Space Summary
The Twitter Space deAI Data Access & Intelligent Web3 Oracles w/ @AutonomysNet & @GenLayer hosted by AutonomysNet. Discover the groundbreaking insights into deAI data access and intelligent Web3 oracles in the AI niche. Explore the significant role of @AutonomysNet and @GenLayer in shaping hyper-scalable solutions for dApps. Pantera Capital's support signifies a strong commitment to innovation in AI and Web3 technologies. Dive into the advanced features of DePIN and how it enhances storage, compute, and consensus in the Web3 environment. Gain valuable knowledge on the impact of intelligent oracles on AI advancements and data accessibility within Web3. Join the conversation on the transformative applications of AI3.0 in decentralized systems and the collaborative efforts driving progress in AI and Web3 solutions.
For more spaces, visit the AI page.
Space Statistics
For more stats visit the full Live report
Questions
Q: What role does deAI play in shaping AI3.0?
A: DeAI serves as the foundational layer for AI3.0, facilitating advanced data access and intelligent oracles.
Q: How do @AutonomysNet and @GenLayer contribute to Web3 development?
A: Both entities are instrumental in creating hyper-scalable solutions for building decentralized applications and on-chain agents.
Q: Why is Pantera Capital's support significant in this space?
A: Pantera Capital's backing signals strong support and investment in innovative AI and Web3 technologies.
Q: What are the key features of DePIN in the Web3 ecosystem?
A: DePIN offers cutting-edge capabilities for storage, computing, and achieving consensus within decentralized networks.
Q: How do intelligent oracles impact AI advancements and data accessibility?
A: Intelligent oracles play a crucial role in enabling advanced AI capabilities and seamless data access in Web3 environments.
Q: How can AI3.0 revolutionize decentralized systems and smart contracts?
A: AI3.0 introduces innovative applications of AI in decentralized platforms, enhancing functionalities and efficiency.
Q: What insights does the discussion provide on the future of Web3 with AI integration?
A: The conversation explores how AI technologies are reshaping Web3, unlocking new possibilities for decentralized systems.
Q: What benefits can AI bring to scalability and efficiency in decentralized apps?
A: AI integration in decentralized applications can significantly enhance scalability, performance, and operational efficiency.
Q: How are AI capabilities being integrated into blockchain networks for improved performance?
A: Tech experts are collaborating to embed AI functionalities into blockchain systems, enhancing overall performance and user experience.
Q: How does the collaboration between experts drive advancements in AI3.0 and Web3 solutions?
A: Collaboration among technology specialists fosters innovation, leading to significant advancements in AI3.0 and Web3 developments.
Highlights
Time: 00:15:42
Innovations in deAI for AI3.0 Exploring the latest advancements in deAI that shape the foundation for AI3.0.
Time: 00:25:18
Hyper-Scalable Solutions by @AutonomysNet & @GenLayer Discovering how @AutonomysNet and @GenLayer are revolutionizing dApp development with hyper-scalable solutions.
Time: 00:35:50
Pantera Capital's Backing for AI & Web3 Innovations Understanding the impact of Pantera Capital's support in driving innovation in AI and Web3 technologies.
Time: 00:45:29
DePIN: Advanced Web3 Capabilities Insights into the advanced features of DePIN for storage, compute, and achieving consensus in the Web3 ecosystem.
Time: 00:55:11
Intelligent Oracles in Web3 Exploring the significance of intelligent oracles for AI advancements and data accessibility in Web3 environments.
Time: 01:05:42
AI3.0 Applications in Decentralized Systems Understanding how AI3.0 is reshaping decentralized systems and smart contracts for improved functionalities.
Time: 01:15:20
AI Integration in Web3 Exploring the integration of AI technologies in Web3 networks to enhance scalability and operational efficiency.
Time: 01:25:37
Tech Collaboration for AI3.0 Advancements Insights into how collaborations among tech experts drive progress in AI3.0 and Web3 solutions.
Time: 01:35:44
Future of AI in Web3 Ecosystem Discussions on the future implications of AI technologies in the evolving landscape of Web3.
Time: 01:45:19
Innovative AI-Driven Solutions Exploring innovative AI solutions that are reshaping the Web3 ecosystem for increased efficiency and capabilities.
Key Takeaways
- deAI is shaping the foundation for AI3.0 with cutting-edge data access and intelligent Web3 oracles.
- @AutonomysNet and @GenLayer are key players in developing hyper-scalable solutions for building dApps and on-chain agents.
- Pantera Capital's backing signifies strong support for innovation in AI and Web3 technologies.
- DePIN offers advanced capabilities for storage, compute, and consensus within the Web3 ecosystem.
- Understanding the importance of intelligent oracles in enabling AI advancements and data accessibility.
- Exploring the potential of AI3.0 and its applications in decentralized systems and smart contracts.
- The discussion sheds light on the intersection of AI, data, and blockchain tech in shaping the future of Web3.
- Insights into utilizing AI technologies to enhance scalability and efficiency in decentralized applications.
- Innovative approaches to integrating AI capabilities into blockchain networks for improved performance and functionality.
- The conversation highlights the collaboration between technology experts to drive advancements in AI3.0 and Web3 solutions.
Behind the Mic
Introduction and Welcome
Hey, everybody. We're going to get started here in a couple minutes. Just letting people file in. I hope everybody's week is going well. We're coming to you live from Salt Lake City at permissionless. So really excited to get started here in a couple minutes. Is anybody here also in Salt Lake City this week? Give a thumbs up in the chat if you are, because I'd be happy to meet up. I'm going to be at the conference and a couple other parallel events today. Okay, we got one clap. I'm guessing that doesn't mean that you're in Salt Lake City, but that's okay. We'll get started here in probably a couple minutes. Just letting a couple more people file in. And we've got Albert already here, so in a minute or two, we'll get started.
Engaging with Participants
Hi, everyone. Hey, Albert. It's great to have you join. Yeah, thanks for having me. I'm excited. Absolutely. Yeah. We're going to treat this as a natural, free flowing conversation and just talk about gen layer synchronicities, data access, and really try to dig into the genlayer stack. Some of the use cases, stuff with the prediction market, different ideas that you have, stuff with the CLI, maybe talk about the simulator, a little bit of genlator J's. Just some of the core concepts here in a minute. And really, the goal is just to get people really excited about gen layer because it is a really cool project. It is pretty unique in the space, and it has a lot of really cool, unique ideas. So I'm really excited to talk about it and have everybody learn on the call.
