom out talk about the market at a high level what's happening and then if we have time for some hot takes and predictions but maybe Dyan starting with you right uh obviously super clear in the world of web 2 value Ural has gone to foundational models um I think certainly can see some parallels here in web 3 uh but you've taken a different approach in terms of your model architecture and using graph neural networks instead of the Transformer architecture can you tell us a little bit about some Transformers Inside has some Transformers Inside salute to Ed so okay okay salute to Ilia yes but for the most part graph no networks curious why um what makes makes that different um and how does that actually show up when people are leveraging and using your model right so uh basically what a AI model is really develop is U actually it has some prediction um predictive capability is really what the AI model is bringing to the world um and for prediction it's really up to uh what type of data you are predicting so for we 2 apparently the majority of the data and the data structure is the language data right that's reason why the foundation model of internet has to be a language model because it's where the value is from is the most of value is from this uh you know language U data but in crypto in web 3 uh we don't have well we have language data mostly on Twitter and you know different other other places but the most Crypton native data actually is really developed onchain data then what is onchain data first of all it's not language it's a it's a type of graphical data so it might sound a little bit but just forget about is is a jargon so what a graph data is it's really about as long as there's a data stru structure that can be represented by two components nodes and edges so nodes in a blockchain context can be wets and can be contracts and the edges are pretty much just interaction between different nodes like transaction you know it's mean to proof and transfer and all that so this a type of graph U data um when we're talking about the uh the blockchain onchain data so just because the the nature of the data is different which means that if you want to develop AI model um the types of the model must be different so that's the reason why we started with developing um you know graphical models and one of the most outstanding model architecture of graph data of graph models is the graph neuro Network um so that's what we are doing yeah amazing and Shu I'm actually going to jump back to you kind of building off everything doing just shared on the actual foundational model side you really planting a flag uh in the data pipeline obviously you've built a really powerful Data curation Network can you expand a little bit more assume folks know nothing about a data pipeline why is what you're building important and why is it impactful and tell us a little bit more about uh how that actually shows up when we're talking about the application ER I can give a very concrete example actually uh for example like uh uh for foundation model if you do pre- training you don't really care about the data quality not not like you don't really need to curate the data so well you still care about data quality for example the token distribution or something like that right so but like when it comes to like the foundation model cannot do everything right it's don't need agent agent and to build the agent you will need some customized data for some specific tasks that's called like vertical area you know you need this kind of specialized data and this kind of data is really the barrier of this kind of applications it's very hard to get this kind of application for example I want to build a AI bot which can help me to find a job in the market then it need to consume tons of resume data or tons of this kind of job market information and normally like as an individual AI developer you don't have that right you don't know where to find this data set and there's no like real time data out there for you to use so what we do is like uh you know Common crawl there's a there's a nonprofit organization called a common crawl and the common crawl craw the whole internet every month and they took like published snapshot out for example like last month common crw cross about like 2.4 billion web pages and including about 300 million domains okay there's tons of data out there but like people just don't know how to parse how to understand this kind of data and if you want to use AI to process the so in total like it's about like 20 terab of the data compressed about like 7 terab of text and if you want to find proper information from seven terabytes of the text it be super expensive I'm telling you if you want to do that even one quy cost about like more than 101k doll we did the calculation okay but if you can leverage this into decentralized data processing Network and you give a query to us and it will help you to cury all like basically we do AI match right to find out all this kind of resume data for you and if you have for example more than tens of thousands of devices running that for you it cost less than $10 about like two hours to finish the query and you found all the resume data you want and you can use that resume data either to build a AI agent bot or use that kind of data to to to build application directly for build a rack and share that to some other applications that's a very concrete example right so you got the data source but like you need this kind of very efficient powerful data processing Network to help you to cure and find out proper data for you to use and just to be frank that Mizu this kind of model Mizu is using now is not that accurate but it's like extremely cheap so even like you know you can fill out like 20 gigaby of data among seven terabytes of data and then you can find another like you know have Downstream pliance to keep curing that that still like you know makes the whole data set analysis being being like completely scalable and affordable yeah yeah no really appreciate that and Jonathan I'll jump to you and then Whit and Don I'll come to you guys next um obviously you can't leverage foundational models or data without compute and you all have scaled an incredible GPU Marketplace curious can you tell us a little bit more about how you've balanced the supply and demand side of that Marketplace knowing that folks are just trying to get their hands on these b100s b200s right now uh flocking to these gpus tell us a little about how you've balanced that as you've scaled particularly in a decentralized manner yeah so for us uh like I said earlier we Leverage The Supply coming from institutional grade data sensors right so so these are T3 T4 data sensors the the quality that you get at AWS Google or or Microsoft uh and and the other tier 2 uh Cloud providers so what that means is when we see that the market is uh having trouble accessing gpus uh high quality gpus uh from our end because we have so much Partners we have over a hundred Partners providing data gr gpus we see an abundance an overabundance uh in fact so the supply from our side uh hasn't been a problem so far right like we have way too many gpus uh uh that that's accessible to the market uh but right now it's uh the demand uh right so uh as far as demand we still aren't seeing the traditional web 2 companies actually believe in uh what we're building right so so they say okay it's much easier uh to pay AWS but then you pay at at a cost right so so for what exhibits is offering uh just to put it out there we're about 40 to 70% cheaper than AWS uh and we sort of emulate the same quality that they do right so what Enterprise companies are looking for is quality of gpus uh uptime uh ensure that it's always running up and running and also the the lack of fault tolerance right so so uh it's it's inevitable that a GPU can break down at any moment and then when that happens uh models usually have to restart uh but we've built a bunch of uh middleware on top of that to ensure that okay if it stops there's a checkpoint that that we can put it up back online as ASAP uh so yeah uh I think the it's more of a narrative problem uh getting Enterprises and and other startups believe that there's an alternative uh to these large incumbents and Don and Whitney just to build off of that obviously you all benefit from uh all the Innovation happening at these kind of core infrastructure layers and maybe don all start with you building what I would consider a very Frontier use case in agent agent payments um how have you educated the market on something that doesn't necessarily yet exist um and tell us a little bit more about how you see the future of agent communication happening sure yeah so I mean it sort of exists right now in fits and spurts yeah so obviously there's Pockets usually regionally dependent where you know there's uh uh a strong foundation for developing out this type of Technology San Francisco there's pockets in China you know these are the dominant places where from