{"id":"ye_2k7dxtCE","title":"NEAR AI Office Hours #5 with Masa and Compute Labs","description":"Decentralizing AI: Exploring Data Sovereignty and Compute Power in Web3\n\nIllia Polosukhin & Alex Skidanov discuss the latest AI research topics and unpack some NEAR.AI focus areas\n\nNEAR AI: https://near.ai/\nIllia Polosukhin https://x.com/ilblackdragon\nAlex Skidanov https://x.com/AlexSkidanov\n\nMasa https://x.com/getmasafi\nCompute Labs: https://x.com/Compute_Labs\n\nThe future of decentralized AI! In this episode, industry leaders from Masa and Compute Labs share insights on the financialization of compute assets, the importance of decentralized data scraping, and the potential of user-owned AI models. Discover how these innovations could reshape the AI and Web3 landscape.\n\nTakeaways:\n• User-Owned AI Models: The shift towards user-owned AI emphasizes user success and financial well-being over company profit.\n• Decentralized Data Scraping: Projects like Masa enable access to valuable data from platforms like Twitter, Discord, and Telegram without high costs.\n• Financialization of Compute Assets: Compute Labs is pioneering the tokenization and financialization of compute resources, providing new investment opportunities.\n• AI Inference and Training: Decentralized networks offer promising solutions for batch inference jobs, synthetic data generation, and potentially training models with high latency.\n• Cross-Chain Interactions: The future of Web3 includes seamless interactions across multiple blockchains, enabling complex transactions and data integration.\n\nTimestamps:\n00:00:00 - Introduction: Musical intro and welcome.\n00:01:20 - Guest Introductions: Introduction of Albert from Compute Labs, and Brendan and Glenn from Masa.\n00:03:54 - Project Overviews: Overview of Compute Labs and Masa projects.\n00:06:13 - NEAR Protocol History: Discussion on the origins and evolution of NEAR Protocol.\n00:11:17 - AI Agents Explained: In-depth look at the concept of AI agents and their applications.\n00:15:07 - Data Scraping and Usage: Brendan explains Masa's approach to data scraping and its applications.\n00:19:18 - Financialization of Compute: Discussion on tokenizing compute resources and investment opportunities.\n00:24:15 - AI Inference and Training: Exploring decentralized AI inference and training capabilities.\n00:28:01 - Web3 Compute Marketplace: Comparing Web3 and Web2 compute demand and supply.\n00:37:16 - User-Owned AI and Economic Models: Benefits of user-owned AI and economic models in Web3.\n00:45:44 - Edge Intelligence and Quantized Models: Potential of edge intelligence and quantized models in decentralized AI.\n00:54:02 - Future of Decentralized AI: Philosophical discussion on the future of decentralized AI and economic incentives.\n00:56:18 - Cross-Chain Solutions: Exploring cross-chain interactions and their practical applications.\n01:00:02 - Closing Remarks: Summary and thank you to participants.\n\nJoin NEAR's ecosystem:\nWebsite: https://near.org/\nTwitter: https://x.com/NEARProtocol\nBlog: https://near.org/blog/","channel_name":"NEAR Protocol","published_at":"2024-06-11T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/ye_2k7dxtCE/hqdefault.jpg","transcript_text":"feel hack the world and make it tight blockchain flows near to one we're close to won't stop us now together we redefine the game with near we'll find the flame empowered Minds rise and sore use your own AI we explore hold the future hand in hand in this Tech driven Wonderland okay uh I feel like what one day this video will uh yeah lead to some interesting reaction but great to be here uh we have a new office hours with amazing mass and compute labs and obviously me and Alex joining in um let's do a quick round of intro uh Albert do want a start and then go to Brandon and com yeah yeah sure um hey guys my name is Alber I'm the founder and CEO of Compu Labs so for Compu Labs we're a r and infra project that we focus on the financializing AI specifically the compute assets and we want to establish the derivative market for compute and I'm currently based in La Pasadena and yeah uh great to be here today and send for an invite great to meet everyone uh Brendan I am the technical co-founder for Massa at Massa we're building a decentralized AI Network where people can contribute data we essentially allow anyone anywhere in the world Builders developers to access any data uh through our decentralized uh data scraping Network and distributed worker Network really excited to be here been in the space for around uh nine or 10 years since 2013 um and thrilled to meet everyone today hey everyone my name is clenia I'm the business co-founder at M Bren already told you everything you need to know about Masa so I'll probably share a fun fact so Masa in Japanese means truth and as a data company I feel like truth and data align quite well super excited to be here and thank you for the invite Alex Andia awesome well great to have both of your companies here uh it's actually really interesting because I think we talked about both of your ideas one way or another on previous office hours and so so uh uh and I've been I've been actually discussing with some folks that from from a compute perspective the real world asset approach makes so much sense because that's actually where money are made right now on uh you know offering compute at like a pretty major uh Premium versus compute marketplaces on their own have a very limited um kind of margin to capture because you know you can always switch and use any one of those and uh yeah combining kind of both makes a lot of sense and then on on the data side for sure kind of how do how do we leverage the fact that we can you know very easily with crypto align a lot of people to uh you know script the internet and bring the data especially the recent data new updates uh to be used for AI uh I think it you know this is a perfect application as well so maybe uh to turn around uh I would love to to hear what kind of questions you guys have and what what's on your mind uh and can and Bernie that we can discuss I'll probably start here I I'm I promise that no one asked me to ask this question um so I know the original story of near was near AI before it became near protocol we love to learn more about how you guys got started in 2017 and what brought you back to the AI strategy more recently yeah it it was actually quite relevant to to Masa because what we were doing is H we were building an AI developer back then quite quite similar to what we try to do right now but uh we needed a lot of data uh we were annotating like we wanted to have people um simplify uh problem statements or Draw sketches of user interfaces or uh there was a couple of interesting data sets we were trying to collect uh and it was very hard to pay them because uh people would usually people people who would do this kind of work would come from the geographies which are not necessarily the geographies that PayPal was supporting well uh and so it was all kind of Adventures right if you send money via PayPal and it says no then there's no support right if you want to use Western Union back in 2017 you have to physically go to the to the branch and pay 15% or something like this I don't remember now but it's probably better now but it was a complete disaster back then right and uh and there was one person who said hey guys you can pay me in Bitcoin and I was like sure that's that sounds great let me figure it out and and it's like we owed him like maybe $5 and the fee was was more than that uh but that's what put you know what made us thinking about uh figuring out how to make it work better and so data notation was always one of the interesting use cases which uh uh actually it's interesting because I was I was uh Googling a little bit around with Masa and it's not just data notation it's generally data contribution right that's something AIA was uh uh was promoting for quite a while now right so uh like you know if I'm browsing Twitter uh I actually don't know what is the license behind the tweets uh when when I tweet uh I I I never read my terms of services even though I said I did uh so I I don't know if uh if it's uh within the legal boundaries of what I can do uh to contribute those tweets back to the network when I'm browsing Twitter but if it is right then it makes sense for me I mean if I was getting a little bit of money for every tweet I I I Doom scrolled I would be quite Rich right now so it's it sounds like yeah yeah so it sounds similar to sort of samasource Mechanical Turk um those data labeling Services yeah they're they're really critical right um I have a question though around that sort of when you're thinking of the uh data ecosystem as a whole and sort of how essential it is in web 3 AI or AI in general where do you see data in the stack like I think the way that we see it Massa I I likened this to metamask the other day to someone where in every AI application you need data and in every web3 application you need something like metamask and metamask really kind of got it start by being consensuses uh you know uh C project that blew up on ethereum still is the biggest wallet today what are those key components of the AI stack that you see as like absolute must haves like we we have like strong conviction that data is obviously compute like these are all intuitive things that you can connect with easily what else is there in the space that you see as critical and and what make up those building blocks to to give birth to a flourishing AI economy you know like you're mentioning here like data labeling you know data labeling is critical like machine learning and computer vision if you're not labeling images correctly computer vision just doesn't work and Samos Source Mechanical Turk really got off the ground in that way so so what are those building blocks would love to hear your kind of opinion and and thoughts on on where that's going yeah yeah you named a few already I'm trying to if I can quickly pull it up I've made this uh I think like already over uh six months ago uh it's just kind of like a pretty basic I don't know if you can see it here um web 3 AI stack and yeah data is kind of at the bottom feeding into everything else uh and it has content kind of reputation data curation crowdsourcing and personal data um so data curation actually had that yeah like what Masa is is kind of I call it the future of Creator economy because actually if you if you think of it after the first step of your bootstrapping where people are contributing tweets to Twitter and then then somebody else puts them to your system you can imagine that the eventual state that people just putting it to your system directly right everybody else consumes it through AI from yeah almost like an archival history of of all of that activity yeah yeah I can dig in later more yeah for journalism for music for video yeah and then you have uh obviously Al Compu and inference and training fincial models on M Compu as well and then you have like various applications and mean this is just like few topics but AI agents is obviously huge topic right now now a governance uh intelligent assets which is like kind of going from just programable assets and then this was old slide but you know kind of application building with AI as well which is what we're working on at neari yeah IIA I'd love to kind of pull on a thread that you say AI agent and like so many people talk about AI agents what does that mean to you like I think everyone has this kind of like either nebulous or very specific definition of what that means when you think of AI agents in this context like what are you thinking about when you say that it's interesting for individual agent it's probably combinate like it's it's some combination of uh you know let's just say code that has access to a language model right and so it's able to you know interact with with some environment potentially like that is shared with a user or users and agents and able to you know do actions in this and use language model to pretty much understand what other people and agents are doing and saying as well as what happens in the environment to like make it more concrete what we're working on on near AI developer side right we have an agent that talks to the user to understand what they're trying to build and writes a specification right so its environment is chat and specification and we have an as agent which is looks at specific and tries to build the you know smart contract and then still may go back to user and like make sure that some of the uh like tradeoffs are valid and so you can have like you know you can almost imagine a group chat with your de team uh where all of them are trying to figure out what you want to be built and then building it and kind of showing you their progress you know have a designers well join out um so but under these the way it works is you know all of the agent is just some python or whatever code that interacts with environment and has a language problem amazing amazing on near at the moment like what are the what are the best like inference uh infrastructure projects and LM sort of infrastructure projects that are currently on there right now where you can use that as your base agent sort of building block to then wrap it in Python to give you that user interface and are you seeing anyone using it for like agents not just for end user interaction but for kind of internal processing I'll give you an example we we scrape a lot of uh volatile web data so all of your tags on websites like hacken noon and these Pages like change a lot they're relatively ephemeral and one day you can have a scraper that passes data out that works and the next day your pipeline could break you've got to constantly monitor we've been um fine-tuning uh llama 3 to pass specifically this volatile data into structured Json so in a way it's like a it's an agent that's doing pipeline work um for the data itself can do a similar thing to tweets like adding metadata to tweets and doing that like second order layer of like processing or post scrape processing we're quite excited about like what that can do there's limitations like latency cost can be a problem tokens per minute throughput and things like that and I think there's a lot of work to be done there to get it to scale but quite exciting what it can do to data pipeline infrastructure um are you seeing any of that being built in near right now and would love to kind of um see if there's any other use cases for agents outside of like end user interaction that you're seeing growing and that's probably all of my questions for a bit otherwise I'm gonna hog the mic yeah I mean there's like few components here one is the in like kind of AI inference infrastructure which we can dive into but we are actually you know working with a few projects that are coming in right now into the ecosystem uh as well as working on something we call AI inference router which is a way for then applications to actually plug in into this and get routed to whoever provides inflence and compute um so solution yeah yeah showcase that as well as then it should serve as a plugin for data providers and other uh kind of VB 3 foundational models Etc who can actually plug in into one system so it's it's a little bit early but like uh we can we'll we'll talk about that more in in coming weeks yeah um have you used open routes is a great tool sounds similar to that yeah yeah I mean I think open routers solves a little bit different problem but yeah like we trying to think how how can we bring all of the projects that they're building right now around near in the I ecosystem how do we like actually com com allows them to compose with each other in you know in a way that Al then serves and user or end application uh like in easy way yeah I like sorry go on go on yeah uh I was just like um having a question for Brandon sort of um ma because um I know like scale AI pretty well and I assume you guys are sort of like doing what they're doing at in web stre right are guys like handling the like data collection or only the data labeling yeah it for us we've really focused on the data collection and scraping so um Alex sort of picked up on a an interesting point here about tweet data so we actually being very transparent we reverse engineered uh Twitter's front-end API the graphql API um as a Twitter worker on our Network you can bring your Twitter credentials you can add them to a node it will log you in save a session cookie on your um machine or your instance um and then we route through a worker layer an ACA framework requests ht2 requests from our API that get fulfilled by those workers there's an element of consensus around the the scrapes um so we've really focused on uh unlocking the Twitter data that's typically like not available through the Twitter API or if it is available you've got to spend an awfully an a like a huge amount of money like it's $42,000 a month just to be able to get a endpoint that gives you followers and right now we do it for free for everyone so you can get any data you want from Twitter completely for free right now um I call it the you Elon product because he completely screwed up the uh the Twitter API very tongue and cheek but it's it's memorable for everyone um and we then do web scrape data Discord Reddit basically anything you see on the public web you would get through a platform like app appify um you can get through uh Massa protocol for free um we're then building on the back end more of like the processing pipeline infrastructure where I have high conviction that instead of investing heavily in data pipelines that transform data using static code using llms to do that and the pace at which people like you are probably bringing compute to the market and other networks there's be so much latent unused compute out there that you can almost run background jobs on data sets um to provide this enrichment L on top so that's something Downstream we're adding to as we get this big Corpus of data that's kind of historical based on what people are scraping so we have two things we store the data that is scraped By Request there's some really interesting stuff there that you can see like what is most valuable kind of surfaces from that data because it's what people are querying the most and you can go deep there and then we're also going to run batch jobs to build data sets in our back end and store them on something like OG laabs or