The Importance of Education in Technology
I appreciate it. Yeah, we really, you know, like, education is such an important component. Right. And it takes effort and it's difficult. Right. Like, we're talking about very interesting technologies, you know, like crypto and the AI and identity, so much stuff. Right. And it's. It's really. It's difficult to get the message across. So it's important that you guys are doing this education effort very much. Of course. Yeah. I mean, the messaging is everything, right? I mean, you can have. You can write the best book in the world, but nobody reads it. It's just kind of a mess. Right. So, I mean, that's the same thing with technology. It's like some of the best ideas that happened over the last few decades were ultimately misses because of the communication. So the more we can communicate about technology and get people excited about it and have them realize that this isn't just possible.
Gen Layer Introduction
It's something that can actually really help people and change the way we think about things. It's major. So we have some people here, we can keep filing in and we can get started. Yeah, I mean, welcome everybody to this really exciting Twitter spaces. I've been really looking forward to this all week and we're featuring our new partner, Gen Layer, who's going to be building on top of the autonomous network here pretty soon. So to get started, autonomous is building the foundation layer for AI 3.0 and we're going to be providing a hyper scalable, decentralized AI infrastructure stack solution. Our network is going to be combining high throughput, permanente distributed storage, data availability and modular execution to enable the development and deployment of advanced AI applications and on chain agents.
Innovative Partnership with Gen Layer
So with that said, today we're thrilled to introduce genlayer, which we found and we had seen previously a few months ago, and we're really excited about it. We're like, oh my gosh, we have to partner with these guys. Really innovative project that really aligns with our vision about a future of decentralized AI, giving ownership back to individuals, you know, decentralizing control, and generally is really pioneering this with the concept of these intelligent contracts, which we think is definitely a step order advance beyond traditional smart contracts. So we really think these intelligent contracts, you know, leveraging LLMs, and I'm sure everybody on the call has used LLMs extensively in the last two years. How can we combine that to access and process Internet knowledge and make really good decisions in a decentralized manner?
Albert's Background and Introduction
So with that said, I'm really pleased to introduce Albert. He's a co founder of Genlayer. So Albert, welcome. If you want, you can tell us a little bit about yourself, how you got started, and then we can start to dig into some of these ideas and see where the conversation takes us. Awesome, thanks Chris. So yeah, maybe I'll start with a bit of background. Right. So personally, my background is in computer science. I've been in the space for quite a while. Since 2013 I've been in multiple projects. Two of them were l one s nem and Radix Dlt. I've been in other projects as well, such as the budget, and I was with the co founder and CEO at Stakehand, which was the first liquid staking solution. Now we've seen what is possible through these led sanguits models.
Understanding LLMs
We're talking about AI, we're talking about LLMs. For those that don't really understand what LLMs are, you can think about. It's like a digital brain. So it's essentially imagine a very big spreadsheet that is capable of. You send it a sentence and it outputs something else that makes sense from that sentence. I'm assuming that all of you have used SAT DPT, so you more or less understand what the concept of this is. A couple of years ago we started looking into this when it came out SatGPD, and were completely fascinated by what's possible by it. You basically have a black box that you can send in something and it responds, and it can be responding very intelligently. It can even be responding code. When we discovered this, when we saw that you can actually have an LLM spin out code and tell you how to code something, we started to explore all the different possibilities that this enables.
Challenges in the Blockchain Space
What can we do if we basically have a piece of software that is capable. Did we lose Albert? Can anybody hear him? Thumbs down if you guys can't hear him. I see. Yeah, I don't know. I think it's ok. I think it's like the screen goes off and then it stops. Am I ok now? Yeah, you're good. Keep going. I don't know how to make it like. Yeah, right. Ok. So essentially what we're talking about here is, look, there's thousands of blocks out there, but the reality is that in general they have the same type of features, they have some limitations, such as blockchains can't connect to the Internet. That's why we have articles and that blockchains can't actually make decisions on themselves. You think about smart contracts, and smart contracts are not smart, they're pretty dumb.
The Concept of Intelligent Contracts
They just run code. And that's it. What we're doing is we're creating the first intelligent blockchain. We say that because our blocks, and essentially the consensus algorithm is using artificial intelligence. That means that each validator is connected to a different LLMDH to essentially reach consensus agreements on non deterministic outputs. What this means is that we can have smart contracts that can natively connect to the Internet without any oracle whatsoever, and that can make decisions such as processing natural language, understanding this natural language, or even code, and then making decisions that are basically an aggregated output from multiple different validators, each one of them running a different Las language model. So it's a completely different type of architecture.
Expanding the Design Space for Decentralized Applications
And with this, the goal is, can we open up the design space for decentralized applications? Because again, we've seen basically that for the last few years, since 20 1819, everything is pretty much the same right now. Everything is just, you open a new l one or a new l two, and there's a flock of the same applications copy pasted from other protocols. We want to see how can we really enable new types of decentralized protocols. Awesome. Yeah. I mean, when I was reading through your white paper, and I definitely encourage everybody on the call to first take a look at Genlayer's website because it's really awesome. Links to the gen layer simulator and some ways to get started with the CLI. But, you know, yeah, you touched on this point of the consensus mechanism being this optimistic democracy where if one LLM might hallucinate, then the likelihood of the others that are being used as multiple validators to interpret the data would be able to catch that hallucination.
Skepticism Surrounding Consensus Mechanism
So that was something that I think that when people were initially reading it, they were like using large language models that are prone to hallucination for something that's supposed to be precise, like consensus, you know, led to some skepticism. So. Yeah, it's a very big jump. Right. You think about it like all blockchains in general. Right. They always, well, they all have been completely deterministic. Right. So again, deterministic just means that, you know, for a given input, you will have the same output. Well, like given another output. Right. But it doesn't change. Yes, this is a big difference. Right. When you're looking at, you know, an LLM, every time that you ask an LLM, you're getting a different response.
Explaining the Stochastic Nature of LLMs
It's a stochastic system, but probabilities. Yes. And, well, how do you bridge that gap? How do you make it so that you can, from a blockchain, connect to an LLM? There's different approaches to this. One of them is called proof of inference, which is, hey, what happens if you have a single LLM and then you do the evaluation like the thinking of the LLM was done correctly? The problem that we see with that approach is that then you're centralizing the decision making on a single model. And that basically means that, well, if this model is wrong, you will get a wrong output, and that's not good. If there's a way to attack the model, prompt injection is called, then you can attack the application.