an agent AI point of view development is actually taking place um but we also see it within genp right a familiarity with what we're doing um and so the question then then becomes well how do you tap into that familiarity I think broadly speaking from our Arena our domain space especially this web 3 AI cross-section we do a disservice most of the time by kind of obfuscating that with our the language that we use right like aii AI finance things like this that are completely foreign and maybe objectionable to most people you know I would argue that like Finance isn't that complicated it's just been off fiscated Away by the language you know people talk about complex derivatives they're not actually that complex when you dig under the hood but you know financiers use a particular language to aisc what they're doing we do that too in our business and so what we've been very conscious about is that abstraction that we create because of the way that we speak and trying to use language that's more familiar right so when when I say we're defining a category I believe we're defining a category but what is that category well it has to be a category that people understand yeah and so we're gravitating towards language like AI Commerce right um in in with our mindset it's U an extension of e-commerce right that most people today now understand and what we're trying to do is just extrapolate this new condition or new agent within this digital economic environment which are AIS right so combine the two AI Commerce holy oh yeah I've got an AI I or I see these things that are starting to emerge they're going to want to buy and sell stuff on the internet like I do AI Commerce I get it right so it's this type of language that I think we need to make a shift towards so that we can really um you know Drive adoption through familiarity right um and then then the question is okay what's the infrastructure that's available in order to enable all of this you know we fundamentally believe that one of the friction points is on the payment site across the board for all of the assets that we're trying to enable you know in particular from uh you know data sets for training as well as the infrastructure that's being made available from a compute point of view um data sets for context so that you can provide relevant inferences against specific use cases but having a mechanism to actually you know uh commercial ize and potentially commoditize those assets as well requires a payment system that you know looks to standardize at least certain um communication and transactional protocols within that ecosystem so that's where our specific Focus lies I know I just went against everything I talked about in the first part of my answer talking a bunch of mumbo jumbo but it's really about you know making sure that these agents understand how they're communicating with one another and then they can buy and sell the assets or services that they're making available yeah and maybe just to build off that exactly what you said with to to go to you in terms of relatability I think building a consumer application incredible opportunity in the age of AI but also can be incredibly challenging sure and so can you tell us a little bit more about a how you've reached consumers to date with what you're building particularly around image generation but also as you're under the hood building a really compelling set of infrastructure as well to actually enable Creator monetization at scale can you tell us a little bit more about how ring fence Works um and unpack some of the components of your architecture sure so when we think about acquiring customers um it is what it is with web 3 I I would say that there are uh there's more of a want for customers and users than there are actual customers and users out there for Consumer applications um generally speaking the the web 2 applications that exist in the world have not yet been surpassed by what what people are building in web 3 fortunately with creators uh we had the nft wave last cycle so they understand the value of digital art and then with this you know wave of AI That's come along and uh the you know the broad scraping of the internet and the court cases that popped up regarding the the theft of uh people's original work it made our go to market much easier so we were able to partner with Sony Alpha Collective to tap into photographers who have millions upon millions of edited photos we were able to tap into the the web3 uh digital artists who you know we're all familiar with because we've lost a bunch of money on their nfts um and and leverage them to come on of the platform first right because now we have this place where they can essentially use us as their free unlimited Cloud right they can upload all of their work they can uh move it on chain they can attach metadata to it so it's provably theirs uh and they can restrict the the use of of AI uh to train with those those images of theirs so that gave us some initial push to get those users on to eer and when we think about um when we think about the the API ring fence really what it is at the core and helping people monetize their data this is something where we are just now starting to educate the market um but you know whether it's good or bad people respond most directly when they're hit in their pocketbook and that's how we think of it right so as we start to talk uh towards the API product and how businesses can integrate it how you know when our our consumer app comes out users will be able to benefit from it we're not going to focus on the the tech word salad we're not going to focus on you know how AI will use their data you know on a technical level we're going to put it in an adjustable way that helps them understand that you know today you have zero doar from these activities tomorrow you can have X amount of dollars from these activities um do you want more money or would you prefer to stay where you're at financially right and that that's that's exactly where we're at we're not looking to tie the financial performance of our project to a token we're looking to tie the financial performance of our project to how much money we can put in the pockets of the users that we're bringing on and we think that that is really what's compelling about web 3 and about about the product that we're building I think if all of us were you know to take a step back and look at why we got into crypto or web 3 in the first place yes the tech is great and it's phenomenal but we all suck around because there was something that led us to uh lifechanging realization that if we put our life's work or our life's effort into this industry it was going to potentially change the ability to create wealth for our family or you know the generations that come and follow on behind us and we want to be able to provide that opportunity to the users that come on to ring fence whether it's B2B users or consumers that we're going to be targeting here in the next 3 to six months so I think each of the solutions that you're building at different layers of the stack are so powerful but each kind of compound together in in our final minute or so here obviously uh there is a lot of innovation happening in web 2 AI uh but I would argue there is an immense amount of innovation happening in the world of decentralized AI and so maybe Shu we can start with you thing you are most excited about for 2025 whether it relates to specifically what you're building in kind of the data processing piece or more in general I will say I I really like looking forward to the Boost of a applications because now I I'm using some of the applications today I'm using cursor to help me write code I think a lot of like programmers are using that but like the usage of that is still like quite limited maybe because of the cost so decentralized AI maybe can bring down the cost but like I really want to see the consequences here basically like how the how it can really like help to motivate and incentivize people to create more applications and can I can use tens of different applications in my daily life that's I'm looking looking forward to yeah yeah we rapid fire go ahead with yeah so I I think the most exciting about the next 12 months is going to be the massive amount of attention that's going to come to the space and what we're going to be able to do with that as a business you know when when you think about getting eyeballs that's the hardest part of a startup right and now essentially we're going to have a market moving in a Direction that's going to capture all those eyeballs for us and now we just have to get up and put up and and show what we've been building for the last year and a half is actually worth somebody downloading and putting on their phone um I'm pretty excited about where we are right now like I've been doing this for so long and people have looked at me like I'm crazy for long enough that you know we filled a panel and everybody's focused on AI and web 3 it's pretty awesome um so I I just