uh space and time something like that as sort of permanent storage so yeah focusing on the data scraping got you got you and like for the data what type of data do you guys support now like only pouch or also like video and audio at the moment yeah it's a good question um video and audio we cut out of scraper just because of the size of the data and storing it was becoming an issue um but it's it's a relatively easy change to add it back in um so we haven't really done much computer vision stuff we've been really focused primarily on the um on the text based uh data because most of the agents we're seeing like IIA was saying are user facing agents that are looking for text either to find tune or go into rag um or in certain cases where people have large budgets they want to trainer or retrain a base model um like Spectra have done this Spectra have retrained a base model on uh GitHub data um specific to Smart contracts so you have a really great uh agent that's able to give a lot of detail on Smart contract uh syntax as opposed to it sort of out of the box and from like a business perspective what we've been looking for and very luckily be able to get as one of the first decentralized AI data product out there is to find what we call data product Market fit so think of it as trying to reach equilibrium between the demand side and the supply side we need to basically learn iterate for our developers specifically AI application and AI agent developers to understand their use case and then we can rapidly ask our network of validators and node workers to spin up new data scrapers so far we find two use cases to be the most prevalent uh the first is uh Financial no surprise there um everyone is trying to use AI agent to capture trading signal Etc and then the second one is social we have seen a huge wave of AI companions they have boyfriends and they have girlfriends emerging right now loneliness is one of the biggest problem facing Millennial and genz right now and it is directly reflected in the use case so based on those two leading categories of use cases we started to develop different kind of data scrapers Social Web data are the most popular amongst them so that's what's informing road map at the moment understood yeah because I mean it's quite interesting because I used to work for a project called delum we a AI agent Network that we automate the entire like workflow for users in crypto for for instance like I think we recently launched a partnership with Galaxy that like because on Galaxy there a a lot of users trying to do like XYZ to farm for the air job or to attend the campaigns right and AI agent could be a really good way way for them to automate the process and just to complete all the stuff with one click and I'm just running are you're saying you pretty much in automating what people are paying for to do user engagement you're automating it with AI and replacing user engagement with AI yes that will be one of the use case like we explore because um so we started as a um a NPC solution for gaming but our tax that is around NLP right and how to interact from a virtual human being to to human uh and then we figure like with such a technology what can we do to automate the workflow um specifically in crypto and one use case we found it just you know like there are a lot of people manually doing xyd like SN their Twitter uh do some transfer and all that and we're essentially having this like Infinity canvas sort of thing for users to DIY cust cize their workflow and in the meantime to have the AI agent to like just I guess complete them with one clip which is super cool I guess but so for that type of dat agent do you foresee any Synergy between like your send them um yeah you know we we we've been through like a bit of a journey with some of this clarification um on our way to realizing that data for AI consumption was really where the best fit in the market is today like data is a pretty unsexy thing like no one typically unless you're a real nerd gets super excited about data um but DJing everybody loves DJing so our first kind of initial use case was uh verifying uh dgen behavior from social like twitter twitter followers is example like are you following XYZ account um yeah we're able to provide verification for that uh through the platform so an agent will be able to understand does Ilia follow Masa or does he follow a you know and get a response back and determine whether or not iler is qualified for that sort of event or something so I believe there are ways in which uh reasoning can come from llms using data of this type and sort of build more complex decision trees that that essentially unlock the use case that you're describing yeah I'm sure there is well I think it it there's even more right like you're bringing on for example Twitter data which you can use to make decisions on than other platforms uh on chain as well and so that's given like Twitter is probably the you know news feed of crypto you can actually have decisions made uh based on that so I think like that's kind of what we're thinking about like how do we plug in kind of the you know some of the data providers into as a context that then other uh you know be agents or users can use directly in their llms you know make it almost like you know set a mware that augments your query or your or your context as you makeing decision love that yeah love that sorry I was gonna say no surprise here right after um Twitter we actually devolve additional data scrapers supporting Discord and telegram data um for example if near users want to um I don't know capture trading signal of all the tokens available on near where they would go to first of all they would go check a few big whale accounts on Twitter secondly they will probably go to their favorite communities on telegram seeing what kind of signal calls are available on telegram that's exactly what we can provide as a data platform what we can do is really give you real-time structure uh Text data um based on telegram group chat history as well as discords so combined once you have the big three Discord telegram Twitter you almost get a pre-completed related to specific token trading yeah I just I just want to add something to what IIA was saying just now about this kind of what I'm almost hearing is like a reasoning framework so we got like inference and there's almost this like reasoning framework built on top of llms I think it's for a lot of people in the web three space there's this Vision it's almost autonomous Behavior within the web three stat constrained by a rule set or parameters that allow for certain actions to take place in a semi-autonomous or fully autonomous way um I think that this discourse and thought process leads to that outcome incrementally uh and opens up like a lot of interesting use cases for you know I'll give you an example we dog food our own Discord scraper internally where we will scrape user feedback user testing groups in Discord um and it will synthesize and output feature requests prioritize them for us add them into like a ticket backlog in GitHub not that we're trying to replace people in the organization but the velocity you now get in terms of taking raw data refining it and then making it actionable in our pipeline just increases like engineering and product disc velocity where you couldn't do this before so it's not just use cases where you're deing it's like how can I introduce both efficiency and observability into my ex existing processes so as I can extract more insights and imagine taking that one one step further you you're collecting information you're observing it and then you're doing reduction through reasoning to understand okay these are the outcomes I want to pursue because of XYZ heuristic you start building up like a very very powerful engine and like you know tie that together with economic incentives where you can program your incentive model on top of that layer you can really bring together those actors to make you know favorable outcomes I strongly strongly believe yeah makes sense I mean there's definitely a lot to develop here maybe Switching gears uh Albert I mean you you mentioned kind of the project you were before right but now with compute Labs you're working more on kind of the base level mization of compute and kind of bringing bringing forward the the value of the future I guess how are you guys seeing the kind of the opportunity with all the roll out right I mean like Amazon is building I don't know 25 new data centers um how do how do you see this on boarding looking and how do we uh how do we is crypto actually participate in that game more effectively well uh I guess like your second part how we can participate that's exactly what we're trying to achieve here or like enable investor to do uh especially like institutional investors and retails right because currently at the market there's no Financial product that provide a a direct exposure to asset to like compute as underlying asset there's Invidia stocks there AI related tokens and you might also invest in AI company from early stage but other than that like there's no direct exposure which is why we're thinking uh we want to focus on the financialization of this entire computer asset because we both know computer is going to be the currency of the future and when we're talking about the financialization or derivatives they're going to be like index ETF perpetu stating Landing uh sort of things and we feel like that market will be huge but of course the Foundation or the first step will be that we have to have this alching representation or like how we can transfer the ownership of that compute because physically it's just not feasible to transfer compute using just regular secondary Marketplace since it's heavy it's hard to be Shi and also easy to be broken which is why uh we feel like using blockchain technology to build this scap for compute as underlying asset makes a lot of sense and that's how we started and going back to like I guess AWS or more web tool Marketplace uh for compute versus web three um I guess I also want to talk to you about this is like currently I think the demand and Supply situation from both worlds uh have a complete different um situation right in web to the demand is much higher than the supply for the web to Cloud computer network however web free um I think like I don't that they just launched their token today right uh for those decentralized Compu Marketplace um I would say due to the physical limitation and all the network latency the actual demand there is uh is limited fairly limited and what's your opinion on like how you know the web stre components can develop longterm in a way that the demander will be more actual and like they're going to be a sustainable economy there yeah I I think I think there's a uh um the problem is that for the use cases that exist today that require a massive number of gpus you do need them to be with the uh infin band and whatnot right so it's one of the uh there are several scenarios where the demand will go up one of them is um presumably you can do inference uh without you know like like that's a viable viable scenario to do inference quickly on uh on machines that do not have that that quick of an interconnect uh especially uh it's one thing to have inference that you want you know like if you want to to be communicating with the uh you know in real time uh you probably want something with a very low latency high throughput but in many cases you want to do batch jobs then you don't care you know about Laten you care about trut so in this case I think the centralized networks already have an advantage uh and I think another one is I'm still very hopeful that there is going to be a breakthrough that will allow training uh with very high latency then of course uh uh the question will be can I de centralized Network provide gpus uh cheaper on a per flop basis right IGN ing the network uh and the bandwidth that is something I don't I don't know if it you know if it's viable I don't know like what's the comparison um between the price per flop on a decentralized network versus you know just getting h100s at like $2 a piece yeah uh I mean like uh 100% current limitation sort of just limit the usage for web Street components more to the AI inferencing for AI training due to the like you know infinite band and all that it's just not visible at the moment y yeah yeah we should invest more into yeah go ahead yeah no I just to mention like on where actually inferencing may be useful at a training time like a kind a background B inferencing is synthetic data generation right because uh I think we'll see a pretty sizable transition for training new models on on synthetic data that's specifically generated that's kind of uh from you know po smarter models distilled into smaller models um or just in general kind of generating distributions of data that are useful and for that you need massive batch of inference uh you don't care about patency at all and you just need like as much Compu as POS as as possible so I think that's probably one of the interesting use cases in the shter term uh if if we can coordinate all deves yeah it's something we've seen initial like green shoot interest on is this notion of like enrichment or synthetic data sets using not the Json model we have but a uh like an enrichment model where natural language conversations from things like YouTube Reddit um conversational uh text is being enriched by uh another protocol to create better conversational AI essentially doing what GPT 40 did to their most recent release right using all of the conversational data from all of our prompts to fine Chen the model so as it had what felt like a more natural uh Pros or vocabulary with what we speak you know the way we speak in in everyday life um and and they don't care about latency um they very much care about price and we we've recently worked on a coordination engine that allows you to imagine spawning a set of workers for an llm that scales horizontally um and allows you to retain the full sort of image of your llm instance but spread work across multiple devices with lower concerns for latency um so you end up using a single llm instance but you can run it over multip multiple workers that are multiple gpus um actually using we use lib peer-to-peer which is like the file coin system to spawn those workers with an ACO framework which is kind of interesting implementation bit like kubernetes for decentralized llms um and I think that we'll start to see better inference infrastructure better infrastructure for llms I don't think our is a final solution I think it's like an initial solution for this um that then allows for higher throughput batch processing and and we're looking to collaborate people like um aash would you know love to see if there's an opportunity to collaborate here to see how that scales and does it work for some of these early you know synthetic data set enrichment uh batch jobs yeah it's interesting 100% um and also I guess like another question for you guys is let's say currently there's a investment opportunity in front of you that can provide a stable 20 to 30 even 40% of yield per year and essentially you will be just putting money and purchasing a GPU yourself that will be custo custodian and also physically managed by computer apps or like by a professional team to generate yours right like to rent it out to AI companies from web to will you consider that as a related investment opportunity yourself as a as an ex Miner and spending years early on mining the minor dgen on me wants like a pool switching equivalent for gpus um you know we're investing in some some of our own Hardware at Massa to be able to run our own um compute just for R&D and to be able to have almost like a a mining pool switching engine where you're always mining the most profitable algo back in the day to basically provisioning your Hardware in the most profitable like use case um sounds interesting and like I'm speaking from someone that can technically set this Harbor up but you know uh collocated hosted miners became a massive thing they still are a massive thing you it feels like it's inevitable um that that's the case like we are going to allow scrapers to be loaded onto an Android device so you'll download an Android app you would just hit like a connect button enter any credentials and then it's almost like tethering to the network with your phone uh and you can essentially monetize the bandwidth on your device people want that really quick oneclick solution right to participate in this economy um yeah yeah yeah yeah I think both compute labs and Masa are trying to make sure that every user of the internet or of the internet will be able to financially benefit from AI right um like I'm probably the only non-technical founder on this call right now I don't know how to set up a hardware nor do I know how to run a very technical node but in the not so far future comput Labs will be able to allow me this opportunity to invest in like a piece of Hardware rented out and then you'll be able to manage it on my behalf and I can generate 20 to 40% yield that sounds pry amazing to me I'll probably only have one request which is to make the yield Dynamic if it's possible because given the speed yeah given the speed AI development I want to see yield not C at 40% yeah and we're actually also thinking about like developing this yield volatility so another derivative on the E itself and I we truly believe that there's going to be a lot of speculation room for that and just like what Brendan said um I think like same for for computer labs basically uh the end goal for us is also going forward uh any user who like can just go on their phone and like maybe go to their wallet and after a few CLS they can purchase the actual run GPU a hardware that's like installed in a physical world but they can have this nft serving as owner proof and also the Ying token right I guess like that's going to be the future we're aiming for here yeah I we're actually literally just brainstorming this uh at consensus a couple weeks ago um both Bren and I dabbled in the fintech world and in the fintech world there's a lot of wallets that gain massive adoption because they were pre-installed on Hardware or phones for example in India Reliance goo is such a conglomerate that promoted PTM amongst other wallets what if in the future every single like Android phone out there will be able to