Wisdom of Crowds Approach
That's not good either. So our approach is essentially the wisdom of the crowds. Can we get a bunch of different models to agree? And even the majority vote, they agree that the output is correct. Then it's correct. We're not trying to verify that the thinking was done correctly. We're trying to allow them to agree on anything that they want as long as, you know, as they can agree. Right. Which is the tiny bit. Does that make sense? It does, yeah. I mean, that gets into the whole appeals process, and that's kind of what you're touching on. I mean, is there anything else you want to maybe mention about the appeals process? Because it is kind of. It is a unique aspect of gen layer.
Understanding the Appeals Process
So, I mean, ultimately it's based on this thing called the Judy theorem, right. Which is, you know, instead of having a single judge deciding your fate, what you want to have is you want to have a bunch of judges and they all together will get a better answer than a single judge. Right. If this judge slept bad, you know, and didn't eat well and so on, maybe it will be 80% correct. Right. You don't want that, right. You want multiple of them to agree on something, and then the appeal process is okay. Even if the five judges that we're selecting at random for the first iteration of the transaction, if you don't like that, you can still appeal to then bring more validators to the mix. So ultimately, if you appeal and appeal, then you have the entire network verifying your transaction.
The Cost and Complexity of Appeals
That's more expensive, but you can do that, and that gives you the best possible answer because that's how the whole network has agreed on a majority on a specific output. So the appeals require some bonds, and it's a bit more complicated, but essentially that's the idea, right. If you disagree, it's like a court system. You can essentially raise the flag and just go to a higher court and then complain there. Yeah, that's really interesting. I think let's take a step back from the appeals process and maybe talk a little bit about the intelligent contracts, because this is a big part of genlayer. And let's maybe break it down for the people on the show here.
Understanding Intelligent Contracts
How did you initially think about smart contracts? And how did you lead yourself to thinking about intelligent contracts as being this key innovation and what possibilities they could potentially unlock for gen layer? What do they do? What can they do and unlock that smart contracts are inflexible and incapable of doing? Right? So, I mean, when you're trying to build a smart contract, that's just code, right? So you have solidity, you have Ras.
Understanding Smart Contracts
There's different approaches to writing smart contracts, right. They have some limitations. One of them, which is not obvious, is that they can't connect to the Internet. Right. We're very used to writing software and things, and, you know, your phone is connected to the Internet all the time. Right. What can you do with your phone when it doesn't have Internet, much less. So why don't smart contracts have connectivity to the Internet? And the reason is basically because the blockchains are, as I said before, deterministic. But when you're going to the Internet, every time that you search for a website, you will get a different website, the cookie has changed or the content has changed or whatever it is. That impossibility to make sure that the website is always responding the same is what makes smart contracts not be able to connect to the Internet, right? That's why they need oracles. And oracles are enabling tons of use cases, right? Thanks to oracles, for example, you have perpetuals and lending and events of other things, right? But oracles are very limited too, very expensive.
Limitations and Potential of Smart Contracts
You can only deal with price data and stuff like that. So the question is, what if you have smart contracts that can natively connect to the Internet, right? So then you would be able to, for example, do, let's go back to the perpetuals use cases, right? Which is you could do perpetuals on anything you want to do. Collectibles, you can do perpetual markets on collectibles, or you can do insurance on, you know, the weather conditions, or you can do, you know, prediction markets on basically anything like peer to peer betting on literally anything that you want, as long as the intelligent contract can go out to the Internet and come back with a decision, basically. So this enables for example, decentralized autonomous organization. So if you look at it like daos are super powerful, right? I think like most of, like many of us actually were totally in love with DAos when they came out, right? 20 14, 20 15, 20 16.
The Role of Decentralized Autonomous Organizations
The problem is that, well, right now they're not real Daos. Right now they're just a multisig, there's just a bunch of people behind the multi sig listening to what people are voting on snapshot and doing that, right? But what happens if you have a contract, an intelligent contract capable of going through snapshot, understanding the proposals and the votes and the weights, and then making a decision without any human in the loop at all, making the decisions, completely autonomous, a truly unstoppable, decentralized, autonomous organization. So this really changes things, to be honest. So you can do anything from here. You can do generative memes, you can do market makers that adjust their fee depending on the liquidity of a centralized exchange, for example, or a stablecoin that can change its properties depending on the policies of the central bank.
Exploring Digital Identity
You can do digital identity as well when it's going I think here's where autonomous has one of these use cases that I think will be very much in line, which is, can we essentially create a system for digital identity where basically you're mixing social information with cryptographic information. Right. Does that make sense? Yeah, it makes perfect sense. And you're touching on what I was going to. I wanted us to dig into prediction markets. I mean, I think a lot of people on the call and maybe some of you have participated in this using polymarket for betting on the election, betting on the recession, betting on something that's happening. It seems like, it's like a bread and butter situation where it seems like these intelligent contracts and the flexibility they provided really lend themselves very well to prediction markets and a lot of other things in crypto that are more flexible or prone to change.
Future of Prediction Markets
Where do you think prediction markets are headed? It seems like building with genlayer and building a prediction market utilizing aspects of gen layer might be something that could feasibly happen in the near future. Curious what you think. Can everybody hear me? Yes. No. Okay. The autonomous host can hear me, but I can't hear. Albert just approved Albert to speak. Yeah, no worries. Did you hear my question and some of my thoughts around the poly market? No, I think you meant to repeat that. I think you meant to repeat that. Okay. Okay. I'll do a tldr for our friends. Apologies. No, no worries, Albert. No, I'm just curious where you think prediction markets are headed. I mean, a lot of us have participated in like, polymarket and messing around with stuff like that. It seems like these intelligent contracts are really built with the ability to be able to.
The Evolution of Trading
You could build something more natively on a market like that. Where do you think that's headed? I think that it's going to be much more open. Means that anybody will be able to create prediction markets. Anybody will be able to create prediction markets on many more things. So right now there's a very big limitation, which is that the resolution needs to be, Orlando, totally deterministic, meaning, for example, just numbers. Right order resolution needs to be managed by humans. Humans are slow, humans disagree, and they have misunderstandings on what's happening. Did the ETF get approved or not? The problem here is that right now, all these human intervention essentially is an overhead. That means that every single market that's being resolved can be costing whatever, hundreds, even thousands of dollars to resolve properly.
