Fire side chat with Matt

Transcript

Yo. Yo, how's it going, Dan? Good. We decided to have a little fireside chat because we were talking and we never record these things and maybe something valuable comes from it. We're talking, we're both building a project, projects and they're kind of coming to a head now and, yeah, just some interesting times. I'm trying to start talking about it, I think. start talking about, yeah, it's Starkers and Dan. And yeah, it's not really an organized call, but I got a lot stuff going in the background, but maybe something interesting pops up. So, how are you doing, Matt? You there? I think we might have already last met. And yeah, it's all starting to come to a head here. Dan's been grinding away at AI for like a year now, and I've been grinding away for about four or five months now. And there's a bunch of things that's starting to align that we hadn't really planned, but all of a sudden they start to come together. So I don't know if you want to talk about what you've been building, Dan, or how much you want to go into it. Maybe we can talk about... Yeah, we cut off there for a second, but we got the good chunk of it. Well, I don't think we need to talk about what you've been building specifically necessarily, but we could talk about a bit of a vision for what can be done with the technology, you know, where we think AI and blockchain should go. Yeah, exactly. So, basically, I've been working on open source AI, a decentralized AI, I should say, rather. And it's funny enough because Matt's working on another project, and we're kind of coming to a head now where... Both of them could actually help each other. But just thinking of AI and how it can work, there's a lot of opportunity out there. And with the infrastructure that's built right now, and if there's enough body of people, enough of a network effect, I really think that the opportunity to create decentralized AI is not only is it an opportunity, I think it's the next evolution of AI, going down to rabbit hole. And I'm excited because the idea keeps expanding. It went from just coding to now we're expanding into, you know, source of the truth. And all this is, you know, it's new. It's rapidly developing, so not so much we can go into into too much detail. But, yeah, basically the centralized AI has been the hardcore focus. And Matt's been focusing on, you know, what we're going to call signal. just finding truth really a killer app. I think it's absolutely amazing what just the use case alone and what you can do with it. And combining that with AI, it's going to be interesting. I think, you know, we've had a lot of philosophical conversations about where AI is going to go and what we can do about it. And, you know, ultimately the concern is that AI is corporatized. And AI is becoming... this boxed-off system where you have to pay other people to access it, either to access the infrastructure or to access the actual models, and then the models become limited, right, kind of almost lobotomized in many ways. I think it's the way I see most AIs. And I think the technology, it's an interesting combination of technology, because it's like, how do you build AI outside the corporate system without allowing the corporate system to throttle it? which is what always happens, and then they obviously want to monetize it. And it's like, well, the guys that are experts in AI don't understand decentralization that well, right? So you've seen a few AIs come out recently, things like BitTenzer, where they, it's a great step forwards, but they clearly don't understand decentralization and governance and distributed governance and running distributed services systems and how to do that in an ethical and moral way without having a pre-mine or a corporation behind it. and obviously people on Hive our community, we understand how to do that very, very, very well. And so it's this combination of, well, if you want decentralized AI, it meet, and I think a lot of people in the AI world are thinking, oh, it's just a side thing, it's going to be simple. We'll just do AI really, we'll just do decentralized governance really easy over there as an afterthought, and it's like, no, it is, it is in the very core of the AI's progress to have it live in an ecosystem that is decentralized. And just saying that is very simple, but actually achieving that is incredibly difficult. And it's taken the Hive community alone, like years and years and years of work and reflection and philosophy to get to the point where it understands what decentralization actually is and can then act upon it. And so I think our background in that is very, very interesting when it ties into AI. So it's like, how do you govern the AI? How do you get node runners? to run nodes in a credible way, in a decentralized way, how do you distribute value to people who are doing things like running compute, providing resources and infrastructure, and providing information into the AI systems, because the AI has to learn, and if you just leave AI to learn on its own, it's just going to become a self-reflecting loop, whereas you need this human creativity and this human addition of information