automatically scrape data contribute data plan where we'll be able to achieve one click or two like uh click uh renting a compute resource that would be pretty amazing and I think the future is actually not that far away so near actually has a application Kai chain which is pre-installed on over 8 billion phones already as pretty much on CI and opa phones and so that that is the part they actually are expanding this into so-called locally store data protocol which allows to kind of run AI on their device but they have a wallet pretty much for every user they have all of that integrated and so that is an opportunity to integrate into that system amazing we would love to integrate that sounds exactly what we have in invion and plan for pursuing that before Apple's announcement yesterday yeah speaking of that like what's your thought on the partnership between Apple and open I saw Elon was like on Twitter yesterday saying like the Privacy concern and all that yeah I mean I got I got strong thoughts I always do about these things yeah I have really strong thoughts um I mean it's it's it's all just uh deals through the back door right like I I think it comes down to I don't know I'm gonna pause I'm G let somebody else go first I'm not gonna go down to Rabbit Hole here because I will go down I will go down to Rabbit Hole but but in this particular case you are on the El site yes I am yes personally I am yes very much so yeah I I I mean I think we see an interesting convergence right and I mean everybody was expecting that announcement too that um although although a bunch of people were telling me like no open AI I mean sorry Apple has you know developed all this AI stuff but uh I mean and they do have some UND device things is built but at the end yeah it seems like they're just integrated um I mean I think realistically the operating system of the future will have ai pretty much as part of the experience layer and so the question is how is that done who is powered on and who is getting all the data right and so I mean that's why like I actually been talking about Edge intelligence um for over a year now kind of focused on how do we bring models on device how do we ensure that they can be as smart as possible leverage in your local data uh your kind of local context um and then potentially even becoming the way you engage with experiences and interfaces going forward right because if you can imagine this model starting to write code um or they can actually just build interface for any experience you want right maybe not on the fly but you can then share those experiences not as applications but actually as a recipe that you can change yourself and so I think like we see that is starting to happen except obviously as a l platform right there's a benefit of them kind of controlling it um and not also opening it up to other developers so uh and then I mean all the privacy concerns for sure will just keep accelerating from here yeah you know I think a lot of the Genesis for opinions yesterday is that we don't know right like the system can't be trusted right now and there's enough evidence to say that it can't be trusted trusted because of past behavior and I think it opens up a lot of philosophical questions as to we're always talking in web 3 whether it be like ZK proofs doing XYZ for like ml or lots and lots of solutions and how they can be applied but it's all well and good us talking about that within web 3 when we really need to have some of these bigger organizations adopting some of this and I just don't ever see that happening and it's because we as users don't put enough pressure and demand on these organizations to change their behavior and it's because the economics don't work so we have two options we go down the river with this kind of thing or we push back and we fight and I certainly want to be fighting I think this might be one of the first times where I reconsider getting an iPhone for the last eight years um much to my wife sugar in she'll see green messages from me not blue messages um and we have to build a different system like we we have to continue building these different systems and quantize models like is talking about for like Edge devices are very very interesting like there's a huge you know Bode of Open Source research that goes into quantise models that could do a lot I I haven't seen a huge amount of projects using quantise models in web 3 or am I missing those projects who out there is using quantized models instead of like llama 3 or another kind of hugging face model is there anyone oh those are quantied in many cases right yeah you you you can yeah you can but who who's who's specifically using those in in the web 3 space the really lightweight ones that have performance devices few projects I've seen who are using on on edge models and through that accessing a lot more data on a person's device I mean they're not yet like you know live but uh they're you know at at C stage pretty much developing their prototypes it's things I I I think emphasis like the emphasis that you made was that they do some things uh on device but I think they do majority of the things on device and and minority through open AI which is which is worth mentioning it is yeah and asks explicitly actually to go to right so yeah yeah it's different from hypothetical Microsoft where where it is you know uh where all your data is is just leaving your machine constantly yeah and and it's the same I I think it's still the same the same argument about iPhone versus Android that on Android by default you know like just like even going to the onboarding sequence where you know uh where every single control is designed in a way for you to give permission to send your data without understanding it I think for for quite a while still Apple will be unfortunately um the choice to make yeah for context Alex and I at some point broke Android well we worked around Android in such a way that people felt like they're approving one thing and actually approved a bunch of other permissions they haven't seen yeah that was not intended behavior and we filed it as an issue it took three years for them to fix it or four wow and I'm 5,000 richer wow that's that's that's that's crazy yeah I I've done I've done data collection from Android with consented permissions that give you very de detailed information on people to build credit models in the past yeah yeah yeah yeah uh and I guess just one more comment regarding like what Brandon just said I feel like the AI landscape is just becoming espe in web to it's becoming more and more oligarchy a little bit that like a few bit company they're controlling all the stuff it's prob now with a partnership between ab and open right yeah yeah I think that's part of the problem right you have to you have to use Android or an iPhone there's no way to live in 2024 without having one of the two right so it's a yeah yeah we we really see ourselves where the entire human race at the crossroads of central AI versus decentralized AI at this specific point in time the good thing is that we are building in parallel although right now centralized AI players have way too much resources compared to what we're building collectively but still we have the first opportunity in human history to be able to build in parallel with centralized players and hopefully surpass them at some point yeah we're definitely living exciting time I think that kind of for me shines this bright light on like what are we all doing here like it's not just the why decentralized AI which for many of us I think seems intuitive and obvious you know it's creating equal opportunity aligning incentives around economic models that you can program to bring the right players together to coordinate and act in a way that's good for the entire system not just one centralized player and I think that where we are in the cycle you know where Bitcoin found its fit for where it is now store of money store of value we need to collectively as a an entire ecosystem building in decentralized a like nail what that single very memorable value proposition is for decentralized that Ai and like what the why is you know ethereum did it with like the world computer even if we can judge with hindsight whether that came to pass or not um near is doing it with what you guys are doing with sort of more scalable uh layer one Solutions but like what is what is this particular one single thing that harmonizes the entire like AI web3 space um that connects with people to make people realize that this is where they should be um I don't know if I've seen that specific thing unless I describe it yeah I describe it as user owned AI because like what what the difference you want is it's not company you know trying to generate more Revenue right it's actually focused on the user and their success their needs their financial well-being and wellbeing more broadly and so that's kind of the flip we want to make um and it's not like decentralization there is the tool right decentralization always is the tool um but do people do people you speak to sucessful yeah do people you speak to outside of AI care about that I think I mean I think if you explain it hey like do you want something that makes you know big company more money or do you want something that makes you more successful uh I think that that seems to resonate um I think on on on its own maybe it's not like immediately used like obvious but so is so is internet computer let's be clear that most people don't absolutely the interet computer me either absolutely I really love I really love the framing of user owned AI by the way had U we have a new slogan we have yet to debut uh it's called Fair AI Power by the people which is exactly to your point right user own AI users should be the owner of their own data their own uh resources their own compute their own bandwidth as St financially benefit from the AI boom I think with anything else in web 3 decentralization and power of the community is always going to be the philosophical angle of all of our uh new Innovations yeah and I think I mean that's been a challenge actually I mean it's it's always been a challenge in web three that if you if you lead with philosophy the you lose a lot of people right it's it's a minority thing and so what is I I mean that's why I think the economics is so important right versus if you leave as economics like hey you can benefit in this way you can access something that you couldn't access right which is you know I'm coming from Ukraine access dollars is one of the very important things that uh everybody understands so I think it's you know we're trying to have this philosophical name but at the end it's going to be like hey here's the opportunity you can get right and here you know you want to benefit from the you know future of compute while compute Labs offering you something and it's like it's only possible to build on that three because you need to coordinate all of this desperate resources and efforts you know you want to benefit from you know your band with pretty much sitting at home uh in the background well you can use Massa so I think that's like at the end is going to be the actual driver for adoption and I I call it like every Revolution starts with economy and gets described with about values like you know America wanted freedom but actually they didn't want to pay taxes to UK right so like it started as economy and they were like yeah we want you know that they they plastered that's a good example that's a good example yeah yeah absolutely yeah I like that everything ends with the IRS in the US but 100% I deeply resonate with that um I grew up in Asia and even within the Masa Community we have now building our community in the past two years and a half um the most active community members actually come from Southeast Asia come from Nigeria come from the CIS region instead of America or Western Europe Etc it's because they see immediate value in what we're giving as token rewards to participate in our Network $1 is probably only one5 of my coffee this morning in New York but could mean a lot a lot more to a lot of people elsewhere yeah I often I often speak about that like the cost of anthropic or gpc4 at $20 a month that's still a huge hurdle for a lot of people and it's only that price because we're giving them their data we sorry we're giving those companies our data in the prompts if we weren't giving them that data it' be even more expensive for those subscriptions so we get access to a lower price because we're feeding the model as well kind of a slave to the system in a way um and I think that opportunity to give Fair access to free AI is something we can really do uh using these economic models especially to like prevent more inequality happening around this particularly like high moving fast moving space where you could very easily leave people behind yeah uh I feel like if uh from that perspective uh I think Masa is doing a really amazing job of because like if we're if we're thinking about it the more data those AI company they have the higher their technology barrier will be so that they can charge more for their subscription so we definitely need you guys to do to like basically decentralize all the data because otherwise it's just going to dep more and more yeah all data anywhere in the world for anyone that's cool love it yeah uh I just have one last question from from me and this is for mostly Alex and Alia I guess um so I have been talking to near team about the change sign nature recently and I really love the idea and it's quite exciting I think at least it fit the I mean our nft or our compute back nft really well the concept that where like no matter which you're from you can easily treat theet without even signing a signature that's super cool but I'm more curious about like hearing from you guys uh what do you think will be the most like applicable or Mak sense use cases for chain sign nature in crypto nowadays I mean there's a whole like already a bunch of people building on top of this Tech a whole set of different use cases um I mean it starts with wallets right the reality is a uh a number of wallets in your ecosystem already uh either deset or in case of hot on mainnet um provide that experience where you log in you get an near account but then you can go and use for example base and applications of Bas right and you know there is a mapped account soal remote account that you're controlling but you know you're interacting from your new account but as this goes further right you can now have Dows for example that can you know facilitate uh various efforts across chains you can have Smart contracts that are trading that are moving assets around we're also exploring this idea we call Omni token where it's pretty much one token that exists on every chain at once right and so uh pretty much the kind of root of it is on near it deploys and erc20 SPL you know even brc2 on bitcoin Mark sosio there uh with information and now you can transact there but it's the total Supply is kind of Interlink between all these chains and you and the importantly you can actually transfer between chains um kind of you know with finality of uh of that chain plus uh like 10 seconds or something and so you can actually now actually have that format where you know if you are on one chain I give you my base colum IL tr. e you can just send to that address not caring which chain you're on and so uh so I think like that's kind of the experience you know we're getting into and obviously there's dexes there's lending protocols there's people who are using it for for example issues of cards because they w't allow users to deposit assets on every chain and then you can pay with a card everywhere so it's really kind of you know like a new piece of tech new building block in the web 3 that really kind of ties the blockchains together and allows to have this kind of composability that I would say didn't exist before between different chains yeah I think like all the use cases you mentioned makes sense any AI specific use case or scenario of your top yeah I mean same for like let's say AI agents that want to interact across chains right um that's something that people are already exploring kind of I mean from payments to actual interactions to transaction right like right now if you're in a agenda you need to implement all the chains that you want to interact on you need rpcs you need gas fees management and all of them right versus just you know go through your new account and uh you know pay fees in one way gets managed by the underlying stack um same for you know as we kind of rolling out some of the inference endpoint ideally you'll be able to you know fetch all all data from all the related chains and then if you want to actually uh for example mbase has this natural language transaction process right where you just say like swap but also you know build me a portfolio of you know top tokens across all chains um and you know Buy $100 worth of it right you can imagine pretty much like a whole whole set of transactions across different chains that get spawned for that to happen so I mean it it is more of a like execution infrastructure not like you know that AI can plug in into that yeah 100% I really like the gas feature uh for chain sign nature that you will just automatically uh convert the gas token for super tool yeah well you kind of need that otherwise it's not very abstract if you need to know the implementation do the chain yeah definely all right I think we're almost at time so I want to thank you guys for coming uh for the discussion hopefully this was helpful uh I'm really excited kind of working with all of you and uh seeing some this opportunities for and users coming in in real life yeah thanks thank you so much for having us","transcript_synced_at":"2026-11-03T05:50:08.665Z","segments":[{"start_seconds":1,"text":"feel hack the world and make it","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1s"},{"start_seconds":9,"text":"tight blockchain","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=9s"},{"start_seconds":17,"text":"flows near to one we're close to won't stop us","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=17s"},{"start_seconds":24,"text":"now together we redefine the game with near we'll find the flame","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=24s"},{"start_seconds":32,"text":"empowered Minds rise and sore use your own AI