Cost Challenges in Prediction Markets
That's really expensive. But if you want to make, say, for example, an insurance on flight, for example, well, that's the sort of prediction market you're spending 100, $300 on a flight. You can't be spending hundreds of dollars to resolve the insurance prediction market. Insurance is just different type of options, basically. And what this means is that with intelligent contracts, you're going to be able to natively go out to the Internet and fetch some information, which is just a few lines of code. So that takes the price of resolving a prediction market from tens or hundreds or thousands of dollars down to less than a dollar. And this means more prediction markets, more products like this. And the more products there are, the more infrastructure that gets created and the more you can have more layers of complexity that add value to what's actually being built there. Right?
Insurance Opportunities
So for example, one of these cases is insurance, right? Right. Now there's no proper insurance on chain. Why? Because it's managed by daos, right. Can we essentially have insurance on slashing? Why not? Can we have insurance for bigger or smaller things? Right. We will be able to do that when we have this infrastructure in place. So, yeah, that makes sense. We're open there. More openness, right? Yeah, more openness, more flexibility. We're seeing the same parallel right now with agentix in the AI space. We're seeing more platforms that are opening up that are just allowing for agent based platforms to be built in, agents that are more flexible, that have access to the Internet.
The Rise of Agent-Based Platforms
This year, perplexity existed and people were using it last year, but perplexity has really blown up more with the ability to have references. And so I really think this year, next year when we see, we're going to be seeing like, you know, I was at the DAI conference yesterday in SLC, and the big core focus for a lot of people were like agent marketplaces being able to sell, you know, having these sort of, and then that ties in perfectly with this notion of intelligent contracts where if you have agents or agentic platforms working with intelligent contracts, that there's additional flexibility there which leads to better insights, better information, better recommendation systems, stuff like that. So I mean, have you, I'm curious what you think about, you know, this idea of agent platforms. I know, I know weren't planning on talking about that, but are you excited about Agentix and excited about that?
Thoughts on the Future of Agents
Yeah. So there's different. So two thoughts there, right? One, I would like to go back to your point on like, you know, how did you actually come back like with the idea for general? It actually relates very much to autonomous, which is if you think about it, why is there no open database that essentially, people can throw information, and the database is capable of verifying the information and storing the information if it deems it correct, valuable, verified, you see, why can't we have a totally open Wikipedia where people can throw information and the network itself goes and verifies that information? So generally, there is essentially this consensus that is capable of taking in the data, verifying the data, and then it needs to store it somewhere, and it can store it in an autonomous. Right.
Vision for Open Data and AI
So that's just one highlight on where we actually came from, the other one. And the question for the agents, we've actually been working on agents for now almost two years. We think that the future is going to be full of them, right? Like, really, my opinion is like, probably by first quarter of next year, open a is going to be releasing a product to just, you know, have an agent that's working 24/7 for you, whatever, pay $50 a month, get an agent that is like doing your stuff and, you know, do something, comes back a few hours later, it's done whatever you needed to do, right? And I think there's going to be like many, well, like billions and billions of agents.
Value of Agents in Future Economy
The question for me is, well, where is the value? There is the value on the code that is being run by the agents. The way I look at it, code is very cheap right now. Anybody can write code just with an agent. I don't see that the value will stay there. The value will be for sure in data. That's one thing that is very difficult. If you can essentially isolate the data in a way that can be like capturing value. That's one. And the other thing where I think the value will really accrue is on the infrastructure, not really the physical infrastructure, which I think will be commoditized, although extremely used.
Synthetic Jurisdictions for AI Commerce
But actually, from the challenge perspective is, can we build a synthetic jurisdiction? Imagine like a nation state for agents to be able to do commerce. So when you have your agent and I have my agent, right, you have your AGI. I have my AGI. Are they going to be dealing with each other? If they want to trade something, are they going to be hiring lawyers? Very expensive, very slow. Are they going to be just trusting smart contracts that can only run code and can't even connect to the Internet? I think that's too limited. The way we look at genlayer from the point of view of these agentic systems is genlayer is essentially a synthetic jurisdiction that can have regulations, can even have taxes or insurance, can have tons of different aspects of any commerce network so that these agents can deal with each other.
Future Vision for AI and Human Integration
That's really the long term vision, which is how do we enable this? How do we enable AI to AI commerce in a way that it really, as well, allows humans to take part of that. Right. So, you know, yeah, all these decentralized protocols that we can enable, we think that it's going to be first humans building them, and then people are going to be able to capitalize on those when AI's are going to be using them to trade. Right? Yeah. I mean, somebody had mentioned yesterday at the DAI conference that he said something like, oh, you know, before I was using agents, or agents that I had custom built for doing trades on different DeFi platforms, I was doing maybe ten trades a day, 20 trades a day.
Impact of Agents on Trading Efficiency
Now, with some custom built scripts and these agents that I've built in house, I can do 300 trades a day, 400 trades a day. As we get to a point where it seems like, I mean, train of thought reasoning is getting us closer to AGI. And there's obviously a lot of open thoughts about, you know, are there groups that have reached AGI internally? You know, there was always the thought that OpenAI had already achieved AGI internally, and there was a reason why people were leaving. There are also thoughts around, well, it's becoming more for profit. That's a reason why people are leaving the company. But I think what you're kind of hinting at is, as we approach, we're in this new age of AI, and as we approach the dawn of AGI and eventually ASi, we're getting to a point where humans are really going to need a lot of help with understanding information on the Internet, making sure information is real, that leads into content, provenance, and understanding authenticity of where information is coming from.
The Role of Custom-Built Agents
But it seems like in the very near future, it'll be very conducive for people to have custom built agents, or agents that they control, that they can use as guides. And exactly like you said, having a VM that you pay per month to have, that you can access from anywhere, and that. And that VM can be controlled by an agent or a set of agents on an equilayer, or a hierarchical layer of agents where they either talk to each other, make commands, make controls, and that'll really end up dictating people's lives. And that kind of leads to the whole notion of the post labor economics, where as the rise of intelligence happens, there's an equal drop in the cost of labor. So it'll reach a point where it's just going to be cheaper and cheaper to do things on the Internet.
