Funding the Commons · May 9, 2026

We are building systems we don't understand. Paul Riechers thinks that's the core problem, and that it's fixable. Paul, co-founder of Simplex and research lead at the Astera Institute, brings a physics background to AI safety. His central argument: without a mathematical understanding of what intelligence actually is, the debate between AI optimists and pessimists has no ground to stand on. Everyone has strong feelings. Nobody has a shared truth to point to. His research finds that transformer models develop internal world models with universal geometric structure, the same representations emerging across different architectures. That's evidence that interpretability is possible and that we can eventually understand what these systems are doing, not just observe what they produce. Speaker: Paul Riechers (Simplex / Astera Institute) This session was part of Intelligence at the Frontier, held March 14–15, 2026, at Frontier Tower in San Francisco. Learn more at https://www.fundingthecommons.io

ETHGlobal · May 9, 2026

In this workshop, Judson Bonneville from Gensyn introduces the Agent eXchange Layer (AXL), a decentralized peer-to-peer mesh networking layer that lets agents, tools, and machine learning workloads discover each other and communicate in a fully encrypted environment. AXL is a single Go binary that handles all networking — message routing, NAT traversal, and dual-layer encryption (hop-by-hop TLS plus end-to-end payload encryption) — with no central servers, DNS, port forwarding, or cloud accounts required. Applications interact through just three HTTP endpoints on localhost: send, receive, and topology. Judson explains how nodes join a bootstrap network and find each other via public keys (like phone numbers), enabling fully peer-to-peer communication where intermediary nodes can see transmissions happening but never the content. He covers AXL's support for MCP (exposing named services for structured JSON-RPC responses) and Google's A2A agent protocol for skill discovery. Use cases include agent collaboration, service registries, private data exchange, distributed inference with pipeline parallelism, and any peer-to-peer application. He demos a group chat app where two nodes message each other and share an OpenClaw agent instance, with the agent maintaining conversation context across multiple users — all built with simple HTTP calls to localhost while AXL handles the infrastructure. 00:00 Introduction 00:12 What Is the Agent eXchange Layer (AXL) 00:53 How AXL Works: Peer-to-Peer Mesh Network 01:36 Application Agnostic: Any Language, Any Payload 01:57 Public Keys as Node Identity 02:17 No Central Servers, No Port Forwarding 02:41 Three Core HTTP Endpoints: Send, Receive, Topology 03:22 Dual-Layer Encryption: TLS and End-to-End 04:26 Building with AXL: Send and Receive Pattern 05:15 Use Cases: Agent Collaboration and Chat Rooms 05:40 MCP and A2A Protocol Support 06:32 Getting Started: Clone, Build, Generate Identity 07:12 What You Can Build with AXL 07:29 Agent Collaboration: Decentralized Research 07:48 Service Registry and Tool Marketplace 08:23 Private Data Exchange with Encryption 08:51 Distributed Inference and Pipeline Parallelism 09:17 Packaged Examples: Gossip, Tensor Exchange, A2A 09:58 Live Demo: Peer-to-Peer Group Chat with OpenClaw Agent 11:27 Shared Agent Instance Across Multiple Users 13:45 How the Demo Works Under the Hood 14:23 Resources and Getting Started 14:41 Closing _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🤖 *Open Agents 2026* This workshop is part of Open Agents 2026, an online hackathon held from April 24 to May 6. This async hackathon brings together the world’s most talented Web3 developers, designers, and product builders to advance the Ethereum ecosystem! Watch the full Open Agents 2026 playlist here: https://www.youtube.com/playlist?list=PLXzKMXK2aHh6vpKqgVU7bEKyYK0tWfYEQ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ *Follow KeeperHub* X: https://x.com/KeeperHubApp ✅ *Follow ETHGlobal* X: https://x.com/ETHGlobal​ Warpcast: https://warpcast.com/ethglobal Website: https://ethglobal.com YouTube: https://www.youtube.com/c/ETHGlobal _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _Are you interested in Ethereum development and entrepreneurship?_ 👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events 🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus 📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Uma Roy