we","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=32s"},{"start_seconds":45,"text":"explore hold the future hand in hand in this Tech driven","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=45s"},{"start_seconds":55,"text":"Wonderland okay uh I feel like what one day this video will uh yeah lead to some","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=55s"},{"start_seconds":63,"text":"interesting reaction but great to be here uh we have a new office hours with amazing mass and compute labs and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=63s"},{"start_seconds":72,"text":"obviously me and Alex joining in um let's do a quick round of intro uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=72s"},{"start_seconds":80,"text":"Albert do want a start and then go to Brandon and com yeah yeah sure um hey","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=80s"},{"start_seconds":87,"text":"guys my name is Alber I'm the founder and CEO of Compu Labs so for Compu Labs we're a r and infra project that we","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=87s"},{"start_seconds":94,"text":"focus on the financializing AI specifically the compute assets and we want to establish the derivative market","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=94s"},{"start_seconds":103,"text":"for compute and I'm currently based in La Pasadena and yeah uh great to be here","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=103s"},{"start_seconds":110,"text":"today and send for an invite great to meet everyone uh Brendan","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=110s"},{"start_seconds":118,"text":"I am the technical co-founder for Massa at Massa we're building a decentralized AI Network where people can contribute","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=118s"},{"start_seconds":125,"text":"data we essentially allow anyone anywhere in the world Builders developers to access any data uh through our decentralized uh data scraping","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=125s"},{"start_seconds":134,"text":"Network and distributed worker Network really excited to be here been in the space for around uh nine or 10 years since 2013 um and thrilled to meet","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=134s"},{"start_seconds":144,"text":"everyone today hey everyone my name is clenia I'm the business co-founder at M Bren already told you everything you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=144s"},{"start_seconds":151,"text":"need to know about Masa so I'll probably share a fun fact so Masa in Japanese means truth and as a data company I feel","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=151s"},{"start_seconds":159,"text":"like truth and data align quite well super excited to be here and thank you for the invite Alex","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=159s"},{"start_seconds":166,"text":"Andia awesome well great to have both of your companies here uh it's actually really interesting because I think we","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=166s"},{"start_seconds":175,"text":"talked about both of your ideas one way or another on previous office hours and so so uh uh and I've been I've been","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=175s"},{"start_seconds":183,"text":"actually discussing with some folks that from from a compute perspective the real world asset approach makes so much sense","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=183s"},{"start_seconds":191,"text":"because that's actually where money are made right now on uh you know offering compute at like a pretty major uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=191s"},{"start_seconds":199,"text":"Premium versus compute marketplaces on their own have a very limited um kind of margin to capture because you know you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=199s"},{"start_seconds":207,"text":"can always switch and use any one of those and uh yeah combining kind of both makes a lot of sense and then on on the data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=207s"},{"start_seconds":215,"text":"side for sure kind of how do how do we leverage the fact that we can you know very easily with crypto align a lot of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=215s"},{"start_seconds":224,"text":"people to uh you know script the internet and bring the data especially the recent data new updates uh to be","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=224s"},{"start_seconds":232,"text":"used for AI uh I think it you know this is a perfect application as well so","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=232s"},{"start_seconds":238,"text":"maybe uh to turn around uh I would love to to hear what kind of questions you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=238s"},{"start_seconds":245,"text":"guys have and what what's on your mind uh and can and Bernie that we can","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=245s"},{"start_seconds":255,"text":"discuss I'll probably start here I I'm I promise that no one asked me to ask this question um so I know the original story","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=255s"},{"start_seconds":264,"text":"of near was near AI before it became near protocol we love to learn more about how you guys got started in 2017","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=264s"},{"start_seconds":271,"text":"and what brought you back to the AI strategy more recently yeah it it was actually quite","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=271s"},{"start_seconds":278,"text":"relevant to to Masa because what we were doing is H we were building an AI developer back then quite quite similar","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=278s"},{"start_seconds":287,"text":"to what we try to do right now but uh we needed a lot of data uh we were annotating like we wanted to have","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=287s"},{"start_seconds":295,"text":"people um simplify uh problem statements or Draw sketches of user interfaces or","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=295s"},{"start_seconds":303,"text":"uh there was a couple of interesting data sets we were trying to collect uh and it was very hard to pay them","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=303s"},{"start_seconds":310,"text":"because uh people would usually people people who would do this kind of work would come from the geographies which are not necessarily the geographies that","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=310s"},{"start_seconds":318,"text":"PayPal was supporting well uh and so it was all kind of Adventures right if you send money via PayPal and it says no","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=318s"},{"start_seconds":327,"text":"then there's no support right if you want to use Western Union back in 2017 you have to physically go to the to the branch and pay 15% or something like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=327s"},{"start_seconds":335,"text":"this I don't remember now but it's probably better now but it was a complete disaster back then right and uh and there was one person who said hey","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=335s"},{"start_seconds":343,"text":"guys you can pay me in Bitcoin and I was like sure that's that sounds great let me figure it out and and it's like we owed him like maybe","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=343s"},{"start_seconds":350,"text":"$5 and the fee was was more than that uh but that's what put you know what made us thinking about uh figuring out how to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=350s"},{"start_seconds":359,"text":"make it work better and so data notation was always one of the interesting use cases which uh uh actually it's interesting because","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=359s"},{"start_seconds":368,"text":"I was I was uh Googling a little bit around with Masa and it's not just data notation it's generally data contribution right that's something AIA","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=368s"},{"start_seconds":376,"text":"was uh uh was promoting for quite a while now right so uh like you know if I'm browsing Twitter uh I actually don't","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=376s"},{"start_seconds":385,"text":"know what is the license behind the tweets uh when when I tweet uh I I I never read my terms of services even though I said I did uh so I I don't know","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=385s"},{"start_seconds":393,"text":"if uh if it's uh within the legal boundaries of what I can do uh to contribute those tweets back to the network when I'm browsing Twitter but if","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=393s"},{"start_seconds":402,"text":"it is right then it makes sense for me I mean if I was getting a little bit of money for every tweet I I I Doom","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=402s"},{"start_seconds":409,"text":"scrolled I would be quite Rich right now so it's it sounds","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=409s"},{"start_seconds":417,"text":"like yeah yeah so it sounds similar to sort of samasource Mechanical Turk um those data labeling Services yeah","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=417s"},{"start_seconds":426,"text":"they're they're really critical right um I have a question though around that sort of when you're thinking of the uh data ecosystem as a whole and sort of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=426s"},{"start_seconds":435,"text":"how essential it is in web 3 AI or AI in general where do you see data in the stack like I think the way that we see","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=435s"},{"start_seconds":442,"text":"it Massa I I likened this to metamask the other day to someone where in every AI application you need data and in","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=442s"},{"start_seconds":451,"text":"every web3 application you need something like metamask and metamask really kind of got it start by being consensuses uh you know uh C project","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=451s"},{"start_seconds":459,"text":"that blew up on ethereum still is the biggest wallet today what are those key components of the AI stack that you see as like absolute must haves like we we","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=459s"},{"start_seconds":469,"text":"have like strong conviction that data is obviously compute like these are all intuitive things that you can connect","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=469s"},{"start_seconds":478,"text":"with easily what else is there in the space that you see as critical and and what make up those building blocks to to give birth to a flourishing AI economy","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=478s"},{"start_seconds":486,"text":"you know like you're mentioning here like data labeling you know data labeling is critical like machine learning and computer vision if you're not labeling images correctly computer","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=486s"},{"start_seconds":494,"text":"vision just doesn't work and Samos Source Mechanical Turk really got off the ground in that way so so what are those building blocks would love to hear","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=494s"},{"start_seconds":501,"text":"your kind of opinion and and thoughts on on where that's going yeah","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=501s"},{"start_seconds":510,"text":"yeah you named a few already I'm trying to if I can quickly pull it up I've made this uh I think like already over uh six","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=510s"},{"start_seconds":520,"text":"months ago uh it's just kind of like a pretty basic","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=520s"},{"start_seconds":527,"text":"I don't know if you can see it here um web 3 AI stack and yeah data is kind of at the bottom feeding into","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=527s"},{"start_seconds":535,"text":"everything else uh and it has content kind of reputation data curation crowdsourcing and personal data um so","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=535s"},{"start_seconds":545,"text":"data curation actually had that yeah like what Masa is is kind of I call it the future of Creator economy because","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=545s"},{"start_seconds":552,"text":"actually if you if you think of it after the first step of your bootstrapping where people are contributing tweets to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=552s"},{"start_seconds":559,"text":"Twitter and then then somebody else puts them to your system you can imagine that the eventual state that people just putting it to your system directly right","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=559s"},{"start_seconds":568,"text":"everybody else consumes it through AI from yeah almost like an archival history of of all of that activity yeah","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=568s"},{"start_seconds":577,"text":"yeah I can dig in later more yeah for journalism for music for video yeah and then you have uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=577s"},{"start_seconds":585,"text":"obviously Al Compu and inference and training fincial models on M Compu as well and then you have like various","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=585s"},{"start_seconds":593,"text":"applications and mean this is just like few topics but AI agents is obviously huge topic right now now a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=593s"},{"start_seconds":601,"text":"governance uh intelligent assets which is like kind of going from just programable assets and then this was old slide but you know kind of application","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=601s"},{"start_seconds":610,"text":"building with AI as well which is what we're working on at neari yeah IIA I'd love to kind of pull","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=610s"},{"start_seconds":617,"text":"on a thread that you say AI agent and like so many people talk about AI agents what does that mean to you like I think","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=617s"},{"start_seconds":626,"text":"everyone has this kind of like either nebulous or very specific definition of what that means when you think of AI","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=626s"},{"start_seconds":632,"text":"agents in this context like what are you thinking about when you say that it's","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=632s"},{"start_seconds":641,"text":"interesting for individual agent it's probably combinate like it's it's some combination of uh you know let's just say code that has access to a language","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=641s"},{"start_seconds":650,"text":"model right and so it's able to you know interact with with some environment potentially like that is shared with a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=650s"},{"start_seconds":657,"text":"user or users and agents and able to you know do actions in this and use language","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=657s"},{"start_seconds":664,"text":"model to pretty much understand what other people and agents are doing and saying as well as what happens in the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=664s"},{"start_seconds":671,"text":"environment to like make it more concrete what we're working on on near AI developer side right we have an agent","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=671s"},{"start_seconds":678,"text":"that talks to the user to understand what they're trying to build and writes a specification right so its environment is chat and specification and we have an","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=678s"},{"start_seconds":687,"text":"as agent which is looks at specific and tries to build the you know smart contract and then still may go back to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=687s"},{"start_seconds":694,"text":"user and like make sure that some of the uh like tradeoffs are valid and so you can have like you know you can almost","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=694s"},{"start_seconds":703,"text":"imagine a group chat with your de team uh where all of them are trying to figure out what you want to be built and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=703s"},{"start_seconds":710,"text":"then building it and kind of showing you their progress you know have a designers well join out um so but under these the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=710s"},{"start_seconds":718,"text":"way it works is you know all of the agent is just some python or whatever code that interacts with environment and has a language","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=718s"},{"start_seconds":726,"text":"problem amazing amazing on near at the moment like what are the what are the best like inference uh infrastructure","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=726s"},{"start_seconds":734,"text":"projects and LM sort of infrastructure projects that are currently on there right now where you can use that as your base agent sort of building block to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=734s"},{"start_seconds":742,"text":"then wrap it in Python to give you that user interface and are you seeing anyone using it for like agents not just for","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=742s"},{"start_seconds":749,"text":"end user interaction but for kind of internal processing I'll give you an example we we scrape a lot of uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=749s"},{"start_seconds":756,"text":"volatile web data so all of your tags on websites like hacken noon and these Pages like change a lot they're","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=756s"},{"start_seconds":763,"text":"relatively ephemeral and one day you can have a scraper that passes data out that works and the next day your pipeline could break you've got to