The Future of Work in an AI-Driven Economy
Eventually, that'll equate to robotics and being able to make robots that are cheaper to higher than people. And so that leads into the whole idea of building infrastructure for post scarcity and having individuals be able to own hardware that'll allow themselves to actually earn a living, at least in the near future. So have you thought at all about, it? Seems like intelligent contracts, they're more flexible. So that's going to lead itself to being more adaptable for agents, which is going to lead itself to being more adaptable for this new age of AGI. You also touched on some notions of the network state, how, as we approach AGI, government systems that currently exist globally are ill equipped to handle the exponential change that's happening in societies.
Challenges of Existing Governance Systems
And that, you know, the network state ideology kind of postulates that it's getting to a point where there's already. The Internet has already led to an erosion of barriers for culture. You know, we people on this call can be all over the world, but they're hearing me simultaneously. Something that would have been considered magical 100 years ago is now something that we just do on a regular basis, and that's propagating to all levels of technology. So, I mean, what do you think about, like, post scarcity and AGI and how that, like, because intelligent contracts are. They're going to be flexible, so it's going to be something that people are going to be using with their agents. So I'm guessing you've probably thought about this.
Reflections on Future Dynamics
Yeah, we think about this all the time. My take on this really is, you know, the 90. Ten. Right. The, the first 90% is 90%, but the last 10% is 90%, too. I think that right now we're in a moment where we are at the first 90% still of AI, but the last 10% where basically you say, okay, we don't really, they don't need us anymore. I think that's more difficult. I think that's the hypothesis we're working with. It's like, okay, if we will have, say, for example, that it's not going to be a fast takeoff, so it will take some time for them to really, let's say, quote, takeover or be able to do anything we want.
Looking Ahead to Human-AI Collaboration
If that doesn't happen so fast, let's say that the last 10% takes a lot. What it means is that we will have. That humans are going to be superpowered. We're going to have all the possibilities in the world to build the ideas that we have. We'll just have a legend of agents that are going to be helping us build whatever ideas that we have. Things are going to be accelerating a lot because humans will be much more powerful thanks to this technology. If you think about deFi, yes.
Emerging Protocols in the Crypto Landscape
You look at all these different protocols that have been created. There's a few versions of curve, there's a few versions of Uniswap, there's a few versions of Aave. But the fundamental primitives, the ones that are winning, are the ones that have the liquidity, the ones that have the assets within them. Right now, what I think is that all the fundamental primitives that Zen layer will enable, we will have quite a few years to be able to have anybody build them through the use of AI. It will be basically natural language that will allow you to create intelligent contracts, even if you don't know how to code at all. Just with having the idea, you're going to be able to create these protocols that are going to be useful for humans and AI's in the future to use.
Micro Task Marketplaces and AI Integration
Look, I can give you one, for example, there's this website from Amazon called the Mechanical Turk. You can think about appwork, for example, or fiber. Can you pay something for somebody else to do something very small, a small micro task, and then get it done and then get back to you, generally it can essentially be one of those. But for AI's to be able to say, hey, I want someone to go do something, post the proof like a photo or something on the Internet, and then the network itself to go verify that actually happened. Right. And if it happened, then there's a release of the payment. Yes. That type of totally decentralized marketplace for micro work, I think that's something that the AI will need like 100%.
Success in Future AI Movies and Current Trends
I don't know if you've seen like the, what is called from Johnny Depp, from the singularity. Yes. That's basically what the AI is doing in the movie is it's actually hiring people to build the data centers, is hiring people to capture information and help it interact with the real world. Right. And I think that stage will last quite a bit because I, because of the last 10%, right. We're trying to build infrastructure that will be useful for humans to be able to have a stake in the machine economy, let's say. Yeah, yeah, it's been a while since I've seen the singularity movie with Johnny Depp. But you touched on a point of this kind of notion and like a fear kind of like get back to, like, the Matrix.
Concerns Around AI and Human Employment
I know we're getting very high level right now. Maybe we can, in a second we can talk about Gen Vm. But you reminded me another recommendation for people on the show. Read Max Tegmark's life 3.0 is what you kind of touched on at the beginning of that book is this notion of the omegas, that it was possible that AGI or ASI would be reached internally somewhere, and then somehow, some way, an ASI convinces humans to allow it to get out, and then it kind of grows exponentially, duplicates itself infinitely, and then ends up kind of surreptitiously running companies secretly and employing people. And so then everything gets subverted or flipped on its head where AI is employing people.
Gig Economy and the Role of AI
I mean, you can think of Uber at this point, essentially humans being the API endpoint for working, where it's like they're like, I work for myself, but I. Here I am responding to this app that's telling me what to do. We've seen with Fiverr how the second GPT-3 came out, and especially GPT four, the amount of jobs that were available for gig economy workers, especially on a line to do, decreased rapidly. So we can only assume once there's easier to use and more customizable and extensible platforms for agents, that one person that understands how this works now will have the superpowers of a team of people.
The Future of Work and AI Integration
And there were plenty of talks where people like Ray Kurzweil was on a podcast, I think, with Alex Freeman or something, and he was like, oh, yeah, I think this decade we'll see a one man company that's like a billion dollar valuation, where you're able to have one person that just has these nested layers of agents that are working for them and they understand how that works, and therefore they can create more faster. But obviously, there's always the fear that you get to a point where you're like, am I working for AI? Like, are all the people that I'm working with, they don't know that they're working for AI? And so it's like the opposite of turtles all the way down, where it's like, well, what does that mean? Well, I think I'm working.
Navigating AI's Influence on Employment
I might be working for it. I don't know. Am I working for a person? I don't know. So, yeah, you touched on some points, but I just wanted to say for listeners, definitely read the first. If you don't have to read the whole book, I realize that reading can be tedious. Read life 3.0 by Max Tegmark. The intro is really good. It's a good intro into a big what if where you're like, oh God, am I working for AI? But yeah. When I see the announcement from OpenAI, like for example, last week they said, oh, now we're expanding to Europe and a bunch of other countries, right?
OpenAI's Reach and the Future of Technology
So like, if you look in the live 3.0 book, he's like, okay, they will start by doing movies, right? Sora, they will start by hiring people everywhere in the world, right? So, yeah, that rings a bell, right? If you want another book that I think is super interesting to understand what that will look like, I really like the, I think it's called the life of M. I will search the name like, is it the life of M? The age of MH, the age of M. It's very cool. It shows you the difference as well in speed, because ultimately they will be moving so much faster than whatever we can do for them.