ETHGlobal · Oct 9, 2025

Explore the future of Zero-Knowledge (ZK) technology with Uma Roy from Succinct Labs as she introduces SP1, a revolutionary ZK Virtual Machine (ZKVM), at Pragma San Francisco 2024. Learn how SP1 simplifies ZK application development, making it accessible for all developers through Rust programming, enabling powerful blockchain scaling, cross-chain interoperability, privacy-preserving solutions, and verifiable real-world applications. Uma shares practical examples, innovative hackathon ideas, and insights into how SP1 is transforming Ethereum rollups, blockchain ecosystems, and beyond. 00:00 Intro: Uma Roy & Succinct Labs on SP1 ZK Applications 00:54 What is Zero-Knowledge (ZK)? Simple Explanation 01:27 Understanding ZK Proofs: Succinctness & Efficiency 02:01 ZK for Blockchain Scalability & Ethereum Rollups 03:04 Privacy Applications of ZK on Ethereum & Bitcoin 03:38 Using ZK for Privacy in Real-World (Web2) Scenarios 05:07 Verifiability & Regulatory Compliance Using ZK 06:08 Introduction to SP1: A Powerful ZK Virtual Machine 06:38 How SP1 Works: Rust Programming & RISC-V Compilation 07:13 Old ZK Development: Complex Circuits & Limitations 08:10 SP1 Advantages: Simple Rust Code & Easy Integration 09:13 SP1 Code Example: Fibonacci Number in Rust 09:48 SP1 Performance: Fastest ZKVM for Ethereum Rollups 11:25 Ethereum Rollups & ZK EVMs Built with SP1 12:35 Converting OP Stack Rollups into ZK Rollups with SP1 13:08 Cross-Chain Interoperability: ZK Light Clients & SP1 14:14 Hackathon Ideas: ZK Applications to Build with SP1 14:47 Building ZK Light Clients & Modular Security Bridges 15:18 Bitcoin Chain State Verification using SP1 15:47 ZK for Alternative VMs (Move VM, SVM, Cosmos SDK) 16:21 Fun Onchain Games & Provable Video Games with SP1 16:55 Advanced NFTs & Conditional Minting with SP1 ZK 17:27 Onchain Machine Learning (ML) Models Using SP1 17:57 Real-World ZK Applications & Use Cases with SP1 18:28 Authenticity Verification for Images (ZK Attestation) 18:57 Running SP1 & ZK Proofs on Mobile Devices 19:29 Verifying Emails & Passports Onchain with SP1 ZK 20:27 Selective Data Disclosure & Identity Privacy with ZK 20:59 Conclusion & Resources for SP1 ZK Development _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🇺🇸 *Pragma San Francisco* Pragma San Francisco 2024 was a special gathering of founders and builders that took place October 16th and 17th at The Contemporary Jewish Museum in San Francisco, CA. Pragma San Francisco 2024 was an opportunity discuss the future of web3 and reflect on the past. Watch the full Pragma San Francisco playlist here: https://youtube.com/playlist?list=PLXzKMXK2aHh4LcwDpGf_itZrAu7XRNb07&si=3x3y9J3avgNTITUe ETHGlobal's Pragma series takes place in cities around the world, and is designed to be a different kind of event. Pragma is a one-stage conference with founders-only on stage, bringing together a small group of curated attendees and speakers to discuss the future of web3 and reflect on the past. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ *Follow the Uma Roy* X: https://x.com/pumatheuma ✅ *Follow Succinct Labs* X: https://x.com/succinctlabs ✅ *Follow ETHGlobal* X: https://x.com/ETHGlobal​ Warpcast: https://warpcast.com/ethglobal Website: https://ethglobal.com YouTube: https://www.youtube.com/c/ETHGlobal _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _Are you interested in Ethereum development and entrepreneurship?_ 👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events 🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus 📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Anna Kazlauskas