constantly","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=763s"},{"start_seconds":770,"text":"monitor we've been um fine-tuning uh llama 3 to pass specifically this volatile data into","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=770s"},{"start_seconds":779,"text":"structured Json so in a way it's like a it's an agent that's doing pipeline work um for the data itself can do a similar","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=779s"},{"start_seconds":786,"text":"thing to tweets like adding metadata to tweets and doing that like second order layer of like processing or post scrape","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=786s"},{"start_seconds":794,"text":"processing we're quite excited about like what that can do there's limitations like latency cost can be a problem tokens per minute throughput and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=794s"},{"start_seconds":802,"text":"things like that and I think there's a lot of work to be done there to get it to scale but quite exciting what it can do to data pipeline infrastructure um are you seeing any of that being","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=802s"},{"start_seconds":810,"text":"built in near right now and would love to kind of um see if there's any other use cases for agents outside of like end user interaction that you're seeing","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=810s"},{"start_seconds":819,"text":"growing and that's probably all of my questions for a bit otherwise I'm gonna hog the mic yeah I mean there's like few","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=819s"},{"start_seconds":827,"text":"components here one is the in like kind of AI inference infrastructure which we can dive into but we are actually you know working with a few projects that","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=827s"},{"start_seconds":836,"text":"are coming in right now into the ecosystem uh as well as working on something we call AI inference router","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=836s"},{"start_seconds":843,"text":"which is a way for then applications to actually plug in into this and get routed to whoever provides inflence and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=843s"},{"start_seconds":850,"text":"compute um so solution yeah yeah showcase that as well as then it should","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=850s"},{"start_seconds":858,"text":"serve as a plugin for data providers and other uh kind of VB 3 foundational models Etc who can actually plug in into","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=858s"},{"start_seconds":866,"text":"one system so it's it's a little bit early but like uh we can we'll we'll talk about that more in in coming weeks","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=866s"},{"start_seconds":874,"text":"yeah um have you used open routes is a great tool sounds similar to that yeah yeah I mean I think open routers solves a little bit different","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=874s"},{"start_seconds":883,"text":"problem but yeah like we trying to think how how can we bring all of the projects that they're building right now around near in the I ecosystem how do we like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=883s"},{"start_seconds":891,"text":"actually com com allows them to compose with each other in you know in a way that Al then serves and user or end","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=891s"},{"start_seconds":899,"text":"application uh like in easy way yeah I like sorry go on go on yeah","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=899s"},{"start_seconds":908,"text":"uh I was just like um having a question for Brandon sort of um ma because um I know like scale AI pretty well and I","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=908s"},{"start_seconds":917,"text":"assume you guys are sort of like doing what they're doing at in web stre right are guys like handling the like data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=917s"},{"start_seconds":925,"text":"collection or only the data labeling yeah it for us we've really focused on","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=925s"},{"start_seconds":933,"text":"the data collection and scraping so um Alex sort of picked up on a an interesting point here about tweet data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=933s"},{"start_seconds":942,"text":"so we actually being very transparent we reverse engineered uh Twitter's front-end API the graphql API um as a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=942s"},{"start_seconds":950,"text":"Twitter worker on our Network you can bring your Twitter credentials you can add them to a node it will log you in save a session cookie on your um machine","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=950s"},{"start_seconds":960,"text":"or your instance um and then we route through a worker layer an ACA framework","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=960s"},{"start_seconds":966,"text":"requests ht2 requests from our API that get fulfilled by those workers there's an element of consensus around the the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=966s"},{"start_seconds":974,"text":"scrapes um so we've really focused on uh unlocking the Twitter data that's typically like not available through the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=974s"},{"start_seconds":982,"text":"Twitter API or if it is available you've got to spend an awfully an a like a huge amount of money like it's $42,000 a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=982s"},{"start_seconds":990,"text":"month just to be able to get a endpoint that gives you followers and right now we do it for free for everyone so you can get any data you want from Twitter","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=990s"},{"start_seconds":998,"text":"completely for free right now um I call it the you Elon product because he completely screwed up the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=998s"},{"start_seconds":1005,"text":"uh the Twitter API very tongue and cheek but it's it's memorable for everyone um","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1005s"},{"start_seconds":1012,"text":"and we then do web scrape data Discord Reddit basically anything you see on the public web you would get through a platform like app appify um you can get","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1012s"},{"start_seconds":1021,"text":"through uh Massa protocol for free um we're then building on the back end more of like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1021s"},{"start_seconds":1028,"text":"the processing pipeline infrastructure where I have high conviction that instead of investing heavily in data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1028s"},{"start_seconds":1036,"text":"pipelines that transform data using static code using llms to do that and the pace at which people like you are","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1036s"},{"start_seconds":1044,"text":"probably bringing compute to the market and other networks there's be so much latent unused compute out there that you can almost run background jobs on data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1044s"},{"start_seconds":1052,"text":"sets um to provide this enrichment L on top so that's something Downstream we're adding to as we get this big Corpus of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1052s"},{"start_seconds":1061,"text":"data that's kind of historical based on what people are scraping so we have two things we store the data that is scraped By Request there's some really","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1061s"},{"start_seconds":1069,"text":"interesting stuff there that you can see like what is most valuable kind of surfaces from that data because it's what people are querying the most and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1069s"},{"start_seconds":1077,"text":"you can go deep there and then we're also going to run batch jobs to build data sets in our back end and store them on something like OG laabs or uh space","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1077s"},{"start_seconds":1086,"text":"and time something like that as sort of permanent storage so yeah focusing on the data scraping got you got you and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1086s"},{"start_seconds":1093,"text":"like for the data what type of data do you guys support now like only pouch or also like video and audio at the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1093s"},{"start_seconds":1102,"text":"moment yeah it's a good question um video and audio we cut out of scraper","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1102s"},{"start_seconds":1109,"text":"just because of the size of the data and storing it was becoming an issue um but it's it's a relatively easy change to add it back in um so we haven't really","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1109s"},{"start_seconds":1118,"text":"done much computer vision stuff we've been really focused primarily on the um on the text based uh data because most of the agents we're seeing like IIA was","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1118s"},{"start_seconds":1126,"text":"saying are user facing agents that are looking for text either to find tune or","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1126s"},{"start_seconds":1132,"text":"go into rag um or in certain cases where people have large budgets they want to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1132s"},{"start_seconds":1138,"text":"trainer or retrain a base model um like Spectra have done this Spectra have retrained a base model on uh GitHub data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1138s"},{"start_seconds":1146,"text":"um specific to Smart contracts so you have a really great uh agent that's able to give a lot of detail on Smart","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1146s"},{"start_seconds":1153,"text":"contract uh syntax as opposed to it sort of out of the box and from like a business perspective what we've been looking for and very","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1153s"},{"start_seconds":1162,"text":"luckily be able to get as one of the first decentralized AI data product out there is to find what we call data product Market fit so think of it as","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1162s"},{"start_seconds":1170,"text":"trying to reach equilibrium between the demand side and the supply side we need to basically learn iterate for our","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1170s"},{"start_seconds":1177,"text":"developers specifically AI application and AI agent developers to understand their use case and then we can rapidly","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1177s"},{"start_seconds":1185,"text":"ask our network of validators and node workers to spin up new data scrapers so far we find two use cases to be the most","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1185s"},{"start_seconds":1193,"text":"prevalent uh the first is uh Financial no surprise there um everyone is trying to use AI agent to capture trading","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1193s"},{"start_seconds":1201,"text":"signal Etc and then the second one is social we have seen a huge wave of AI companions they have boyfriends and they have girlfriends emerging right now","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1201s"},{"start_seconds":1210,"text":"loneliness is one of the biggest problem facing Millennial and genz right now and it is directly reflected in the use case so based on those two leading categories","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1210s"},{"start_seconds":1218,"text":"of use cases we started to develop different kind of data scrapers Social Web data are the most popular amongst","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1218s"},{"start_seconds":1226,"text":"them so that's what's informing road map at the moment understood yeah because I mean","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1226s"},{"start_seconds":1233,"text":"it's quite interesting because I used to work for a project called delum we a AI agent Network that we automate the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1233s"},{"start_seconds":1241,"text":"entire like workflow for users in crypto for for instance like I think we recently launched a partnership with","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1241s"},{"start_seconds":1248,"text":"Galaxy that like because on Galaxy there a a lot of users trying to do like XYZ to farm for the air job or to attend the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1248s"},{"start_seconds":1255,"text":"campaigns right and AI agent could be a really good way way for them to automate the process and just to complete all the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1255s"},{"start_seconds":1262,"text":"stuff with one click and I'm just running are you're saying you pretty much in automating what people are paying for to do user engagement you're","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1262s"},{"start_seconds":1271,"text":"automating it with AI and replacing user engagement with AI yes that will be one of the use case like we explore because","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1271s"},{"start_seconds":1279,"text":"um so we started as a um a NPC solution for gaming but our tax that is around","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1279s"},{"start_seconds":1286,"text":"NLP right and how to interact from a virtual human being to to human uh and then we figure like with such a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1286s"},{"start_seconds":1294,"text":"technology what can we do to automate the workflow um specifically in crypto and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1294s"},{"start_seconds":1301,"text":"one use case we found it just you know like there are a lot of people manually doing xyd like SN their Twitter uh do","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1301s"},{"start_seconds":1308,"text":"some transfer and all that and we're essentially having this like Infinity canvas sort of thing for users to DIY","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1308s"},{"start_seconds":1318,"text":"cust cize their workflow and in the meantime to have the AI agent to like just I guess complete them with one clip","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1318s"},{"start_seconds":1326,"text":"which is super cool I guess but so for that type of dat agent do you foresee any Synergy between","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1326s"},{"start_seconds":1333,"text":"like your send them um yeah you know we we we've been","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1333s"},{"start_seconds":1341,"text":"through like a bit of a journey with some of this clarification um on our way to realizing that data for AI consumption","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1341s"},{"start_seconds":1349,"text":"was really where the best fit in the market is today like data is a pretty unsexy thing like no one typically unless you're a real nerd gets super","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1349s"},{"start_seconds":1357,"text":"excited about data um but DJing everybody loves DJing so our first kind","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1357s"},{"start_seconds":1364,"text":"of initial use case was uh verifying uh dgen behavior from social like twitter","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1364s"},{"start_seconds":1371,"text":"twitter followers is example like are you following XYZ account um yeah we're able to provide verification for that uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1371s"},{"start_seconds":1380,"text":"through the platform so an agent will be able to understand does Ilia follow Masa or does he follow a you know and get a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1380s"},{"start_seconds":1387,"text":"response back and determine whether or not iler is qualified for that sort of event or something so I believe there are ways in which uh reasoning can come","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1387s"},{"start_seconds":1396,"text":"from llms using data of this type and sort of build more complex decision trees that that essentially unlock the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1396s"},{"start_seconds":1405,"text":"use case that you're describing yeah I'm sure there is well I think it it there's even more right like you're bringing on for example Twitter data which you can use","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1405s"},{"start_seconds":1414,"text":"to make decisions on than other platforms uh on chain as well and so","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1414s"},{"start_seconds":1421,"text":"that's given like Twitter is probably the you know news feed of crypto you can actually have decisions made uh based on","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1421s"},{"start_seconds":1431,"text":"that so I think like that's kind of what we're thinking about like how do we plug in kind of the you know some of the data providers","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1431s"},{"start_seconds":1439,"text":"into as a context that then other uh you know be agents or users can use directly","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1439s"},{"start_seconds":1446,"text":"in their llms you know make it almost like you know set a mware that augments your query or your or your context as","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1446s"},{"start_seconds":1455,"text":"you makeing decision love that yeah love that sorry I was gonna say no surprise","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1455s"},{"start_seconds":1462,"text":"here right after um Twitter we actually devolve additional data scrapers supporting Discord and telegram data um","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1462s"},{"start_seconds":1469,"text":"for example if near users want to um I don't know capture trading signal of all the tokens available on near where they","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1469s"},{"start_seconds":1477,"text":"would