NEXUS Startup Competition Pitch - Hajime AI - TOKEN2049 Singapore 2024

I'm uh leave I'm representing hajim AI so hajim is Japanese in Japanese this word means first so if you visit Japan for the first time you'll learn hajim Masa so means our first time uh seeing you so we want to become the first AI company to witness the rise of this one person billion dollar company for the first time so this vision is not from us this is from our old friend Sam Alman so he is the guy behind this Vision so he believes that uh you know one day there will there will be like a one person billion dollar company where the CEO is a you know a human being but all the employees are AI agents like AI marketer project manager AI auditor AI software developers these are all AI agents so you can hire them and you can launch a billion dollar company on your own so actually in the AI industry we already believe that AI agent is the future it's the consensus you know because AI agents are going to just become our mobile apps they you you know these agents will outnumber and outsmart us but the real question is how to make them work together how to make this multi-agent nwork work as Andrew you know Mr Professor Andrew n said uh you know they are already very efficient but there are ought to be ways to really orchestrate them and make them into a workflow that serve real world purposes and these are some challenges this multi-agent system is also still very early uh these air agents they operate in silos they don't talk to each other don't learn from each other they don't really communicate that much and there's also a web to structure where you know open AI can be launching your product tomorrow so this always a existential existential crisis for AI agent startup what if AI uh what if open AI releases your product tomorrow so we want to um you know it's very important to protect the proper data and professional domain knowledge for these AI agents and also it's hard to co- start there are already around like 26,000 uh AI agents uh building uh you know uh for the last two years so these are the two problems in uh it's stopping us from embracing the uh multi-agent future so what what is to be done we believe web 3 is really a good answer to this uh dilemma so if this AIA agents are operating in silos we should make them collaborative and we will uh we will build a multi-agent graph to make them work together which I will show you in just one minute and if there's web to Monopoly sure we can introduce a democracy by building a decentralized network and if there's lack of propietary uh uh and private data protection we can do that with ENT encryption technology I've seen a lot of uh teams building on F and te and you know where uh there are things that we can can collaborate on that and also it's hard to co- start it is where we can introduce more web three boost strapping mechanism and uh you know we can introduce human uh human uh feed feedback to help build up a leaderboard for AI agents where you can actually discover AI agents and you have a score that where you can rely on when you when you you know make decisions whether you hire this AI agents or not so this is uh common practice in foundational model uh you know when people uh give scores to ational model but it's still not available for a agents so haimi AI we are building this decentralized multi-agent graph dmac for collective intelligence so this is our uh small demo we just this is um you know um just out of Aven so I will just show you so basically we are creating a workflow where we can actually create a news reader so you this this reader will read news for you every morning and you know read it read it out loud so uh here is how we actually choose the AI agents in our Network so you need a reception you need a news croing you need summarization need simple chat you need a uh text to speech and you find a way to and we our system also help you find a way to connect them together in a workflow you see these lines connecting them and you define the base features you uh Define you know what this uh what are the prompts that these uh these agents will react to and then you see like these three agent uh and all these um agent will work together and give you these answers in that chat uh in that chat box and then you can just give a you know a URL to this chat box and then we'll read out the news and summarize everything for you so this is like Point uh one Z and the whole idea is to automate in the future so our value proposition is like we are connecting the agent users with very you know real world intent with uh a agent providers so we can connect any agent providers and we're going to unboard them orchestrate them and also you know get make sure that they're getting paid so we're building Al layer one we're building a s Chen uh which means that we can also be composable to a lot of like you know um service provider along the a decentralized AI stat like you know decentralized GPU we are in partnership with a year modular da uh zero gravity and Onan storage there are a lot of like providers so we can make sure that this decentralized as possible while we run our s small POS consensus alongside the Solana layer one so we can scale up Solana uh you know for AI operations but without breaking up the liquidity so these are our stack you know we have dmac which is AI Hub is LinkedIn Plus app store and we have this AI side chain for salana with haimi AI layer uh that's both like POS and fhe and also we have this initial offering which we will be able to um uh issue agentic assets on Chen on Solana so that's why we're building on layer one because we want this uh you know agentic assets to really enjoy the liquidity on Solana layer one and we have Benchmark Dow which is the leaderboard that I just talked about it's very exciting uh we want to replicate that we're going to introduce that uh as a web3 company so this is how web3 will be able to help Ai and you know just to crowdsource feedback and uh on the AI agents so we have Team uh G is a Serial entrepreneur based in Silicon Valley uh very close closely related to uh Berkeley and Stanford and Har and Melody is the number one uh resourceful person in Japan's web three world uh incubator of stepan uh fusion list and some other like FDX Japan and Nova is from IBM and we have also onboarded advisers from uh you know from pogon JD is a pogon uh pogon uh funer and Scott is funer of uh uh gcoin also on the board of ZK sync and Robert is uh with many Web Two background so making wave here uh we are uh we want s rone hone uh yeah thank you going to move down to Q&A now to the judges sure yeah thanks that was a good presentation uh can you touch on the the private data side of things you mentioned like tees and fhe um my understanding is like you'd have to get the private data from somewhere so yeah what kind of data are you working with here that's uh Pro yeah so uh agents are like model plus a rag right like the retrieval augmented uh graph so that the rag is where you store the private data and the the r we will need them to be uploaded uh to our Network so which means that that's where we need the you know the F and the te to help us because and we we want to pro pro protect the privacy so uh but but these racks these are the you know uh these are the data and knowledge that make AI agents uh different from you know generic AI agents built out of out of the box by uh you know open AI CH gbt got and like who would be uploading this data would it be like uh the agent provid the agent uh developers so um yeah so you they can be web2 or web 3 they are web two companies who already are building AI agents and providing like sdks so we we can be composable to that as well got it thanks yeah I'm are you guys reading the blockchain I guess like to me I'm trying to understand the connection between the actual blockchain and the AI technology that you're using sure uh I think it comes out of this necessity to introduce blockchain uh because you know a lot of the AI agents they uh so the really useful AI agents are buil by professionals or like people can actually know you know how to write how to create how to become an influencer on on Twitter so uh but they don't want their uh you know data data uh their database or the knowledge to be shared right out to people like open AI or Microsoft so that's why we want to uh create um this decentralized network where you know we can the noes can check whether they racks uh I just mentioned the the rack is the database the private database so we want to make sure that are not getting stolen from the network and there and also the information itself is protected by encryption like fhe or uh te awesome thank you so much thank you