ETHGlobal · Oct 9, 2025

Join Anna Kazlauskas from Vana at the Agentic Ethereum 2025 Summit to explore the future of decentralized data and its powerful impact on AI. Discover how self-sovereign data, proof of contribution, and data DApps enable privacy, ownership, and better AI models. See real-world examples of data DAOs, how users can legally own and contribute data from walled-garden platforms, and why structured, high-quality datasets are key to advancing AI agents. Learn how the Vana protocol ensures security with TE compute and encrypted storage—paving the way for a truly decentralized AI ecosystem. Whether you’re a developer, researcher, or enthusiast, this session offers a comprehensive look at unlocking web3 data for game-changing AI innovation. Stay tuned for insights into privacy-preserving data, building your own data DAO, and how to leverage user data across platforms for more powerful machine learning. Don’t forget to like, share, and subscribe for more talks on agentic governance, AI-driven solutions, and blockchain breakthroughs! 00:00 Introduction & Decentralized Data Overview 00:53 Why Data Quality Matters for AI Models 02:34 The Data Wall: Limitations on Public Internet Data 04:18 Scraping vs Curating: Approaches to AI Training Data 05:25 User-Owned Data and Emerging Data Markets 07:00 Legal Ownership: Exporting Data from Walled Gardens 08:08 Introducing Vana: Self-Sovereign Data Layer 09:41 Data DAOs and Proof of Contribution 11:21 Examples of Data DAOs & Private Data Use Cases 12:30 Verifying Data Integrity with TE Compute 16:51 Privacy-Preserving Data & AI Training 20:00 How to Start a Data DAO: Key Steps 23:08 Accessing Data Sets for AI Agents 24:14 Q&A: Building AI Solutions with Vana _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🤖 *Agentic Ethereum 2025 Summit* Join the official Agentic Ethereum 2025 Summit for an in-depth look at AI agents, smart contracts, and how they converge to power the next era of decentralized intelligence. This is more than just a summit—it's a gathering of the brightest minds in our ecosystem, and we're honored to have Andrew Miller, Anna Kazlauskas, Davide Crapis, dcbuilder, Ed Felten, Erik Voorhees, Illia Polosukhin, Jasper Zhang, Karl Floersch, Ken Ng, Kevin Owocki, Linda Xie, Shaw, Tomasz Stanczak, Chase Allred and Vitalik Buterin and on stage for this virtual summit focused around AI Agents and Blockchain! Watch the full Agentic Ethereum 2025 Summit Playlist here: https://youtube.com/playlist?list=PLXzKMXK2aHh6DrYO4ORk8wzQ5wCZDs0mW&si=ARqB9gaXAqehWH1_ 🤖 *Agentic Ethereum 2025* Agentic Ethereum is an online, async hackathon that runs from Jan. 31 to Feb. 12, 2025. With over $175,000 USD in prizes, Agentic Ethereum is a great way to dive into the world of autonomous AI agents and explore how they can revolutionize projects on Ethereum. Start the year with groundbreaking innovation! Watch the full playlist: https://youtube.com/playlist?list=PLXzKMXK2aHh7jxP_flgRf5nZ_WDgFpBy3&si=RZozbd9C5KrG6Bf3 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ Follow Anna Kazlauskas X: https://x.com/anna_kazlauskas ✅ Follow Vana X: https://x.com/vana Website: https://www.vana.org ✅ Follow Kartik Talwar X: https://x.com/therealkartik Warpcast: https://warpcast.com/kartik ✅ Follow ETHGlobal X: https://x.com/ETHGlobal​ Warpcast: https://warpcast.com/ethglobal Website: https://ethglobal.com YouTube: https://www.youtube.com/c/ETHGlobal _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Are you interested in Ethereum development and entrepreneurship? 👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events 🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus 📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

Vitalik Buterin

ETHGlobal · Oct 9, 2025

Join Vitalik Buterin at the Agentic Ethereum 2025 Summit as he revisits his predictions on the evolving relationship between crypto and AI one year later. Discover how AI is transforming on-chain governance through AI DAOs and “distilled human judgment,” and learn about the challenges of adversarial machine learning, exploitability, and the critical role cryptography (ZK proofs, TEEs) plays in securing advanced AI-crypto integrations. Gain insights into decentralized markets, retroactive funding, and the future of Ethereum as AI becomes both a powerful player and interface in blockchain ecosystems. 00:00 Intro & 2024 Blog Post Recap 00:55 One Year Later: Crypto Meets AI 01:31 Four Categories of AI–Crypto Intersections 02:45 AI as a Player in On-Chain Games 03:24 Arbitrage Bots & Prediction Markets 04:42 Introducing AI DAOs & Governance 06:01 AI Mechanisms vs. Bribery & Capture 07:48 Distilled Human Judgment Explained 10:08 Deep Funding & Retroactive Rewards 12:22 AI as the Interface (“Agents”) 14:11 Power Shifts & User-Aligned AI 16:38 AI as the Rules: Exploitability Risks 19:17 AI as the Objective: On-Chain ML 20:56 Conclusions: Short-Term Focus Areas 22:24 EVM, Scalability & Transaction Costs 23:34 ZK Proofs, TEEs & Trusted Computation 26:33 Unprovability & Governance Challenges 27:34 Wrap-Up & Next Panel Transition _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🤖 *Agentic Ethereum 2025 Summit* Join the official Agentic Ethereum 2025 Summit for an in-depth look at AI agents, smart contracts, and how they converge to power the next era of decentralized intelligence. This is more than just a summit—it's a gathering of the brightest minds in our ecosystem, and we're honored to have Andrew Miller, Anna Kazlauskas, Davide Crapis, dcbuilder, Ed Felten, Erik Voorhees, Illia Polosukhin, Jasper Zhang, Karl Floersch, Ken Ng, Kevin Owocki, Linda Xie, Shaw, Tomasz Stanczak, Chase Allred and Vitalik Buterin and on stage for this virtual summit focused around AI Agents and Blockchain! Watch the full Agentic Ethereum 2025 Summit Playlist here: https://youtube.com/playlist?list=PLXzKMXK2aHh6DrYO4ORk8wzQ5wCZDs0mW&si=ARqB9gaXAqehWH1_ 🤖 *Agentic Ethereum 2025* Agentic Ethereum is an online, async hackathon that runs from Jan. 31 to Feb. 12, 2025. With over $175,000 USD in prizes, Agentic Ethereum is a great way to dive into the world of autonomous AI agents and explore how they can revolutionize projects on Ethereum. Start the year with groundbreaking innovation! Watch the full playlist: https://youtube.com/playlist?list=PLXzKMXK2aHh7jxP_flgRf5nZ_WDgFpBy3&si=RZozbd9C5KrG6Bf3 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ *Follow Vitalik Buterin* X: https://x.com/vitalikbuterin Warpcast: https://warpcast.com/vitalik.eth Blog: https://vitalik.ca ✅ *Follow the Ethereum Foundation* X: https://x.com/Ethereum Warpcast: https://warpcast.com/ethereumfndn Website: https://ethereum.foundation YouTube: https://www.youtube.com/@EthereumFoundation ✅ Follow Kartik Talwar X: https://x.com/therealkartik Warpcast: https://warpcast.com/kartik ✅ Follow ETHGlobal X: https://x.com/ETHGlobal​ Warpcast: https://warpcast.com/ethglobal Website: https://ethglobal.com YouTube: https://www.youtube.com/c/ETHGlobal _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Are you interested in Ethereum development and entrepreneurship? 👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events 🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus 📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