go to first of all they would go check a few big whale accounts on Twitter secondly they will probably go to their favorite communities on","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1477s"},{"start_seconds":1485,"text":"telegram seeing what kind of signal calls are available on telegram that's exactly what we can provide as a data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1485s"},{"start_seconds":1493,"text":"platform what we can do is really give you real-time structure uh Text data um based on telegram group chat history as","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1493s"},{"start_seconds":1501,"text":"well as discords so combined once you have the big three Discord telegram Twitter you almost get a pre-completed","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1501s"},{"start_seconds":1509,"text":"related to specific token trading yeah I just I just want to add something to what IIA was saying just now about this kind of what I'm almost","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1509s"},{"start_seconds":1517,"text":"hearing is like a reasoning framework so we got like inference and there's almost this like reasoning framework built on top of llms I think it's for a lot of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1517s"},{"start_seconds":1525,"text":"people in the web three space there's this Vision it's almost autonomous Behavior within the web three stat constrained by a rule set or parameters","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1525s"},{"start_seconds":1534,"text":"that allow for certain actions to take place in a semi-autonomous or fully autonomous way um I think that","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1534s"},{"start_seconds":1543,"text":"this discourse and thought process leads to that outcome incrementally uh and opens up like a lot","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1543s"},{"start_seconds":1550,"text":"of interesting use cases for you know I'll give you an example we dog food our","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1550s"},{"start_seconds":1557,"text":"own Discord scraper internally where we will scrape user feedback user testing groups in Discord um and it will","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1557s"},{"start_seconds":1566,"text":"synthesize and output feature requests prioritize them for us add them into like a ticket backlog in","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1566s"},{"start_seconds":1574,"text":"GitHub not that we're trying to replace people in the organization but the velocity you now get in terms of taking","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1574s"},{"start_seconds":1581,"text":"raw data refining it and then making it actionable in our pipeline just increases like engineering and product disc velocity where you couldn't do this","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1581s"},{"start_seconds":1590,"text":"before so it's not just use cases where you're deing it's like how can I introduce both efficiency and observability into my ex existing","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1590s"},{"start_seconds":1598,"text":"processes so as I can extract more insights and imagine taking that one one step further you you're collecting information you're observing it and then","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1598s"},{"start_seconds":1606,"text":"you're doing reduction through reasoning to understand okay these are the outcomes I want to pursue because of XYZ heuristic you start building up like a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1606s"},{"start_seconds":1615,"text":"very very powerful engine and like you know tie that together with economic incentives where you can program your","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1615s"},{"start_seconds":1623,"text":"incentive model on top of that layer you can really bring together those actors to make you know favorable outcomes I strongly strongly","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1623s"},{"start_seconds":1633,"text":"believe yeah makes sense I mean there's definitely a lot to develop here maybe Switching gears uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1633s"},{"start_seconds":1643,"text":"Albert I mean you you mentioned kind of the project you were before right but now with compute Labs you're working","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1643s"},{"start_seconds":1649,"text":"more on kind of the base level mization of compute and kind of bringing bringing forward the the value","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1649s"},{"start_seconds":1658,"text":"of the future I guess how are you guys seeing the kind of the opportunity with all the roll out right I mean like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1658s"},{"start_seconds":1666,"text":"Amazon is building I don't know 25 new data centers um how do how do you see this on","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1666s"},{"start_seconds":1673,"text":"boarding looking and how do we uh how do we is crypto actually participate in that game more","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1673s"},{"start_seconds":1680,"text":"effectively well uh I guess like your second part how we can participate that's exactly what we're trying to achieve here or like enable investor to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1680s"},{"start_seconds":1689,"text":"do uh especially like institutional investors and retails right because currently at the market there's","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1689s"},{"start_seconds":1696,"text":"no Financial product that provide a a direct exposure to asset to like compute as underlying asset there's Invidia","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1696s"},{"start_seconds":1704,"text":"stocks there AI related tokens and you might also invest in AI company from early stage but other than that like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1704s"},{"start_seconds":1712,"text":"there's no direct exposure which is why we're thinking uh we want to focus on the financialization of this entire","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1712s"},{"start_seconds":1720,"text":"computer asset because we both know computer is going to be the currency of the future and when we're talking about the financialization or derivatives","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1720s"},{"start_seconds":1729,"text":"they're going to be like index ETF perpetu stating Landing uh sort of things and we feel like that market will","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1729s"},{"start_seconds":1736,"text":"be huge but of course the Foundation or the first step will be that we have to have this alching representation or like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1736s"},{"start_seconds":1744,"text":"how we can transfer the ownership of that compute because physically it's just not feasible to transfer compute","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1744s"},{"start_seconds":1750,"text":"using just regular secondary Marketplace since it's heavy it's hard to be Shi and also easy to be broken which is why uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1750s"},{"start_seconds":1759,"text":"we feel like using blockchain technology to build this scap for compute as underlying asset makes a lot of sense","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1759s"},{"start_seconds":1767,"text":"and that's how we started and going back to like I guess AWS or more web tool","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1767s"},{"start_seconds":1774,"text":"Marketplace uh for compute versus web three um I guess I also want to talk to you about this is like currently I think","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1774s"},{"start_seconds":1782,"text":"the demand and Supply situation from both worlds uh have a complete different","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1782s"},{"start_seconds":1789,"text":"um situation right in web to the demand is much higher than the supply for the web to Cloud computer network however","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1789s"},{"start_seconds":1796,"text":"web free um I think like I don't that they just launched their token today right uh for those decentralized Compu","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1796s"},{"start_seconds":1806,"text":"Marketplace um I would say due to the physical limitation and all the network latency","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1806s"},{"start_seconds":1812,"text":"the actual demand there is uh is limited fairly limited and what's your opinion","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1812s"},{"start_seconds":1820,"text":"on like how you know the web stre components can develop longterm in a way that the demander will be more actual","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1820s"},{"start_seconds":1829,"text":"and like they're going to be a sustainable economy","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1829s"},{"start_seconds":1836,"text":"there yeah I I think I think there's a uh um the problem is that for the use cases","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1836s"},{"start_seconds":1846,"text":"that exist today that require a massive number of gpus you do need them to be with the uh infin band and whatnot right","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1846s"},{"start_seconds":1853,"text":"so it's one of the uh there are several scenarios where the demand will go up one of them is um presumably you can do","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1853s"},{"start_seconds":1862,"text":"inference uh without you know like like that's a viable viable scenario to do inference quickly on uh on machines that","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1862s"},{"start_seconds":1870,"text":"do not have that that quick of an interconnect uh especially uh it's one thing to have inference that you want you know like if you want to to be","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1870s"},{"start_seconds":1877,"text":"communicating with the uh you know in real time uh you probably want something with a very low latency high throughput","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1877s"},{"start_seconds":1885,"text":"but in many cases you want to do batch jobs then you don't care you know about Laten you care about trut so in this","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1885s"},{"start_seconds":1893,"text":"case I think the centralized networks already have an advantage uh and I think another one is I'm still very hopeful","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1893s"},{"start_seconds":1902,"text":"that there is going to be a breakthrough that will allow training uh with very high latency then of course uh uh the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1902s"},{"start_seconds":1910,"text":"question will be can I de centralized Network provide gpus uh cheaper on a per flop basis","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1910s"},{"start_seconds":1917,"text":"right IGN ing the network uh and the bandwidth that is something I don't I don't know if it you know if it's viable","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1917s"},{"start_seconds":1926,"text":"I don't know like what's the comparison um between the price per flop on a decentralized network versus you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1926s"},{"start_seconds":1932,"text":"know just getting h100s at like $2 a piece yeah uh I mean like uh 100%","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1932s"},{"start_seconds":1940,"text":"current limitation sort of just limit the usage for web Street components more to the AI inferencing for AI training","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1940s"},{"start_seconds":1947,"text":"due to the like you know infinite band and all that it's just not visible at the moment y yeah yeah we should invest more into","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1947s"},{"start_seconds":1957,"text":"yeah go ahead yeah no I just to mention like on where actually inferencing may be useful at a training time like a kind","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1957s"},{"start_seconds":1965,"text":"a background B inferencing is synthetic data generation right because uh I think","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1965s"},{"start_seconds":1971,"text":"we'll see a pretty sizable transition for training new models on on synthetic data that's specifically generated","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1971s"},{"start_seconds":1980,"text":"that's kind of uh from you know po smarter models distilled into smaller models um or just in general kind of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1980s"},{"start_seconds":1987,"text":"generating distributions of data that are useful and for that you need massive batch of inference uh you don't care","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1987s"},{"start_seconds":1995,"text":"about patency at all and you just need like as much Compu as POS as as possible so I think that's probably one of the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=1995s"},{"start_seconds":2002,"text":"interesting use cases in the shter term uh if if we can coordinate all deves","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2002s"},{"start_seconds":2010,"text":"yeah it's something we've seen initial like green shoot interest on is this notion of like enrichment or synthetic","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2010s"},{"start_seconds":2019,"text":"data sets using not the Json model we have but a uh like an enrichment model","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2019s"},{"start_seconds":2027,"text":"where natural language conversations from things like YouTube Reddit um","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2027s"},{"start_seconds":2033,"text":"conversational uh text is being enriched by uh another protocol to create better","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2033s"},{"start_seconds":2041,"text":"conversational AI essentially doing what GPT 40 did to their most recent release right using all of the conversational","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2041s"},{"start_seconds":2048,"text":"data from all of our prompts to fine Chen the model so as it had what felt like a more natural uh Pros or","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2048s"},{"start_seconds":2056,"text":"vocabulary with what we speak you know the way we speak in in everyday life um and and they don't care about","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2056s"},{"start_seconds":2063,"text":"latency um they very much care about price and we we've recently worked on a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2063s"},{"start_seconds":2071,"text":"coordination engine that allows you to imagine spawning a set of workers for an","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2071s"},{"start_seconds":2078,"text":"llm that scales horizontally um and allows you to retain the full sort of image of your llm instance but spread","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2078s"},{"start_seconds":2087,"text":"work across multiple devices with lower concerns for latency um so you end up using a single llm","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2087s"},{"start_seconds":2095,"text":"instance but you can run it over multip multiple workers that are multiple gpus um actually using we use lib","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2095s"},{"start_seconds":2103,"text":"peer-to-peer which is like the file coin system to spawn those workers with an ACO framework which is kind of interesting implementation bit like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2103s"},{"start_seconds":2109,"text":"kubernetes for decentralized llms um and I think that we'll start to see better inference infrastructure better","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2109s"},{"start_seconds":2118,"text":"infrastructure for llms I don't think our is a final solution I think it's like an initial solution for this um that then allows for higher throughput","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2118s"},{"start_seconds":2125,"text":"batch processing and and we're looking to collaborate people like um aash would you know love to see if there's an opportunity to collaborate here to see","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2125s"},{"start_seconds":2133,"text":"how that scales and does it work for some of these early you know synthetic data set enrichment uh batch","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2133s"},{"start_seconds":2141,"text":"jobs yeah it's interesting 100% um and also I guess","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2141s"},{"start_seconds":2152,"text":"like another question for you guys is let's say currently there's a investment opportunity in front of you that can","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2152s"},{"start_seconds":2160,"text":"provide a stable 20 to 30 even 40% of yield per year and essentially you will","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2160s"},{"start_seconds":2167,"text":"be just putting money and purchasing a GPU yourself that will be custo custodian and also physically managed by","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2167s"},{"start_seconds":2176,"text":"computer apps or like by a professional team to generate yours right like to rent it out to AI companies from web to will you consider that as a related","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2176s"},{"start_seconds":2184,"text":"investment opportunity yourself as a as an ex Miner and spending years","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2184s"},{"start_seconds":2192,"text":"early on mining the minor dgen on me wants like a pool switching equivalent for gpus","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2192s"},{"start_seconds":2201,"text":"um you know we're investing in some some of our own Hardware at Massa to be able to run our own um compute