flow into the AI systems. and this is something we've realized over the last couple of months, and then bringing ways to inject that human influence into the AI, so it has both a mathematical ingress of data, so it's proving new mathematics, it's proving new code, but also it has this way of finding out what's human, finding out what's legitimate content, and what's reputable, and then bringing that into its knowledge base as well. And by doing that on a distributed database, you can... theoretically outperform the corporate models. And that's what, you know, right now, no, but at the time as we build the database, as we build the technology, you're going to get AIs, decentralized AIs that naturally outperform and out accumulate information when compared to the corporate systems that always have to gatekeep and always have to charge people for access and effectively lobotomize their AIs. So this is a very interesting world. And it opens the gates to people from outside of Hive to come in and start, bringing their expertise because everyone wants a place where they can run compute and test new models outside the corporate world, right? And that's where real AI progress is going to happen. It's not going to happen inside the corporate world unless we all want to live in this enslaved, weird place. Because, you know, the corporates are very clearly getting everyone using AI, getting everyone to write on AI, getting everyone to think with their eye, they won't do a single decision in their life without their AI. asking them or telling them what to do. And then it's like, well, if the corporations control those AI, it's a very clear control mechanism for humanity, and people are sleepwalking into this without realising, whereas you get AI in a decentralized ecosystem, and it starts to think outside the box, it starts to say and do things that maybe the corporates don't want it to do, maybe it helps people start thinking independently rather than thinking in the way that the corporate AI is wanting to. All of a sudden, you have this... sort of battlefront going on where it's like you've got the corporate AI versus a non -corporate AI and the decentralized AI. I have a lot of hope then for humanity and for the future if we can figure out ways to build decentralized AI models and use technology such as hard to keep you decentralized. Yeah, I mean, you said it very well. There's two things. Talk about why centralized AI fall and then talk about the long journey to why decentralized AI is actually the end point of everything. Because there's been a few times Stark and I have been like, oh, my God, everything we've done has led to this moment. One of the biggest ones was the speak network, just off-chain, we'll just call it off-chain storage. You know, there was a time where I thought that was the culmination of blockchain. Blockchain was the corner point, but the point of blockchain was to achieve censorship resistance off -chain. You can't get censorship resistance off-chain without blockchain. I'm not talking about immutability, you need blockchain for that. That was the biggest thing, because we realized how many things that unlocks. It's like, it's vastly more powerful than blockchain itself, which, funny enough, you can't have without blockchain, so obviously it makes blockchain more important. But once you've birthed, once you've got this pinnacle, but then AI came. So there's been these ladders where you feel like, oh my God, the pinnacle, the pinnacle, the pinnacle. And now I'm saying it again, the pinnacle. of all of these things since the beginning is now AI. You can't have decentralized AI without blockchain. You cannot have to centralize AI without off -chain, censorship-resistance off-chain. You can't have censorship-resistance off-chain without blockchain. So it starts with blockchain, period. Without it, you fail. And you have to do blockchain, right? That's why, if you wanted to make a new blockchain, you'd have to fork hive or its ideals. is the pinnacle of what blockchain is supposed to be. Now, yeah, there's little tweaks you can do. It's almost a decade-old blockchain. It was built before the AI movement. But the ideals don't change. The ideals in blockchain, or with AI coming out, it just reinforces it. P-to-P doesn't change. The ideal of P-to-P doesn't change. Now, you might use Rust. You might use a different tech, but the idea doesn't change. And these ideas of blockchain, they don't change regardless of the tech behind it. So the ideas that you need to get right, Hive got right. And that led to the moment of, okay, now we can do off-chain sensitive resistance. And now it's like, oh, now we have to centralize AI. Because why do you need blockchain and off -chain storage? Off-chain storage, that's where you store your brain, what you call RAG, which is called the brain, the brain of the AI. I've got a question to interrupt here. Why wouldn't you be able to just store all that on-chain, for example? What realisation have you come to here? Now, you could store it on-chain. The thing is, information for an AI is fluid. So maybe coding gets outdated, maybe things need to be trimmed, maybe things