want to you know I think the the continuity on this it's you know it's going to continue to explode more specifically I'm looking forward to AI agent New Year uh you know more like uh the the the truth terminals coming coming live and getting paid let's do it let's do it yeah for me uh it's it's like what Don said um AI agents I'm super excited on that uh just because what you're seeing on chat GPT when they opened up like their their their uh API to for anyone to create different subsets of chat GPT right like chat GPT for Bangkok chat GPD for doing uh uh anything you want right there there's there's a solution for that um and for exhibits what we're trying to do is enable anyone to innovate it by by uh focusing on the cheapest compute possible uh it's the market so anyone can build what they want uh with AI yeah so I definitely want to see more and more models more and more agents and uh also at the same time I want to see more and more builders in this AI space and if you are building anything related to Ai and I want to see more and more people applying for near Horizon it's been an amazing incubation program I think all of us have can pretty much just agree with uh with this um yeah this is the future I want to see let near be the AI hub off five 3 amazing well folks I have successfully gone over time uh and so thank you all for coming out today and Round of Applause for our founders","transcript_synced_at":"2026-11-03T05:20:28.430Z","segments":[{"start_seconds":5,"text":"all right well we have everyone here now great to see you all uh folks my name is Jared Barnes I am uh the head of founder","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=5s"},{"start_seconds":12,"text":"success at the New Foundation have the privilege and opportunity of working with all of our Founders uh and builders on near and uh I could not be more proud","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=12s"},{"start_seconds":20,"text":"about this panel and these uh amazing Founders on stage here 6 months ago uh we really doubled down on growing and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=20s"},{"start_seconds":27,"text":"scaling our AI ecosystem and as we were in that process of growing uh we said who are the best Founders in the space","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=27s"},{"start_seconds":35,"text":"that we can go find and work with uh and those are these folks sitting on stage and so we have teams building at all layers of uh this decentralized AI stack","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=35s"},{"start_seconds":44,"text":"that I'm really excited to to get into and really excited for you all to hear a little bit more um about each and every single one of them so Dylan I'm going to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=44s"},{"start_seconds":52,"text":"start with you and if we can go down the line uh just sharing a little bit about not only what you're building but your origin story to how for yourself and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=52s"},{"start_seconds":60,"text":"pond came about sure uh thank you Jared and thank you nice to meet you everyone so I'm uh I'm Dyan one of the co-founders at Pond and also we are one","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=60s"},{"start_seconds":68,"text":"of the incubes of near AI Horizon uh incubation Pro uh program so uh at Point what we are doing is we are developing","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=68s"},{"start_seconds":76,"text":"Crypton native models trained on onchain data um so for examp I can give you some examples because AI model seems to be a","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=76s"},{"start_seconds":84,"text":"little bit vague um so when we are saying saying that one token is about to pump uh based on um you know all the training","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=84s"},{"start_seconds":92,"text":"results that we have on the onchain behaviors of previous tokens what might happen and based on the behavioral prediction so we can say you maybe this","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=92s"},{"start_seconds":100,"text":"token has some potential in some ways so this is one of the use cases of unchain data and also unchain data models um so","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=100s"},{"start_seconds":108,"text":"we started from developing user search engine because we feel like it's very difficult to find who knows who and who knows who well and it seems like it's","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=108s"},{"start_seconds":116,"text":"very it's very hard to do such a thing in web to because all the data structures across different platforms are totally different like today I'm","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=116s"},{"start_seconds":123,"text":"using Facebook data but your LinkedIn profile also has some of the social information of your of yourself but there's no such a third party can do","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=123s"},{"start_seconds":132,"text":"this so we started from onchain data because all the financial data and social data all at one place and we developed graphical algorithm to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=132s"},{"start_seconds":139,"text":"calculate the you know distance between different nodes which are different users and uh from calculating the distance we'll be able to find who knows","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=139s"},{"start_seconds":147,"text":"who and who knows who well to provide a such a user service uh user search service and later on we feel like since all the social data and financial data","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=147s"},{"start_seconds":155,"text":"are at the same place why don't we just train an industrial Scale Models so that's how we started this project um we are more like a model developer","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=155s"},{"start_seconds":164,"text":"developing Crypton native AI models to power all the Crypton native use cases right hey guys I'm Jonathan uh one of","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=164s"},{"start_seconds":172,"text":"the four core members of exhibits and we got our origin story from building having experience in building over a 100 data sensors in the in our past lives","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=172s"},{"start_seconds":181,"text":"and now we are leveraging that to produce uh compute uh for the decentralized compute Market uh and also","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=181s"},{"start_seconds":187,"text":"decentralized AI companies uh web 2 and web 3 gaming companies uh and and really focusing on gen AI companies uh who want","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=187s"},{"start_seconds":195,"text":"to uh just build uh with us so uh yeah so leveraging that uh data center","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=195s"},{"start_seconds":202,"text":"experience we decided to provide compute uh to everyone in the market um some of the things and compan we are working","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=202s"},{"start_seconds":210,"text":"with uh web 2 gen companies uh web 3 gen companies and also a lot of the cohort members inside uh near","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=210s"},{"start_seconds":221,"text":"Foundation all right I guess it's me first off uh thanks for having me near I'm Dawn I'm the co-founder and CEO of","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=221s"},{"start_seconds":229,"text":"nevermind so my origin story my parents met in 1968 oh no I'm just kidding um so I've","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=229s"},{"start_seconds":237,"text":"been focused on the ml and AI space for going on 20 years now studied engineering at University and then went","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=237s"},{"start_seconds":245,"text":"into Commodities trading um on the risk side doing back then what was called statistical modeling which then became","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=245s"},{"start_seconds":253,"text":"machine learning and now has been co-opted by the AI brand um did that for a few years and then parlayed that into","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=253s"},{"start_seconds":261,"text":"a career for the better part of a decade and a half in IT consulting as a subject matter expert in data and analytics um that had me TR travel all","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=261s"},{"start_seconds":270,"text":"over the world building large scale data States for ML purposes for some of the biggest companies on the planet HSBC","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=270s"},{"start_seconds":276,"text":"L'Oreal AXA mizuo yada yada um caught the Crypt bug