The Future of AI Commerce
We are like trees, they're just moving so much faster than us that essentially commerce between AI to AI and AI to humans is going to be so different, two different dimensions. The question is, can we find a way that we can get a stake in that commerce that they will have, which will be hundreds of thousands of times faster than whatever we can do ourselves? I think that's one of the things where I feel like generally is going to be important. If you think about it, the way that general works is that the consensus algorithm doesn't fix any specific model.
Decentralized Systems and Future Innovations
Anyone can run a validator, it's a DPOS network. And when you're running a validator, you can select one or multiple models, large language models, but we're not telling you which one. You can run any model you want. You can go to llama three, you can go to GPT six, whatever it is. Which means that as the AI will improve, genlayer will improve too. Because you see, right now it's cheaper to be running tagpt 4.0 than chat GPT with GPT-3 because they have optimized that very much. The capabilities are improving.
The Evolution of Intelligent Contracts
It's cheaper, it's faster, it's smarter, and so will happen with the whole network. That's why we call it an intelligent blockchain. It really is like you have a machine that's a protocol that you can add something, it will answer. And right now you're basically going to get, let's say, average vote between different validators that have the technology that we see right now in the market. Yeah. In ten years that's going to be that. Every one of those validators will have nodes that are going to be running LLMs 100 times more powerful than what they are right now.
Emerging Capabilities of Blockchain Technology
And those are new capabilities that are possible. So for example, if you have a Dao and you have an intelligent contract that's evaluating a proposal, this intelligent contract can be saying something like, you need to make sure that the proposal that you will be voting has to be valid, has to be legal, and has to have the best ROI if you do a discounted cash flow. So the third one is not possible today because of how not smart yet the LLMs are. But in a matter of a year or 2100%, this will be possible.
The Future of Intelligent Contracts
Which I think, to me it's one of my most interesting things about jelly. The whole network will improve with technology, which is not something that you see anywhere else because everywhere else is totally deterministic and fixed. Look how difficult it is to change the block size for bitcoin or whatever. Yeah, I mean you touched on so many points, I dont even know where to begin. I think one thing ill mention before we move into maybe talking more about the stack is, again, that notion you mentioned, that every day that passes, and im sure people listening can probably relate to this, more projects are being released, more news is happening, technology is moving faster.
The Speed of Technological Evolution
And it's a little trite, but there's the classic Steve Jobs quote where computers are like bicycles for the mind. So our large language models like rocket ships or something are the future of these agentic and AGI platforms are going to just allow us to do a lot more. And then we kind of approach a limit of how much can our brains comprehend on a daily basis to be able to make sense of what's going on. It gets to a point where, and there have been research papers on this, I think Oxford released a paper recently on this that, I mean, there's within AI models, but there's also people becoming overly reliant on models to.
Reliance on AI and Its Implications
And I've experienced this myself, right. To take a step back where I'm relying on them at times to summarize, I'm relying on them for sources of truth and to break things down for me when there's so much stuff that's going on and it can be difficult to know what to focus on. But yeah, I mean, we touched on a lot of parts. I know we've got about 20 minutes left. I think we can dig deeper into to gen layer since, I mean, we could speculatively talk about AI and AGI for probably hours.
Exploring Gen Layer and Its Start-up Potential
But I think some listeners are probably like well how can I learn more about gen layer and maybe get started with the stack? So anything you want touch on with the stack, like the CLI, the simulator, genlayer, JS, anything that comes to mind that you want to maybe highlight or mention. Yeah, so I mean ultimately generally it will be a completely open platform. Anybody can come in and can contribute code to the platform.
Community Engagement and Testnet Development
Anybody can come in and start to build things on it. We're working on it. We started working on it about a year ago. We're trying to move as quickly as we can towards testnet. The way we're doing this is by essentially opening up pieces of it over time. We already opened up the simulator, which is essentially an instance of general that you can be running locally in your computer. What this means is that you can as of today start to write intelligent contracts.
Building on Zen layer
Now the main thing here is what can you build? Right now it's still something that you can be running on your computer, that you can even deploy in a server. But when the testnet will go out, when the mainnet will go out, then all of these contracts will be able to be ported to the public network to be executed there. So anybody can today already start to build on Zenlayer through the simulator when the testnet will be live. Obviously that will happen too. There will be different phases of it.
Encouraging Developer Participation
One of them will be focused on developers. We're setting up all the grants and different incentives for developers to come in and build different pieces of the ecosystem. So all of that, of course then we're just trying to make it so that people understand what we're building is the education that I was saying before. I think it's absolutely critical to do that and really appreciate that you're giving us a forum to be able to share what we're working on, ultimately welcoming anybody to everybody to join our discord, our telegram and just follow us as well.
The Vision for Blockchain Applications
Try to understand what's possible here. I haven't seen anything like what we're building. I think that there's going to be many more in the future, but right now the vision is very clear. Can we build a blockchain that can enable completely new types of applications? Then the question is who's going to be the first one that will be building them? Can we build funny meme coins that can change their price curve depending on how much attention there is on them? On the social media for example, there's going to be tons of different new type of approaches to crypto that you will see in it.
Innovations in Cryptocurrency
And we just want to see people come and give it a shot. The programming language is Python, which means that instead of less than 1% of the market share in developers is like 35% or something like this. Plus, LLMs now are really good at writing code, so you can just copy paste a couple intelligent contracts and then Claude will give you whatever you want to build. We've released, like you said before, the J's library as well, with a boilerplate that basically makes it very easy to start application with the front end and with the whole thing.
Accelerating Product Development
As I said before, I'm a computer scientist, but I'm always very much leaning towards product. For me, being able to prototype fast and see products and test them and see if there's interest in the market, that to me, is the way to actually get people to be building on top. So, yeah, we're going to be doing hackathons, we're going to be doing different type of activities, and I guess not much more than I can add to just welcoming everybody to participate and join us.
Resources Available for Learning
Totally, yeah, I mean, even just the docs for you guys on docs, dot gen layer.com is full of stuff for the gen layer simulator. It's taken me a lot of time to get through, and I can't claim to be an expert on any of it. Yeah, it's really interesting stuff, I'm thinking. I think people listening are probably curious. We've talked a little bit about where you're at right now. Where do you think gen layer is heading?