just for R&D","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2201s"},{"start_seconds":2210,"text":"and to be able to have almost like a a mining pool switching engine where you're always mining the most profitable algo back in the day","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2210s"},{"start_seconds":2218,"text":"to basically provisioning your Hardware in the most profitable like use case um sounds interesting and like I'm speaking","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2218s"},{"start_seconds":2226,"text":"from someone that can technically set this Harbor up but you know uh collocated hosted miners became","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2226s"},{"start_seconds":2235,"text":"a massive thing they still are a massive thing you it feels like it's inevitable um that that's the case like we are going to allow scrapers to be loaded","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2235s"},{"start_seconds":2244,"text":"onto an Android device so you'll download an Android app you would just hit like a connect button enter any credentials and then it's almost like tethering to the network with your phone","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2244s"},{"start_seconds":2253,"text":"uh and you can essentially monetize the bandwidth on your device people want that really quick oneclick solution","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2253s"},{"start_seconds":2260,"text":"right to participate in this economy um yeah yeah yeah yeah I think both compute labs and Masa are trying to make sure","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2260s"},{"start_seconds":2269,"text":"that every user of the internet or of the internet will be able to financially benefit from AI right um like I'm","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2269s"},{"start_seconds":2277,"text":"probably the only non-technical founder on this call right now I don't know how to set up a hardware nor do I know how to run a very technical node but in the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2277s"},{"start_seconds":2284,"text":"not so far future comput Labs will be able to allow me this opportunity to invest in like a piece of Hardware","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2284s"},{"start_seconds":2292,"text":"rented out and then you'll be able to manage it on my behalf and I can generate 20 to 40% yield that sounds pry amazing to me I'll probably only have","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2292s"},{"start_seconds":2300,"text":"one request which is to make the yield Dynamic if it's possible because given the speed yeah given the speed AI","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2300s"},{"start_seconds":2307,"text":"development I want to see yield not C at 40% yeah and we're actually also thinking about like developing this","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2307s"},{"start_seconds":2315,"text":"yield volatility so another derivative on the E itself and I we truly believe that there's going to be a lot of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2315s"},{"start_seconds":2323,"text":"speculation room for that and just like what Brendan said um I think like same for for computer labs","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2323s"},{"start_seconds":2330,"text":"basically uh the end goal for us is also going forward uh any user who like can just go on their phone and like maybe go","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2330s"},{"start_seconds":2338,"text":"to their wallet and after a few CLS they can purchase the actual run GPU a hardware that's like installed in a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2338s"},{"start_seconds":2347,"text":"physical world but they can have this nft serving as owner proof and also the Ying token right I guess like that's going to be the future we're aiming for","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2347s"},{"start_seconds":2355,"text":"here yeah I we're actually literally just brainstorming this uh at consensus a couple weeks ago um both Bren and I","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2355s"},{"start_seconds":2363,"text":"dabbled in the fintech world and in the fintech world there's a lot of wallets that gain massive adoption because they were pre-installed on Hardware or phones","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2363s"},{"start_seconds":2371,"text":"for example in India Reliance goo is such a conglomerate that promoted PTM amongst other wallets what if in the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2371s"},{"start_seconds":2378,"text":"future every single like Android phone out there will be able to automatically scrape data contribute data plan where we'll be able to achieve one click or","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2378s"},{"start_seconds":2387,"text":"two like uh click uh renting a compute resource that would be pretty amazing and I think the future is actually not that far","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2387s"},{"start_seconds":2395,"text":"away so near actually has a application Kai chain which is pre-installed on over 8 billion phones already as pretty much","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2395s"},{"start_seconds":2404,"text":"on CI and opa phones and so that that is the part they actually are expanding this into so-called locally store data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2404s"},{"start_seconds":2412,"text":"protocol which allows to kind of run AI on their device but they have a wallet","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2412s"},{"start_seconds":2419,"text":"pretty much for every user they have all of that integrated and so that is an opportunity to integrate into that system","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2419s"},{"start_seconds":2428,"text":"amazing we would love to integrate that sounds exactly what we have in invion and plan","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2428s"},{"start_seconds":2435,"text":"for pursuing that before Apple's announcement yesterday yeah speaking of that like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2435s"},{"start_seconds":2443,"text":"what's your thought on the partnership between Apple and open I saw Elon was like on Twitter yesterday saying like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2443s"},{"start_seconds":2450,"text":"the Privacy concern and all that yeah I mean I got I got strong thoughts","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2450s"},{"start_seconds":2457,"text":"I always do about these things yeah I have really strong thoughts um I mean it's it's it's all just uh deals through","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2457s"},{"start_seconds":2464,"text":"the back door right like I I think it comes down to I don't know I'm gonna pause I'm G let somebody else go first I'm not gonna go down to Rabbit Hole here because I will go down I will go","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2464s"},{"start_seconds":2472,"text":"down to Rabbit Hole but but in this particular case you are on the El site yes I am yes personally I am yes","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2472s"},{"start_seconds":2481,"text":"very much so yeah I I I mean I think we see an interesting convergence right and I mean everybody","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2481s"},{"start_seconds":2489,"text":"was expecting that announcement too that um although although a bunch of people were telling me like no open AI I mean","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2489s"},{"start_seconds":2498,"text":"sorry Apple has you know developed all this AI stuff but uh I mean and they do have some UND device things is built but","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2498s"},{"start_seconds":2507,"text":"at the end yeah it seems like they're just integrated um I mean I think realistically the operating system of the future will have ai pretty much as","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2507s"},{"start_seconds":2516,"text":"part of the experience layer and so the question is how is that done who is powered on and who is getting all","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2516s"},{"start_seconds":2523,"text":"the data right and so I mean that's why like I actually been talking about Edge intelligence um for over a year now kind","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2523s"},{"start_seconds":2531,"text":"of focused on how do we bring models on device how do we ensure that they can be as smart as possible leverage in your","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2531s"},{"start_seconds":2539,"text":"local data uh your kind of local context um and then potentially even becoming the way you engage with experiences and","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2539s"},{"start_seconds":2548,"text":"interfaces going forward right because if you can imagine this model starting to write code um or they can actually just build","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2548s"},{"start_seconds":2556,"text":"interface for any experience you want right maybe not on the fly but you can then share those experiences not as","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2556s"},{"start_seconds":2564,"text":"applications but actually as a recipe that you can change yourself and so I think like we see that is starting","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2564s"},{"start_seconds":2573,"text":"to happen except obviously as a l platform right there's a benefit of them","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2573s"},{"start_seconds":2579,"text":"kind of controlling it um and not also opening it up to other developers so uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2579s"},{"start_seconds":2588,"text":"and then I mean all the privacy concerns for sure will just keep accelerating from here yeah you know I think a lot of the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2588s"},{"start_seconds":2596,"text":"Genesis for opinions yesterday is that we don't know right like the system can't be trusted right now and there's","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2596s"},{"start_seconds":2605,"text":"enough evidence to say that it can't be trusted trusted because of past behavior and I think it opens up a lot of philosophical questions as to we're","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2605s"},{"start_seconds":2613,"text":"always talking in web 3 whether it be like ZK proofs doing XYZ for like ml or lots and lots of solutions and how they","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2613s"},{"start_seconds":2621,"text":"can be applied but it's all well and good us talking about that within web 3 when we really need to have some of these bigger organizations adopting some","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2621s"},{"start_seconds":2628,"text":"of this and I just don't ever see that happening and it's because we as users don't put enough pressure and demand on","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2628s"},{"start_seconds":2635,"text":"these organizations to change their behavior and it's because the economics don't work so we have two options we go down the river with this kind of thing","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2635s"},{"start_seconds":2644,"text":"or we push back and we fight and I certainly want to be fighting I think this might be one of the first times where I reconsider getting an iPhone for","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2644s"},{"start_seconds":2651,"text":"the last eight years um much to my wife sugar in she'll see green messages from me not blue messages um and we have to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2651s"},{"start_seconds":2659,"text":"build a different system like we we have to continue building these different systems and quantize models like is","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2659s"},{"start_seconds":2667,"text":"talking about for like Edge devices are very very interesting like there's a huge you know Bode of Open Source","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2667s"},{"start_seconds":2674,"text":"research that goes into quantise models that could do a lot I I haven't seen a huge amount of projects using quantise","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2674s"},{"start_seconds":2681,"text":"models in web 3 or am I missing those projects who out there is using quantized models instead of like llama 3 or another kind of hugging face model is","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2681s"},{"start_seconds":2689,"text":"there anyone oh those are quantied in many cases right yeah you you you can yeah you can","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2689s"},{"start_seconds":2697,"text":"but who who's who's specifically using those in in the web 3 space the really lightweight ones that have performance","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2697s"},{"start_seconds":2705,"text":"devices few projects I've seen who are using on on edge models and through that accessing a lot more data on a person's","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2705s"},{"start_seconds":2713,"text":"device I mean they're not yet like you know live but uh they're you know at at","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2713s"},{"start_seconds":2721,"text":"C stage pretty much developing their prototypes it's things I I I think","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2721s"},{"start_seconds":2728,"text":"emphasis like the emphasis that you made was that they do some things uh on device but I think they do majority of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2728s"},{"start_seconds":2737,"text":"the things on device and and minority through open AI which is which is worth mentioning it is yeah and asks","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2737s"},{"start_seconds":2745,"text":"explicitly actually to go to right so yeah yeah it's different from hypothetical Microsoft where where it is","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2745s"},{"start_seconds":2752,"text":"you know uh where all your data is is just leaving your machine","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2752s"},{"start_seconds":2759,"text":"constantly yeah and and it's the same I I think it's still the same the same argument about iPhone versus Android that on Android by default you know like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2759s"},{"start_seconds":2769,"text":"just like even going to the onboarding sequence where you know uh where every single control is designed in a way for you to give permission to send your data","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2769s"},{"start_seconds":2778,"text":"without understanding it I think for for quite a while still Apple will be","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2778s"},{"start_seconds":2784,"text":"unfortunately um the choice to make yeah for context Alex and I at some","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2784s"},{"start_seconds":2791,"text":"point broke Android well we worked around Android in such a way that people felt like they're approving one thing","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2791s"},{"start_seconds":2799,"text":"and actually approved a bunch of other permissions they haven't seen yeah that was not intended behavior and we filed it as an issue it took three years for","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2799s"},{"start_seconds":2807,"text":"them to fix it or four wow and I'm 5,000 richer wow that's that's that's that's","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2807s"},{"start_seconds":2816,"text":"crazy yeah I I've done I've done data collection from Android with consented","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2816s"},{"start_seconds":2823,"text":"permissions that give you very de detailed information on people to build credit models in the past yeah yeah","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2823s"},{"start_seconds":2836,"text":"yeah yeah uh and I guess just one more comment regarding like what Brandon just said I feel like the AI landscape is","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2836s"},{"start_seconds":2844,"text":"just becoming espe in web to it's becoming more and more oligarchy a little bit that like a few bit company","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2844s"},{"start_seconds":2852,"text":"they're controlling all the stuff it's prob now with a partnership between ab and open right yeah yeah I think that's part of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2852s"},{"start_seconds":2860,"text":"the problem right you have to you have to use Android or an iPhone there's no way to live in 2024 without having one","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2860s"},{"start_seconds":2867,"text":"of the two right so it's a yeah yeah we we really see ourselves where the entire","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2867s"},{"start_seconds":2874,"text":"human race at the crossroads of central AI versus decentralized AI at this specific point in time the good thing is","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2874s"},{"start_seconds":2881,"text":"that we are building in parallel although right now centralized AI players have way too much resources compared to what we're building","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2881s"},{"start_seconds":2888,"text":"collectively but still we have the first opportunity in human history to be able to build in parallel with centralized","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2888s"},{"start_seconds":2896,"text":"players and hopefully surpass them at some point yeah we're definitely living","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2896s"},{"start_seconds":2905,"text":"exciting time I think that kind of for me shines this bright light on like what are we all doing here like it's not just the why","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2905s"},{"start_seconds":2914,"text":"decentralized AI which for many of us I think seems intuitive