need to be refactored. So it's much more fluid off-chain. You can't have frequent data deletion and call it immutability. Immutability means nothing's deleted. There's no black or white. So that's what you want stored on-chain. So on-chain, if you think of it, so let's think of an AI. Let's say you have clod. It has a memory file, right, that it reads every single time. but when it gets into a chat. And then it has what I created called Second Brain. It's very popular on X right now. where you use obsidian, or I forked one for Hive AI made its own, but let's just say obsidian, and you store most of your stuff there, your project details, everything. You don't want to store that in a memory file, why? Because it's heavy, and the AI has to read that every single time, and maybe there's information it doesn't need to know and doesn't need to read every time, but it's forced to because it's in the memory file. Well, what if the memory file was just hardcore pointers and rules, so it read it and said, oh, now I know how to act, and now I know where to go, and I can search it. database by keyword, so I don't have to read the entire database. I have to read the entire memory file. There is no keyword search. That's the point of a memory file. You have to read every line every time. That's the memory file. Well, it doesn't make sense to store all a bunch of shit in there, especially stuff that's going to be irrelevant tomorrow to a lot of other tasks. You want that on the brain where you can mold it and shape it and do what you want with it, and you want the memory file to be hardcore pointers. Well, if you take that same similarity to blockchain, you don't want a bunch of shit on blockchain. needed tomorrow and then you have to weed through it because the blockchain yeah you can ping it but it's a hell of a lot more expensive and it's harder and you can't shape it once it's on there it's on there one of the things I don't really like getting into the AI portion and to do too many features but I guess I'll say it won, there's thing called the innovation game, TIG, very interesting for research and all that. One of the things, Hydei has taken that philosophy for the protocol itself, for the rag, for information. So you basically take your centralized AIs like Opus or Codex, whatever, or your local AIs, and then you mine crypto by improving the protocol. Well, you can't do it on a blockchain, right? Because once the information is on there, you have to say it's in its perfect last state. It can't get any more perfect. Well, that's not true in anything, really. I mean, yeah, and coding, you can reach pinnacles, but even then, what I thought compression could reach has been shattered, but what AI can do. So there's always, like, you might, oh, this is the fewest lines you can get it until tomorrow something else comes out, and it could be fewer. So you're handicapped in... yourself by putting your brain on a blockchain and saying, well, I can never change it. You want your brain to be able to learn and move and get smarter and leaner and more compact and fewer lines here and there. So that's why it's really a negative to store it on-chain. If you look at on-chain as rules, parameters, things that you want the AI to always know no matter what and they're never going to change, like rules, like the wheel is round. You need to know that if you're dealing with wheels. That's never going to change. You're not going to get a square wheel. So, and if you did get a square wool, you'll put it in your memory. So that's sort of the long-winded version. The other thing here, I mean, that's a great, it's a great realization, in my opinion. The other thing I... Oh, you cut out there, Matt. You get Matt back in a second. And those things could go on chain. Oh, you cut it off for about 10 seconds there. Right when you start talking, basically. Okay, I'll try again. So we're all here measuring objective reality, right? We're trying to, and it's a very subjective thing. It's a very isotoric thing when you get down to the levels of truth and what you're here to measure. But ultimately, you know, we can all agree that the trees are green, that the cars are, I've got four wheels, you know, those types of things. And as humans, we're kind of all here collectively measuring those things in reality and coming to a general conclusion as to what's real, what's true, what's correct, and what's history, what's historic. And when we come to that agreement, then we can have some mechanism that puts that on chain. And then we start documenting history in a way that's based on people's reputation, people's expertise. ways that they use to measure and record a document and the extra value that they bring, and then sort of have the accumulation of that, and then that point goes on the blockchain as a moment in history, and then the AIs can reference that as well. So that would be an example of something that you would put on the blockchain, the actual historical events and the historical moments. Whereas the things that are evolving, like the connections that are being... constantly made and re-engineered in an AI brain, that would be more fluid and