in","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=276s"},{"start_seconds":283,"text":"2016 uh went down that rabbit hole um was enamored with the um I think","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=283s"},{"start_seconds":293,"text":"sophistication um and elegance of using a blockchain for providential integrity purposes um so in 2017 I co-founded a","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=293s"},{"start_seconds":303,"text":"project called ocean protocol which was one of the first attempts to bring Ai and blockchain together so I've been","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=303s"},{"start_seconds":310,"text":"doing this for the better part of a decade now before it was ever cool um a lot of people used to look at me side","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=310s"},{"start_seconds":318,"text":"eyed like what the are you doing um but yeah I've been in this space for a long time and where we've landed now is","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=318s"},{"start_seconds":325,"text":"with it we've really distilled basically a decade's worth of into a hyperfocus on AI payments um so","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=325s"},{"start_seconds":333,"text":"never mind we have a very specific view of the world we believe what we're witnessing right now is the rise of a new consumer that's going to manifest as","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=333s"},{"start_seconds":340,"text":"trillions of AI agents and that these agents are going to change Commerce forever um we believe that this is a category defining moment um the new","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=340s"},{"start_seconds":349,"text":"category being AI Commerce and in response we're building the PayPal of AI to be the engine behind AI Commerce and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=349s"},{"start_seconds":356,"text":"get these agents paid boom Jared thanks for uh for moderating and it's been an amazing experience working","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=356s"},{"start_seconds":363,"text":"with near Horizon uh in this previous incubator my name's Whitney I'm the co-founder and CEO of ring fence AI uh I've been in the space since","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=363s"},{"start_seconds":372,"text":"2017 and you know as a an observer at first and uh then a Community member for","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=372s"},{"start_seconds":378,"text":"a lot of now defunct projects um but I you know I really found my passion in mining at first and over the course of a","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=378s"},{"start_seconds":386,"text":"few years I built a a pretty successful mining podcast called the hash rate podcast which is now the mining podcast","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=386s"},{"start_seconds":393,"text":"um still in operation today and uh built a company called Compass mining from you know zero to about 100 employees Compass","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=393s"},{"start_seconds":400,"text":"is now one of one of Bitcoin uh bitcoin's largest hash rate managers and we serve thousands and thousands of","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=400s"},{"start_seconds":407,"text":"retail Bitcoin miners um in 2022 I really caught the AI bug we were getting a lot of inbound from people who wanted","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=407s"},{"start_seconds":415,"text":"us to stand up GPU facilities for them and you know eth mining was was dying so got really curious as to why so I","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=415s"},{"start_seconds":423,"text":"started looking into where all of this uh compute power was was being directed and really caught the AI bug stepped","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=423s"},{"start_seconds":430,"text":"down as CEO of compass and focused all efforts on uh learning more about the AI space became a student of uh of the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=430s"},{"start_seconds":439,"text":"process of how to build and and scale one of these businesses uh developed a couple of different uh AI products that are now in operation mostly in the UAE","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=439s"},{"start_seconds":448,"text":"um and then in 23 started ring fence with a couple of uh a couple of friends and our goal with ring fence was to help people monetize their data simply put we","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=448s"},{"start_seconds":456,"text":"found that you know right now there's a tremendous push for AI but the barrier is going to be the the quality data that people are able to to gather and much","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=456s"},{"start_seconds":464,"text":"like the thesis of never mind that Don has uh we believe that there's going to be trillions and trillions of these agents that are out there in operation and each one of them are going to need","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=464s"},{"start_seconds":473,"text":"different types of data and most of that data is going to be unlocked by tapping into consumers uh which presents an","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=473s"},{"start_seconds":480,"text":"opportunity for these consumers to get paid which is really where we focus at ring fence so uh the first product that we built was geared towards creators and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=480s"},{"start_seconds":488,"text":"that is eer the eer app uh went live in September of this year we had over 140,000 creators that were excited to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=488s"},{"start_seconds":495,"text":"start using the platform so that's been in operation for a couple of months now and today actually if you go to ring fence. we've just released our API so","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=495s"},{"start_seconds":504,"text":"that anyone any platform any company has the ability to integrate ring fences data governance and monetization into","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=504s"},{"start_seconds":512,"text":"their own platform so you can collect protect and monetize your own users data and we believe that this is going to be the real unlock for Consumer data to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=512s"},{"start_seconds":520,"text":"help us break through the current data wall that a lot of companies are facing um while paying people for the data that they provide right because we've all been giving our data away for you know","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=520s"},{"start_seconds":529,"text":"as long as the internet has existed and now I think it's only right that people actually get paid for what they contribute so that's really where our Focus lies and you know it's been my","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=529s"},{"start_seconds":538,"text":"journey up to this point thanks yeah hi everyone uh my name is Michu uh you can see that my name is not","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=538s"},{"start_seconds":545,"text":"listed here because I almost cannot make it but like uh finally like I made it and then thanks shared to bring me back into the into the stage yeah I'm super","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=545s"},{"start_seconds":553,"text":"glad to to join the Horizon program and share my story so Mizu was originated by uh it's actually founded like last year","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=553s"},{"start_seconds":561,"text":"and uh uh like that very first we were thinking like what's the most problem like what's most fundamental like","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=561s"},{"start_seconds":568,"text":"problem want to solve for the space and uh the problem we see is that there's very few AI applications coming out we see like a lot of progress in the found","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=568s"},{"start_seconds":576,"text":"found Foundation model side the application ecosystem is still like not boosted and we look into that and see why and uh uh we look back into this","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=576s"},{"start_seconds":586,"text":"kind of uh Web Two like normal application side we see that the open source code really boosted the application like web applications or","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=586s"},{"start_seconds":594,"text":"whatever this kind of applications for this kind of space and for AI side we see the op this story most most of the data is actually multiplied valuable","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=594s"},{"start_seconds":602,"text":"data I say by Enterprises by this kind of foundation model companies that's why we decided to