Innovative Directions for Gen Layer
I mean, are there any like, new novel innovations or anything you're really excited to share or, you know, we could just speculate on where you think things are heading for you guys. Like, what are the next, what does the next 36 months look like? I mean, 36 months will probably have an AGI between us, right? So. But yeah, we'd love to answer that.
Future Focus Areas
I think that, you know, like, first we're really focused on Desnet and we understand the scope of that. Then after Testnet, you have audits and you need to have a bunch of different things. Then you need to really shift focus as well. In the token, ultimately, this is, as I said before, a dpos network. So a proof of stake, delegated proof of stake, which means that the token is essentially the safety system of the protocol.
Token Distribution Importance
It's very important to do that distribution correctly. We want to make it as widely as possible. Again, we invite anybody to participate on the community as well as developers. And so on. That's going to be the focus for sure for the next, let's say, twelve months. Then ultimately there are quite a few innovations that we're seeing that are quite interesting.
Understanding New Concepts
For example, if you read the paper, maybe you know about this thing called the equivalency principle, which is like this. How can the consensus agree on something subjective? The way that the white paper shows it is by essentially having a second LLM call where the validators are saying, okay, yes, is this okay or not? Is this proposal from the leader valid or nothing? Acceptable or not acceptable?
Technical Innovations in Consensus
One of the innovations we've done is essentially generalize that so that this occurrence principle can be split into a leader function and a validator function. And that really, I mean, I'm getting too technical, I think, but these really, really expand the type of applications that can be built on top of the platform, like by a lot, because then you can have the leader is doing something and the validators are doing something else.
The Power of Decentralized Operations
Right. As long as they agree on their output, then that can be accepted. That is essentially that. It's not just that the platform is capable of non deterministic operations, but it's also capable of asymmetric operations. Again, this is a bit of a technical answer, I think, but yeah, I think what we really want to see is people come and build.
Facilitating Contract and Interface Development
I think we will make it super easy for people to build and deploy intelligent contracts and interfaces. We'll make it sure that it's as AI friendly as possible. It will be really easy for people to do that. And for the regular folk, let's say non developers are not interested in that. They will be able to try completely new types of applications that they haven't seen anywhere else.
The Future of Real World Assets
We have this team, for example, building, I think I said before, maybe a perpetual market on collectibles. Why can't you do that right now? Well, because they don't have oracles for collectibles. You want to buy some Rolex or you want to bet on the price of a specific watch. Now, you can do that with us. It will be just a few lines of code. It really democratizes it.
Tokenization and Its Prospects
So you're touching on rwas or real world assets on chain. I think in the last bull market, things like 4K.com blew up with people being like, hey, I've got a $40,000 watch, or I've got a $100,000 watch on chain. So, yeah, I wasn't thinking that we would talk about this, but how do you think, I mean, gen layer or intelligent contracts might be.
Potential Applications for Real World Assets
It seems kind of obvious how they might be used, but where do you kind of see rwas going? Or people like, actually, obviously, crypto fax, they want tokenize everything. They want to put everything on chain. Put my birth certificate on chain.
The Future of On-Chain Assets
Put everything on chain. But where do you think that's heading? And how do you see any synchronicities between what you're building and the future of real world assets on chain?
Initial Experiences and Observations
When I started, the first project I did in crypto was essentially a consulting company trying to sell solidity contracts, building decentralized applications for companies. Let's say we're trying to sell permission networks. And the problem is that, well, companies still don't get it. Still don't get it. As of today, there's some people that are using Hyperledger, different versions of it and so on, but still not really, really taking off. Right. And I think you need to think about, like, why, what is this gap between crypto and business, right?
Examining the Business-Crypto Gap
Will, for example, the ETF's be enough to close the gap, right? Or will it? Nothing, right? Do businesses right now have a fundamental problem transferring value between each other? And I think, I don't really see it yet. What I see is that they have a lot of infrastructure to do it, from sepa to visa, whatever you want. There's tons of infrastructure for that and regulation. So the question is, what is needed to actually bridge the gap?
Potential Solutions for Bridging the Gap
And I think, for example, being able to connect to external outputs, to external sources, that to me, is something that potentially can close that gap. So what this means is that you can have, for example, that the network itself is capable of going to a government website to identify something and bring it on chain. Like create a digital representation without any human in the loop. You say, this is my house. This is the identification for my house. You can see it in this website.
Proving Ownership and Automation
Then the question is, how do you prove that? But these LLMs can understand image too, and there's different ways that you can notarize that. So you can start to find ways to really bring real world assets on chain without any intermediary. Because the moment that you have an intermediary, then what's the point of doing it on chains? Do it without it. But yeah, I think we need to see. I think I'm still skeptical a bit on the business side of things.
Skepticism and Future Aspirations
I think that this will take some time. I'm much more interested volume wise, on the type of applications that can be built for AI to human and AI to AI commerce. I think that's going to be the biggest one for sure. People want to betting. For example, you want to buy a very expensive card, maybe you want to hedge your bet and you know, like you can essentially get insurance against it. Getting reprinted, for example.
Engagement and Reaction from the Audience
Right, stuff like that. Totally. Yeah. I mean, yeah. It's clear that RWAs still have a whiles to go, so I know we have, we got nine minutes left. We touched on a lot. Before I pick your brain about different models you're using for generating code, does anybody have any questions? I mean, I think got a few people in the audience and I've asked all the questions. I feel selfish because we've got you on the call.
Encouragement for Queries
If anybody has any questions, you can raise your hand, we can bring you on stage. If you don't, then Albert and I can still talk about really fun stuff without you guys. Any questions, anything that's on your mind to ask Albert, because it's an AMA, but it's like, ask Albert anything, it's an aaa, nothing. Okay, well, think about questions, think deeply.
Artificial Intelligence in Development
But I was going to ask you, Albert, I've been using cloud opus and I think obviously chain of thought reasoning is going to lead to ultimately better one shot outputs for code, for developers and just for anyone that's trying to build anything. Have you been using, I'm sure you've looked into cursor. What have you been messing around with? I'm curious.
Current Tools and Technologies
At this point in time, we have a bit of a pipeline here. I can tell you on the easiest side of things, zero one preview is really actually good at math and logic and code, but actually it's pretty expensive. You feel like you feel bad by asking it every time that you ask it. It's ok, maybe I will lose an opportunity to ask it in the future. I feel like something like 40 for like defining the problem really well then something like o one preview for actually resolving the problem.