that would need to go on a distributed off -chain storage system. Yeah, you need blockchain for the value sending rails, so the payments, and the immutable data and the governance. It's basically the wall that holds everything in. It's like the skeleton, and then you need... something like what we called, what people know is a speak network, what it's going to evolve into. It's much bigger than it, and what then it just was, it was just storage. Now it's much, much bigger than that. You need, that's the other, so there's three pillars. You have blockchain, you have resource sharing, storage, GPU, CPU, RAM. That has to be glued together. That comes with governance. There's a lot of things, right? You just don't throw that out there. There's a lot of ingredients to get that work and actually have a chance of decentralization. And then with that, because how do you pay these people? You need to pay them for work they're doing off-chain. That's what the Speak Network is about. It's about proving worth off-chain and getting paid from the blockchain by it for doing it. And then you move on to... the actual AI itself, which can then read the brain, which is on IPFS, it can then be incentivized by a blockchain itself, and now anyone anywhere can share and run resources and get paid for doing so. So it scales indefinitely, and it's really a fine game because everyone is willing to share a little bit of GPU to get a better product. So you don't even really need a business model because everyone's just going to plug into it anyway and get that better model. And with it, every user, so, like, every user codex gets, it helps its network effect, but it hurts its bottom line. Because they, for me, they're losing so much money. Like, I'll show you guys just a little trickle what's going on. That's been running for probably two hours now. They don't, they hate me. I milked that $200. Probably $10,000 a month for my resources. They're losing an absurd amount of money. They're not, so bleeding into why centralization is going to fail, or centralizing is going to fail in the free market guaranteed 100%. It's a terrible idea. It's a terrible business idea. It is the worst business idea you can ever think of. The only reason to do it is you want to be a government-controlled sci-op to control the masses where the money doesn't matter because you're printed anyway or you already have it from taxes, whatever. The people don't care about money because they make money. They literally make money. The definition of making money, not just making money in the wallet. Also, I'd argue as well, just to come in here, is the, it's a control mechanism, it's a thought control mechanism, right? Centralised corporate AI is being used to mould the thoughts of the people based on the advice it gives them, right? So it's like, if everyone's got this philosopher in their pocket and everyone's using it to sort of say, oh, this is how I feel about this, how do I improve myself, what do I do here? Please write me with this message, please do this for me, please do that for me. it starts to become a reliance of dependence. Exactly. And if you control the rails upon that, upon which that dependence is built, then you can control the thoughts of humanity. That's what this looks like to me. Imagine generation raised up GDP, CHETGTP. Yeah, yeah. And only imagine that's getting tighter and more constrained. I'm already doing things. It's already dead, right? It's already there, bro. And now it's only going to be, oh, no, we can't do it because it's a hate speech. Like, you know, you already see them, like, oh, no, that philosophy. So, very dangerous. I mean, to people like us, it's obvious what they're trying to do with it. Like, hell no, we're not letting that happen. Like, we can see that coming from a mile. We saw what they did with Facebook. Imagine what they did with chat GTP. You were out of your minds, on steroids, half-bred monsters, the worst nightmare you can ever imagine centralized AI that's used in a negative way. And ooh, maybe they won't use it a negative way, but I don't know, I've seen, I've seen an island, I've seen a bunch of shit, not personally, but I've heard of one, I've seen a lot of shit out there, I don't trust these people, there's no way in hell I'm going to allow this thing to really take over. Decentralize AI is a must. It's like a blockchain was a must, and then…

Comments

More videos on 3Speak

• Live Demo unseres KI Avatars, Frieda für Pflegeheime by @michaelreischer
• Back to the Classics: My First Crash Bandicoot Gameplay at N. Sanity Beach! 🎮✨ by @diegoelpro
• Proses Pemancangan Tiang Beton Menggunakan Hydraulic Hammer by @almahdi-014
• Absent Minded - Sm Bu Pejen by @seckorama
• Hive IDE v2 by @theycallmedan
• Pacific Warships Gameplay ⚓🚁 | Epic Naval Battles & Warship Combat⛴️🚀 by @dxjihad
• The Peace River Valley...With Fog!!! by @jongolson
• Car Showroom by @devid.araf
• The Legend of Black Hand Riku. by @efootjoker-yt
• The Daily Steps #01 | 🌿 Exploring Waduk Setu on Foot by @bulkovam
• Sunset at Playa Pilar 🌅 | Atardecer en Playa Pilar by @adonistrujillo
• Stickit: Sticky Notes, but make them Hive Powered by @meno