go to the direction that we want to do some open data stuff not","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=602s"},{"start_seconds":610,"text":"like private data and we believe that private data can also be open sourced if we can introduce some synthetics like technology here to anonymize it or by","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=610s"},{"start_seconds":619,"text":"extract this kind of pattern and so we can share the knowledge and share the structure out for other application","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=619s"},{"start_seconds":626,"text":"developers to to use so uh then like Mizu started to uh investigate in this kind of uh synthetic data Direction and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=626s"},{"start_seconds":635,"text":"also like the op General open source data Direction and uh then uh we realize that to make that happen like we really need a scalable Network and uh it","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=635s"},{"start_seconds":644,"text":"naturally leads us to this kind of deepin solution because uh I really like the saying that cloud service is for scale and uh deeping is for hypers scale","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=644s"},{"start_seconds":652,"text":"if we can Engage The efforts not only like the data but also computation resources across this kind of multiple uh like basically the user in the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=652s"},{"start_seconds":661,"text":"community from all over the world I think we can make like something really cool and very different we can accumulate the data accumulate this kind","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=661s"},{"start_seconds":668,"text":"of uh uh computation resources together to build a very hyper scale but like efficient and transparent and open","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=668s"},{"start_seconds":675,"text":"source open source data and share that to the like you know to to the to the uh bigger community and for the greater good that's like what we're building so","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=675s"},{"start_seconds":684,"text":"the very first product we're building is basically a DP Network to uh basically like we be a Min like tegram mini app","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=684s"},{"start_seconds":691,"text":"and a Chrome extension so people can run the code across mobile devices and laptops to help us to process data it's a decentralize the data search and data","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=691s"},{"start_seconds":699,"text":"processing Network to help people to fill out this kind of valuable data across a major like more giant data set and then like give people like valuable","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=699s"},{"start_seconds":708,"text":"insights that's what we're building yeah amazing so I'm excited to dig in a little bit more with kind of pointed questions for each of you and then we'll","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=708s"},{"start_seconds":715,"text":"zoom out talk about the market at a high level what's happening and then if we have time for some hot takes and predictions but maybe Dyan starting with","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=715s"},{"start_seconds":723,"text":"you right uh obviously super clear in the world of web 2 value Ural has gone to foundational models um I think certainly can see some parallels here in","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=723s"},{"start_seconds":731,"text":"web 3 uh but you've taken a different approach in terms of your model architecture and using graph neural networks instead of the Transformer architecture can you tell us a little","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=731s"},{"start_seconds":739,"text":"bit about some Transformers Inside has some Transformers Inside salute to Ed so okay okay salute to Ilia yes but for the most part graph no networks curious why","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=739s"},{"start_seconds":748,"text":"um what makes makes that different um and how does that actually show up when people are leveraging and using your model right so uh basically what a AI","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=748s"},{"start_seconds":756,"text":"model is really develop is U actually it has some prediction um predictive capability is really what the AI model is bringing to the world um and for","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=756s"},{"start_seconds":765,"text":"prediction it's really up to uh what type of data you are predicting so for we 2 apparently the majority of the data and the data structure is the language","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=765s"},{"start_seconds":773,"text":"data right that's reason why the foundation model of internet has to be a language model because it's where the value is from is the most of value is","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=773s"},{"start_seconds":781,"text":"from this uh you know language U data but in crypto in web 3 uh we don't have well we have language data mostly on","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=781s"},{"start_seconds":789,"text":"Twitter and you know different other other places but the most Crypton native data actually is really developed onchain data then what is onchain data","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=789s"},{"start_seconds":797,"text":"first of all it's not language it's a it's a type of graphical data so it might sound a little bit but just forget about is is a jargon so what a graph","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=797s"},{"start_seconds":805,"text":"data is it's really about as long as there's a data stru structure that can be represented by two components nodes","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=805s"},{"start_seconds":812,"text":"and edges so nodes in a blockchain context can be wets and can be contracts and the edges are pretty much just interaction between different nodes like","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=812s"},{"start_seconds":821,"text":"transaction you know it's mean to proof and transfer and all that so this a type of graph U data um when we're talking","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=821s"},{"start_seconds":828,"text":"about the uh the blockchain onchain data so just because the the nature of the data is different which means that if you want to develop AI model um the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=828s"},{"start_seconds":836,"text":"types of the model must be different so that's the reason why we started with developing um you know graphical models and one of the most outstanding model","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=836s"},{"start_seconds":844,"text":"architecture of graph data of graph models is the graph neuro Network um so that's what we are doing yeah amazing and Shu I'm actually going to jump back","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=844s"},{"start_seconds":852,"text":"to you kind of building off everything doing just shared on the actual foundational model side you really planting a flag uh in the data pipeline","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=852s"},{"start_seconds":860,"text":"obviously you've built a really powerful Data curation Network can you expand a little bit more assume folks know nothing about a data pipeline why is","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=860s"},{"start_seconds":869,"text":"what you're building important and why is it impactful and tell us a little bit more about uh how that actually shows up when we're talking about the application","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=869s"},{"start_seconds":877,"text":"ER I can give a very concrete example actually uh for example like uh uh for foundation model if you do pre- training you don't really care about the data","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=877s"},{"start_seconds":885,"text":"quality not not like you don't really need to curate the data so well you still care about data quality for example the token distribution or","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=885s"},{"start_seconds":892,"text":"something like that right so but like when it comes to like the foundation model cannot do everything right it's don't need agent agent and to build the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=892s"},{"start_seconds":899,"text":"agent you will need some customized data for some specific tasks that's called like vertical area you know you need","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=899s"},{"start_seconds":906,"text":"this kind of specialized data and this kind of data is really the barrier of this kind of applications it's very hard to get this kind of application for