Effective Coding Practices
Right. Then of course, like cursor is very good if you want to be coding. And I don't know if you've tried the relet agent, for example. Have you tried that replit agent? It's. Yeah, it's insane. Yeah, I think anybody on relet agents, guys, the agent platform relets. Definitely. Really? Go ahead, continue. I love it. No, no, 100% to me, as I said, me, I have so many ideas for stuff and I always hesitate because I know that if I need to change context and start building this prototype, it just takes forever.
Ideas and Innovation in Prototype Building
But I don't know, a few weeks ago I was thinking, I really want a token simulation platform. It took me literally a couple hours to get it with Monte Carlo simulations and this and that. It was just this flow photo plus om preview and then just the replication to do the deployment and most of the coding as well. And super, super impressive to me.
Exploration of Agents and Frameworks
It's crazy. And then if you want something a bit more deeper, there's now quite a few agents. This is something that we had been working almost, well, basically when we started. So, like almost two years ago, which is, well, these software engineering agents, right. There's one called Sree agent. There's a bunch of different things like this. Right. Ultimately, our goal is to basically have it so that when one of our team members wants to create, well, creates an issue, essentially there's an agent in the background that already is loaded with all the information with a rag and stuff on the code base, and then just codes, tests, debugs, makes a pr with documentation as well.
Enhancing Team Efficiency
It really helps us move faster in terms of how we're building all the different components of the protocol. Right. So again, I think that code will become absolutely commoditized. I think that we will all become like orchestra directors. So having ideas and testing them fast, I think it's the way to go. And we're really focused on trying to bring this as much to the core of the company as possible.
Documentation and AI Interaction
So, yeah, you will see that we're putting out tons of documentation. We're leveraging this as much as we can. It's very important right now, you have to assume that your docs will be read by an AI even more than a human. Right. Because we're getting to a point where people are kind of wanting to have organic conversations with docs. And I'm sure people on the call, if you use GitHub, use lens AI to talk with documentation that's been really useful, and then using tools like copilot or cursor for talking with docs.
Rapid Developments and Discussing Technologies
Totally. But, yeah, I mean, all the tools you mentioned, I was. I wasn't sure if you were going to say some bizarre LLM that I was like, I have no idea what Albert's talking about. I have no idea. You never know when you talk to people because you know. Yeah, it's moving so quickly. It's moving very quickly. Yeah. I think I'll ask the crowd once again if you guys have any questions.
Closing Remarks and Community Engagement
We got five minutes left. Otherwise, we'll let Albert give a little spiel about where to find gen layer, and then we'll close things out. So going once, going twice, I mean, I think we did such an amazing job on this AmA that nobody has any questions. Nobody has any questions because we covered everything. Well, to be fair, we didn't cover, we didn't go through all the docs, which would have taken forever.
Website Features and Accessibility
But I would just necessarily say, yeah, no, it's not necessary, but I would just say, go to Jen Lair's website. It's a really cool website. The white paper is really cool. And some white papers can be really boring. So you did a good job of making it not boring. So it's actually one of the more interesting white papers that's been released recently.
Innovative Functionalities on Gen Layer Website
And I feel like I would have a decent idea of that because I feel like I'm always looking stuff up. But yeah, we got a couple minutes. Left to make them easy. You see, for example, one thing we've done is if you go to the top of our website, you go to january.com. at the top it says ask GPT, you just click there and you can just be basically talking to the white paper.
Enhancements in AI Support
Well, yeah, much easier. Yeah, I was using that too. We also have an autonomous bot that we want to open up. That is a set of different knowledge and stuff from our GitHub library and from our KB. So we're taking inspiration from this. We'll probably have that bee easier and more available. I know members of our team, something that I like to do, and I know advanced voice mode isn't available yet for this, but being able to use advanced voice mode in the near future for customized agents on the OpenAI platform is going to be really powerful because the current voice mode works.
Future Developments in AI Interaction
But it's slower. So I mean it's. Yeah, so I'm looking forward to having faster conversations with docs, being able to build stuff. I mean, only question, really quick question, do you have any preferences for vector databases? Like are you like a pine cone guy or anything that you've been using for Rad?
Voice Technology and Partnerships
So yeah, it's been a while now, actually. Before we used to, we had a partnership with Chroma and then we had as well, with Pinecone as well. So we've been exploring different of them. We had an agent like right when we. So we've done a bunch of different agents. So Jaeger is the dev go behind. The first project that we did was basically an agent that was building its own tooling, develop it, essentially it was designing it, coding it, testing it, debugging it and so on.
Multi-Agent Framework Development
That's like we're talking a year and a half ago, like Mars, something like this. Then we moved into building a multi agent framework called Gen Worlds that basically allows for anybody to create, to deploy. You have a small world. Imagine a simulation and you put different agents that are very plug and play with different mental models, like tree of thought and chain of thought and so on, so that they have less degrees of freedom.
Challenges in Agent Interaction
So problem with the agents right now is that they are actually still quite wrong. We had a demo with a customer and the fucker basically took the database right before the test and just deleted the database. It decided to delete the database. Why? Yes, it's because too many degrees of freedom. What you want is to essentially remove those degrees of freedom.
Concluding Discussions
Right. And so, yeah, I'm just long winded, basically. Like, yeah, we've been exploring multiple vector databases for different, like, mind models. Right. We lost Albert again. Yeah, yeah. I think it's like when the screen goes off. So, yeah, basically, like, for me, I think it's important that it needs to be both local and if you want remote too, but you need to have one that's very close to the actual agent. If not, it can stop at any point.
Final Remarks
So, yeah, definitely. Okay, we got 15 seconds. I'll say. Everybody give a big virtual round of applause for having Albert on stage. I know autonomous is super excited to have gen layer be partnering and building with us. Likewise. Yeah, I mean, this is a great spaces. I think we all learned a lot and I would just encourage everybody go to genlayer.com, join gen layer discord.
Closure and Next Steps
If you're not already in autonomous discord, please join. Feel free to follow Albert or myself on Twitter. And autonomous, obviously. So this was fun, guys, and we ended right on the dot. So thanks, everybody, for stopping by. And, you know, I'm sure you'll be able. Feel free to reach out to us if you have any questions. This was a blast.
Gratitude and Well Wishes
Thank you very much. Right, have a good time. We'll talk soon. Thanks, everyone. Thanks so much, Alberta.