and obvious you know it's creating equal opportunity","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2914s"},{"start_seconds":2922,"text":"aligning incentives around economic models that you can program to bring the right players together to coordinate and act in a way that's good for the entire","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2922s"},{"start_seconds":2930,"text":"system not just one centralized player and I think that where we are in the cycle you know where Bitcoin found its","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2930s"},{"start_seconds":2937,"text":"fit for where it is now store of money store of value we need to collectively as a an entire ecosystem building in","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2937s"},{"start_seconds":2946,"text":"decentralized a like nail what that single very memorable value proposition is for decentralized that Ai and like what the why is you know ethereum did it","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2946s"},{"start_seconds":2955,"text":"with like the world computer even if we can judge with hindsight whether that came to pass or not um near is doing it","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2955s"},{"start_seconds":2963,"text":"with what you guys are doing with sort of more scalable uh layer one Solutions but like what is what is this particular one single thing that harmonizes the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2963s"},{"start_seconds":2972,"text":"entire like AI web3 space um that connects with people to make people realize that this is where they should","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2972s"},{"start_seconds":2978,"text":"be um I don't know if I've seen that specific thing unless I describe it yeah","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2978s"},{"start_seconds":2985,"text":"I describe it as user owned AI because like what what the difference you want is it's not company you know trying to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2985s"},{"start_seconds":2993,"text":"generate more Revenue right it's actually focused on the user and their success their needs their financial","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=2993s"},{"start_seconds":3000,"text":"well-being and wellbeing more broadly and so that's kind of the flip we want to make um and it's not like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3000s"},{"start_seconds":3008,"text":"decentralization there is the tool right decentralization always is the tool um but do people do people you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3008s"},{"start_seconds":3016,"text":"speak to sucessful yeah do people you speak to outside of AI care about","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3016s"},{"start_seconds":3023,"text":"that I think I mean I think if you explain it hey like do you want something that makes you know big company more money or do you want","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3023s"},{"start_seconds":3031,"text":"something that makes you more successful uh I think that that seems to resonate um I think on on on its own","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3031s"},{"start_seconds":3040,"text":"maybe it's not like immediately used like obvious but so is so is internet computer let's be clear that most people don't","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3040s"},{"start_seconds":3048,"text":"absolutely the interet computer me either absolutely I really love I really love the framing of user owned AI by the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3048s"},{"start_seconds":3055,"text":"way had U we have a new slogan we have yet to debut uh it's called Fair AI Power by the people which is exactly to","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3055s"},{"start_seconds":3063,"text":"your point right user own AI users should be the owner of their own data their own uh resources their own compute","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3063s"},{"start_seconds":3070,"text":"their own bandwidth as St financially benefit from the AI boom I think with anything else in web 3 decentralization","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3070s"},{"start_seconds":3078,"text":"and power of the community is always going to be the philosophical angle of all of our uh new Innovations","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3078s"},{"start_seconds":3088,"text":"yeah and I think I mean that's been a challenge actually I mean it's it's always been a challenge in web three that if you if you lead with","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3088s"},{"start_seconds":3096,"text":"philosophy the you lose a lot of people right it's it's a minority thing and so what is I I mean that's why I think the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3096s"},{"start_seconds":3104,"text":"economics is so important right versus if you leave as economics like hey you can benefit in this way you can access something that you couldn't access right","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3104s"},{"start_seconds":3113,"text":"which is you know I'm coming from Ukraine access dollars is one of the very important things that uh everybody","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3113s"},{"start_seconds":3122,"text":"understands so I think it's you know we're trying to have this philosophical name but at the end it's going to be like hey here's the opportunity you can","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3122s"},{"start_seconds":3129,"text":"get right and here you know you want to benefit from the you know future of compute while compute Labs offering you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3129s"},{"start_seconds":3137,"text":"something and it's like it's only possible to build on that three because you need to coordinate all of this desperate resources and efforts you know","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3137s"},{"start_seconds":3146,"text":"you want to benefit from you know your band with pretty much sitting at home uh in the background well you can use Massa","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3146s"},{"start_seconds":3154,"text":"so I think that's like at the end is going to be the actual driver for adoption and I I call it like every","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3154s"},{"start_seconds":3162,"text":"Revolution starts with economy and gets described with about values like you know America wanted freedom but actually","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3162s"},{"start_seconds":3170,"text":"they didn't want to pay taxes to UK right so like it started as economy and they were like yeah we want you know that they they plastered that's a good","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3170s"},{"start_seconds":3178,"text":"example that's a good example yeah yeah absolutely yeah I like that everything ends with the IRS in the US but 100% I","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3178s"},{"start_seconds":3188,"text":"deeply resonate with that um I grew up in Asia and even within the Masa Community we have now building our community in the past two years and a","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3188s"},{"start_seconds":3196,"text":"half um the most active community members actually come from Southeast Asia come from Nigeria come from the CIS","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3196s"},{"start_seconds":3204,"text":"region instead of America or Western Europe Etc it's because they see immediate value in what we're giving as","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3204s"},{"start_seconds":3212,"text":"token rewards to participate in our Network $1 is probably only one5 of my coffee this morning in New York but","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3212s"},{"start_seconds":3219,"text":"could mean a lot a lot more to a lot of people elsewhere yeah I often I often speak about that like the cost of anthropic or","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3219s"},{"start_seconds":3228,"text":"gpc4 at $20 a month that's still a huge hurdle for a lot of people and it's only that price because we're giving them","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3228s"},{"start_seconds":3235,"text":"their data we sorry we're giving those companies our data in the prompts if we weren't giving them that data it' be even more expensive for those","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3235s"},{"start_seconds":3243,"text":"subscriptions so we get access to a lower price because we're feeding the model as well kind of a slave to the system in a way um and I think that","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3243s"},{"start_seconds":3252,"text":"opportunity to give Fair access to free AI is something we can really do uh using these economic models especially","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3252s"},{"start_seconds":3260,"text":"to like prevent more inequality happening around this particularly like high moving fast moving space where you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3260s"},{"start_seconds":3266,"text":"could very easily leave people behind yeah uh I feel like if uh from that perspective uh I think Masa is","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3266s"},{"start_seconds":3275,"text":"doing a really amazing job of because like if we're if we're thinking about it the more data those AI company they have the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3275s"},{"start_seconds":3283,"text":"higher their technology barrier will be so that they can charge more for their subscription so we definitely need you guys to do to like basically","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3283s"},{"start_seconds":3291,"text":"decentralize all the data because otherwise it's just going to dep more and more yeah all data anywhere in the world for","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3291s"},{"start_seconds":3299,"text":"anyone that's cool love it yeah uh I just have one last question from from me and this is for mostly Alex","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3299s"},{"start_seconds":3309,"text":"and Alia I guess um so I have been talking to near team about the change sign nature recently and I really love","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3309s"},{"start_seconds":3316,"text":"the idea and it's quite exciting I think at least it fit the I mean our nft or our compute back nft really well the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3316s"},{"start_seconds":3325,"text":"concept that where like no matter which you're from you can easily treat theet without even signing a signature that's super cool","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3325s"},{"start_seconds":3333,"text":"but I'm more curious about like hearing from you guys uh what do you think will be the most like applicable or Mak sense","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3333s"},{"start_seconds":3342,"text":"use cases for chain sign nature in crypto nowadays I mean there's a whole like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3342s"},{"start_seconds":3350,"text":"already a bunch of people building on top of this Tech a whole set of different use cases um I mean it starts","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3350s"},{"start_seconds":3357,"text":"with wallets right the reality is a uh a number of wallets in your ecosystem already uh either deset or in case of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3357s"},{"start_seconds":3365,"text":"hot on mainnet um provide that experience where you log in you get an near account but then you can go and use","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3365s"},{"start_seconds":3374,"text":"for example base and applications of Bas right and you know there is a mapped account soal remote account that you're","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3374s"},{"start_seconds":3383,"text":"controlling but you know you're interacting from your new account but as this goes further right you can now have","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3383s"},{"start_seconds":3390,"text":"Dows for example that can you know facilitate uh various efforts across chains you can have Smart contracts that","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3390s"},{"start_seconds":3398,"text":"are trading that are moving assets around we're also exploring this idea we call Omni token where it's pretty much","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3398s"},{"start_seconds":3407,"text":"one token that exists on every chain at once right and so uh pretty much the kind of root of it is on near it deploys","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3407s"},{"start_seconds":3416,"text":"and erc20 SPL you know even brc2 on bitcoin Mark sosio there uh with","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3416s"},{"start_seconds":3425,"text":"information and now you can transact there but it's the total Supply is kind of Interlink between all these chains and you and the importantly you can","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3425s"},{"start_seconds":3432,"text":"actually transfer between chains um kind of you know with finality of uh of that","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3432s"},{"start_seconds":3440,"text":"chain plus uh like 10 seconds or something and so you can actually now actually have that format where you know","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3440s"},{"start_seconds":3448,"text":"if you are on one chain I give you my base colum IL tr. e you can just send to that address not caring which chain","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3448s"},{"start_seconds":3457,"text":"you're on and so uh so I think like that's kind of the experience you know we're getting into and obviously there's","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3457s"},{"start_seconds":3464,"text":"dexes there's lending protocols there's people who are using it for for example issues of cards because they w't allow","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3464s"},{"start_seconds":3471,"text":"users to deposit assets on every chain and then you can pay with a card everywhere so it's really kind of you know like a new","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3471s"},{"start_seconds":3479,"text":"piece of tech new building block in the web 3 that really kind of ties the blockchains together and allows to have","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3479s"},{"start_seconds":3486,"text":"this kind of composability that I would say didn't exist before between different","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3486s"},{"start_seconds":3496,"text":"chains yeah I think like all the use cases you mentioned makes sense any AI specific use case or scenario of your","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3496s"},{"start_seconds":3505,"text":"top yeah I mean same for like let's say AI agents that want to interact across chains right um that's something that people are already exploring kind of I","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3505s"},{"start_seconds":3513,"text":"mean from payments to actual interactions to transaction right like right now if you're in a agenda you need to implement all the chains that you","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3513s"},{"start_seconds":3521,"text":"want to interact on you need rpcs you need gas fees management and all of them right versus just you know go through your new account and uh you know pay","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3521s"},{"start_seconds":3530,"text":"fees in one way gets managed by the underlying stack um same for you know as","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3530s"},{"start_seconds":3537,"text":"we kind of rolling out some of the inference endpoint ideally you'll be able to you know fetch all all data from","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3537s"},{"start_seconds":3544,"text":"all the related chains and then if you want to actually uh for example mbase has this natural language transaction","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3544s"},{"start_seconds":3552,"text":"process right where you just say like swap but also you know build me a portfolio of you know top tokens across","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3552s"},{"start_seconds":3559,"text":"all chains um and you know Buy $100 worth of it right you can imagine pretty much like a whole whole set of","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3559s"},{"start_seconds":3566,"text":"transactions across different chains that get spawned for that to happen so I mean it it is more of a like","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3566s"},{"start_seconds":3573,"text":"execution infrastructure not like you know that AI can plug in into that yeah","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3573s"},{"start_seconds":3582,"text":"100% I really like the gas feature uh for chain sign nature that you will just automatically uh convert the gas token","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3582s"},{"start_seconds":3590,"text":"for super tool yeah well you kind of need that otherwise it's not very abstract if you need to know the implementation do the","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3590s"},{"start_seconds":3598,"text":"chain yeah definely all right I think we're almost at time so I want to thank you guys for","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3598s"},{"start_seconds":3605,"text":"coming uh for the discussion hopefully this was helpful uh I'm really excited kind of working with all of you and uh","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3605s"},{"start_seconds":3614,"text":"seeing some this opportunities for and users coming in in real life yeah thanks thank you so much for","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3614s"},{"start_seconds":3621,"text":"having us","url":"https://www.youtube.com/watch?v=ye_2k7dxtCE&t=3621s"}],"youtube_url":"https://youtube.com/watch?v=ye_2k7dxtCE","url":"https://near-pedia.xyz/videos/ye_2k7dxtCE"}