example I want to build a AI bot which","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=906s"},{"start_seconds":916,"text":"can help me to find a job in the market then it need to consume tons of resume data or tons of this kind of job market information and normally like as an","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=916s"},{"start_seconds":924,"text":"individual AI developer you don't have that right you don't know where to find this data set and there's no like real time data out there for you to use so","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=924s"},{"start_seconds":932,"text":"what we do is like uh you know Common crawl there's a there's a nonprofit organization called a common crawl and the common crawl craw the whole internet","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=932s"},{"start_seconds":941,"text":"every month and they took like published snapshot out for example like last month common crw cross about like 2.4 billion","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=941s"},{"start_seconds":947,"text":"web pages and including about 300 million domains okay there's tons of data out there but like people just","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=947s"},{"start_seconds":955,"text":"don't know how to parse how to understand this kind of data and if you want to use AI to process the so in total like it's about like 20 terab of","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=955s"},{"start_seconds":963,"text":"the data compressed about like 7 terab of text and if you want to find proper information from seven terabytes of the text it be super expensive I'm telling","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=963s"},{"start_seconds":972,"text":"you if you want to do that even one quy cost about like more than 101k doll we did the calculation okay but if you can leverage this into decentralized data","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=972s"},{"start_seconds":980,"text":"processing Network and you give a query to us and it will help you to cury all like basically we do AI match right to find out all this kind of resume data","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=980s"},{"start_seconds":988,"text":"for you and if you have for example more than tens of thousands of devices running that for you it cost less than","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=988s"},{"start_seconds":995,"text":"$10 about like two hours to finish the query and you found all the resume data you want and you can use that resume","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=995s"},{"start_seconds":1001,"text":"data either to build a AI agent bot or use that kind of data to to to build application directly for build a rack and share that to some other","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1001s"},{"start_seconds":1009,"text":"applications that's a very concrete example right so you got the data source but like you need this kind of very efficient powerful data processing","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1009s"},{"start_seconds":1016,"text":"Network to help you to cure and find out proper data for you to use and just to be frank that Mizu this kind of model","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1016s"},{"start_seconds":1024,"text":"Mizu is using now is not that accurate but it's like extremely cheap so even like you know you can fill out like 20 gigaby of data among seven terabytes of","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1024s"},{"start_seconds":1032,"text":"data and then you can find another like you know have Downstream pliance to keep curing that that still like you know makes the whole data set analysis being","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1032s"},{"start_seconds":1040,"text":"being like completely scalable and affordable yeah yeah no really appreciate that and Jonathan I'll jump to you and then Whit and Don I'll come","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1040s"},{"start_seconds":1048,"text":"to you guys next um obviously you can't leverage foundational models or data without compute and you all have scaled","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1048s"},{"start_seconds":1055,"text":"an incredible GPU Marketplace curious can you tell us a little bit more about how you've balanced the supply and demand side of that Marketplace knowing","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1055s"},{"start_seconds":1063,"text":"that folks are just trying to get their hands on these b100s b200s right now uh flocking to these gpus tell us a little","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1063s"},{"start_seconds":1070,"text":"about how you've balanced that as you've scaled particularly in a decentralized manner yeah so for us uh like I said","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1070s"},{"start_seconds":1078,"text":"earlier we Leverage The Supply coming from institutional grade data sensors right so so these are T3 T4 data sensors","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1078s"},{"start_seconds":1085,"text":"the the quality that you get at AWS Google or or Microsoft uh and and the other tier 2 uh Cloud providers so what","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1085s"},{"start_seconds":1093,"text":"that means is when we see that the market is uh having trouble accessing","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1093s"},{"start_seconds":1100,"text":"gpus uh high quality gpus uh from our end because we have so much Partners we have over a hundred Partners providing","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1100s"},{"start_seconds":1108,"text":"data gr gpus we see an abundance an overabundance uh in fact so the supply from our side uh hasn't been a problem","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1108s"},{"start_seconds":1116,"text":"so far right like we have way too many gpus uh uh that that's accessible to the market uh but right now it's uh the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1116s"},{"start_seconds":1124,"text":"demand uh right so uh as far as demand we still aren't seeing the traditional","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1124s"},{"start_seconds":1130,"text":"web 2 companies actually believe in uh what we're building right so so they say okay it's much easier","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1130s"},{"start_seconds":1139,"text":"uh to pay AWS but then you pay at at a cost right so so for what exhibits is offering uh just to put it out there","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1139s"},{"start_seconds":1148,"text":"we're about 40 to 70% cheaper than AWS uh and we sort of emulate the same quality that they do right so what","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1148s"},{"start_seconds":1156,"text":"Enterprise companies are looking for is quality of gpus uh uptime uh ensure that it's always running up and running and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1156s"},{"start_seconds":1164,"text":"also the the lack of fault tolerance right so so uh it's it's inevitable that a GPU can break down at any moment and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1164s"},{"start_seconds":1172,"text":"then when that happens uh models usually have to restart uh but we've built a bunch of uh middleware on top of that to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1172s"},{"start_seconds":1180,"text":"ensure that okay if it stops there's a checkpoint that that we can put it up back online as ASAP uh so yeah uh I","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1180s"},{"start_seconds":1188,"text":"think the it's more of a narrative problem uh getting Enterprises and and other startups believe that there's an alternative uh to these large","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1188s"},{"start_seconds":1197,"text":"incumbents and Don and Whitney just to build off of that obviously you all benefit from uh all the Innovation happening at these kind of core","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1197s"},{"start_seconds":1204,"text":"infrastructure layers and maybe don all start with you building what I would consider a very Frontier use case in agent agent payments um how have you","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1204s"},{"start_seconds":1213,"text":"educated the market on something that doesn't necessarily yet exist um and tell us a little bit more about how you see the future of agent communication","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1213s"},{"start_seconds":1222,"text":"happening sure yeah so I mean it sort of exists right now in fits and spurts yeah","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1222s"},{"start_seconds":1229,"text":"so obviously there's Pockets usually regionally dependent where you know there's uh uh a strong foundation for","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1229s"},{"start_seconds":1239,"text":"developing out this type of Technology San Francisco there's pockets in China you know these are the dominant places where from an agent AI point of view","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1239s"},{"start_seconds":1247,"text":"development is actually taking place um but we also see it within genp right a","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1247s"},{"start_seconds":1254,"text":"familiarity with what we're doing um and so the question then then becomes well how do you tap into that familiarity I","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1254s"},{"start_seconds":1262,"text":"think broadly speaking from our Arena our domain space especially this web 3 AI cross-section we do a disservice most","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1262s"},{"start_seconds":1270,"text":"of the time by kind of obfuscating that with our the language that we use right like aii AI finance things like this","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1270s"},{"start_seconds":1278,"text":"that are completely foreign and maybe objectionable to most people you know I would argue that like Finance isn't that","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1278s"},{"start_seconds":1287,"text":"complicated it's just been off fiscated Away by the language you know people talk about complex derivatives they're","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1287s"},{"start_seconds":1294,"text":"not actually that complex when you dig under the hood but you know financiers use a particular language to aisc what","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1294s"},{"start_seconds":1303,"text":"they're doing we do that too in our business and so what we've been very conscious about is that abstraction that","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1303s"},{"start_seconds":1311,"text":"we create because of the way that we speak and trying to use language that's more familiar right so when when I say","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1311s"},{"start_seconds":1318,"text":"we're defining a category I believe we're defining a category but what is that category well it has to be a category that people understand yeah and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1318s"},{"start_seconds":1326,"text":"so we're gravitating towards language like AI Commerce right um in in with our mindset it's U an extension of","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1326s"},{"start_seconds":1335,"text":"e-commerce right that most people today now understand and what we're trying to do is just extrapolate this new","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1335s"},{"start_seconds":1343,"text":"condition or new agent within this digital economic environment which are AIS right so combine the two AI Commerce","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1343s"},{"start_seconds":1351,"text":"holy oh yeah I've got an AI I or I see these things that are starting to emerge they're going to want to buy and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1351s"},{"start_seconds":1359,"text":"sell stuff on the internet like I do AI Commerce I get it right so it's this type of language that I think we need to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1359s"},{"start_seconds":1366,"text":"make a shift towards so that we can really um you know Drive adoption","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1366s"},{"start_seconds":1373,"text":"through familiarity right um and then then the question is okay what's the infrastructure that's available in order","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1373s"},{"start_seconds":1381,"text":"to enable all of this you know we fundamentally believe that one of the friction points is on the payment site across the board for all of the assets","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1381s"},{"start_seconds":1389,"text":"that we're trying to enable you know in particular from uh you know data sets for training as well as the infrastructure that's being made","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1389s"},{"start_seconds":1396,"text":"available from a compute point of view um data sets for context so that you can provide relevant inferences against","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1396s"},{"start_seconds":1403,"text":"specific use cases but having a mechanism to actually you know uh commercial ize and potentially","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1403s"},{"start_seconds":1410,"text":"commoditize those assets as well requires a payment system that you know","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1410s"},{"start_seconds":1418,"text":"looks to standardize at least certain um communication and transactional protocols within that ecosystem so","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1418s"},{"start_seconds":1426,"text":"that's where our specific Focus lies I know I just went against everything I talked about in the first part of my answer talking a bunch of mumbo jumbo","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1426s"},{"start_seconds":1435,"text":"but it's really about you know making sure that these agents understand how they're communicating with one another and then they can buy and sell the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1435s"},{"start_seconds":1443,"text":"assets or services that they're making available yeah and maybe just to build off that exactly what you said with to to go to you in terms of relatability I","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1443s"},{"start_seconds":1451,"text":"think building a consumer application incredible opportunity in the age of AI but also can be incredibly challenging sure and so can you tell us a little bit","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1451s"},{"start_seconds":1458,"text":"more about a how you've reached consumers to date with what you're building particularly around image generation but also as you're under the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1458s"},{"start_seconds":1466,"text":"hood building a really compelling set of infrastructure as well to actually enable Creator monetization at scale can you tell us a little bit more about how","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1466s"},{"start_seconds":1473,"text":"ring fence Works um and unpack some of the components of your architecture sure so when we think about acquiring customers","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1473s"},{"start_seconds":1482,"text":"um it is what it is with web 3 I I would say that there are uh there's more of a","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1482s"},{"start_seconds":1489,"text":"want for customers and users than there are actual customers and users out there for Consumer applications um generally","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1489s"},{"start_seconds":1497,"text":"speaking the the web 2 applications that exist in the world have not yet been surpassed by what what people are building in web 3 fortunately with","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1497s"},{"start_seconds":1506,"text":"creators uh we had the nft wave last cycle so they understand the value of digital art and then with this you know","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1506s"},{"start_seconds":1513,"text":"wave of AI That's come along and uh the you know the broad scraping of the internet and the court cases that popped","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1513s"},{"start_seconds":1520,"text":"up regarding the the theft of uh people's original work it made our go to market much easier so we were able to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1520s"},{"start_seconds":1527,"text":"partner with Sony Alpha Collective to tap into photographers who have millions upon millions of edited photos we were","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1527s"},{"start_seconds":1534,"text":"able to tap into the the web3 uh digital artists who you know we're all familiar with because we've lost a bunch of money on their","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1534s"},{"start_seconds":1541,"text":"nfts um and and leverage them to come on of the platform first right because now we have this place where they can essentially use us as their free","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1541s"},{"start_seconds":1549,"text":"unlimited Cloud right they can upload all of their work they can uh move it on chain they can attach metadata to it so it's provably theirs uh and they can","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1549s"},{"start_seconds":1557,"text":"restrict the the use of of AI uh to train with those those images of theirs so that gave us some initial push to get those","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1557s"},{"start_seconds":1566,"text":"users on to eer and when we think about um when we think about the the API ring fence really what it is at the core and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1566s"},{"start_seconds":1573,"text":"helping people monetize their data this is something where we are just now starting to educate the market um but you know whether it's good or bad people","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1573s"},{"start_seconds":1582,"text":"respond most directly when they're hit in their pocketbook and that's how we think of it right so as we start to talk uh towards the API product and how","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1582s"},{"start_seconds":1590,"text":"businesses can integrate it how you know when our our consumer app comes out users will be able to benefit from it we're not going to focus on the the tech","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1590s"},{"start_seconds":1598,"text":"word salad we're not going to focus on you know how AI will use their data you know on a technical level we're going to put it in an adjustable way that helps","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1598s"},{"start_seconds":1606,"text":"them understand that you know today you have zero doar from these activities tomorrow you can have X amount of dollars from these activities um do you","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1606s"},{"start_seconds":1614,"text":"want more money or would you prefer to stay where you're at financially right and that that's that's exactly where we're at we're not looking to tie the financial performance of our project to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1614s"},{"start_seconds":1623,"text":"a token we're looking to tie the financial performance of our project to how much money we can put in the pockets of the users that we're bringing on and we think that that is really what's","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1623s"},{"start_seconds":1631,"text":"compelling about web 3 and about about the product that we're building I think if all of us were you know to take a step back and look at why we got into","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1631s"},{"start_seconds":1639,"text":"crypto or web 3 in the first place yes the tech is great and it's phenomenal but we all suck around because there was something that led us to uh lifechanging","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1639s"},{"start_seconds":1648,"text":"realization that if we put our life's work or our life's effort into this industry it was going to potentially change the ability to create wealth for","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1648s"},{"start_seconds":1657,"text":"our family or you know the generations that come and follow on behind us and we want to be able to provide that opportunity to the users that come on to ring fence whether it's B2B users or","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1657s"},{"start_seconds":1666,"text":"consumers that we're going to be targeting here in the next 3 to six months so I think each of the solutions that you're building at different layers of the stack are so powerful but each","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1666s"},{"start_seconds":1674,"text":"kind of compound together in in our final minute or so here obviously uh there is a lot of innovation happening","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1674s"},{"start_seconds":1681,"text":"in web 2 AI uh but I would argue there is an immense amount of innovation happening in the world of decentralized AI and so maybe Shu we can start with","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1681s"},{"start_seconds":1689,"text":"you thing you are most excited about for 2025 whether it relates to specifically what you're building in kind of the data","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1689s"},{"start_seconds":1696,"text":"processing piece or more in general I will say I I really like looking forward to the Boost of a applications because","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1696s"},{"start_seconds":1705,"text":"now I I'm using some of the applications today I'm using cursor to help me write code I think a lot of like programmers are using that but like the usage of","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1705s"},{"start_seconds":1713,"text":"that is still like quite limited maybe because of the cost so decentralized AI maybe can bring down the cost but like I really want to see the consequences here","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1713s"},{"start_seconds":1720,"text":"basically like how the how it can really like help to motivate and incentivize people to create more applications and can I can use tens of different","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1720s"},{"start_seconds":1729,"text":"applications in my daily life that's I'm looking looking forward to yeah yeah we rapid fire go ahead with yeah so I I","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1729s"},{"start_seconds":1737,"text":"think the most exciting about the next 12 months is going to be the massive amount of attention that's going to come to the space and what we're going to be able to do with that as a business you","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1737s"},{"start_seconds":1744,"text":"know when when you think about getting eyeballs that's the hardest part of a startup right and now essentially we're going to have a market moving in a Direction that's going to capture all those eyeballs for us and now we just","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1744s"},{"start_seconds":1753,"text":"have to get up and put up and and show what we've been building for the last year and a half is actually worth somebody downloading and putting on their","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1753s"},{"start_seconds":1760,"text":"phone um I'm pretty excited about where we are right now like I've been doing this for so long and people have looked","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1760s"},{"start_seconds":1767,"text":"at me like I'm crazy for long enough that you know we filled a panel and everybody's focused on AI and web 3 it's","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1767s"},{"start_seconds":1775,"text":"pretty awesome um so I I just want to you know I think the the continuity on this it's you know it's going to","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1775s"},{"start_seconds":1782,"text":"continue to explode more specifically I'm looking forward to AI agent New Year uh you know more like uh the the the","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1782s"},{"start_seconds":1791,"text":"truth terminals coming coming live and getting paid let's do it let's do it yeah for me uh it's it's like what Don","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1791s"},{"start_seconds":1798,"text":"said um AI agents I'm super excited on that uh just because what you're seeing on chat GPT when they opened up like","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1798s"},{"start_seconds":1805,"text":"their their their uh API to for anyone to create different subsets of chat GPT right like chat GPT for Bangkok chat GPD","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1805s"},{"start_seconds":1814,"text":"for doing uh uh anything you want right there there's there's a solution for that um and for exhibits what we're","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1814s"},{"start_seconds":1822,"text":"trying to do is enable anyone to innovate it by by uh focusing on the cheapest compute possible uh it's the market so anyone can build what they","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1822s"},{"start_seconds":1829,"text":"want uh with AI yeah so I definitely want to see more and more models more and more agents and uh also at the same time I want to see more and more","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1829s"},{"start_seconds":1837,"text":"builders in this AI space and if you are building anything related to Ai and I want to see more and more people applying for near Horizon it's been an","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1837s"},{"start_seconds":1846,"text":"amazing incubation program I think all of us have can pretty much just agree with uh with this um yeah this is the future I want to see let near be the AI","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1846s"},{"start_seconds":1854,"text":"hub off five 3 amazing well folks I have successfully gone over time uh and so thank you all for coming out today and","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1854s"},{"start_seconds":1862,"text":"Round of Applause for our founders","url":"https://www.youtube.com/watch?v=z124wIkumIA&t=1862s"}],"youtube_url":"https://youtube.com/watch?v=z124wIkumIA","url":"https://near-pedia.xyz/videos/z124wIkumIA"}