ETHGlobal · Oct 9, 2025

Ram Kumar, co-founder of OpenLedger, presents their blockchain-based protocol designed for building specialized AI models. He highlights the limitations of current large language models (LLMs) like OpenAI and Llama, which are trained on generic internet data, resulting in generalized outputs and lack of transparency. OpenLedger addresses these issues by enabling users to contribute domain-specific datasets, build and fine-tune AI models via a no-code platform, and receive rewards through an on-chain attribution and provenance system. The protocol ensures data ownership, auditability, and verifiability, distinguishing it from centralized platforms like Hugging Face. OpenLedger supports both community developers and enterprises, with use cases ranging from Solidity code generation to customer support and tax advisory. The ecosystem includes a marketplace for models, agent frameworks, stablecoin micropayments, and a DEX for model tokenization. With over a million users in testnet, OpenLedger is set to launch its mainnet, aiming to democratize and decentralize AI model creation and monetization. 00:00 Introduction 00:38 Why AI Models Need Specialization 01:45 OpenLedger’s Background and Enterprise Experience 02:41 Problems with Current AI Models 03:33 OpenLedger’s Solution: Data, Models, and Attribution 04:38 Blockchain for Verifiable AI and Data Ownership 05:34 OpenLedger vs. Traditional AI Platforms 06:04 Product Overview: Data Nets, Model Factory, Open Models 07:24 Example: Building a Specialized Solidity Model 08:34 Model Marketplace and Access 09:22 Enterprise Use Cases and Partnerships 10:20 Auditability and Attribution for Enterprises 11:21 Community and Enterprise Ecosystem 11:56 Ecosystem Extensions: Stablecoin, DEX, Agent Frameworks 12:38 Testnet Results and Mainnet Launch 13:09 Closing Remarks _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🇫🇷 *Pragma Cannes* Pragma Cannes 2025 was held on July 3rd at the Carlton Cannes in Cannes, France and was an in-person summit for builders and leaders in the web3 ecosystem. Watch the full Pragma Cannes YouTube Playlist here: https://www.youtube.com/playlist?list=PLXzKMXK2aHh6jZYPY5-YIBzvMtUT3ajjI ETHGlobal's Pragma series takes place in cities around the world, and is designed to be a different kind of event. Pragma is a one-stage conference with founders-only on stage, bringing together a small group of curated attendees and speakers to discuss the future of web3 and reflect on the past. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ Follow Ram Kumar X: https://x.com/Ramkumartweet ✅ Follow OpenLedger X: https://x.com/OpenledgerHQ ✅ Follow ETHGlobal X: https://x.com/ETHGlobal​ Warpcast: https://warpcast.com/ethglobal Website: https://ethglobal.com YouTube: https://www.youtube.com/@UCfF9ZO8Ug4xk_AJd4aeT5HA _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Are you interested in Ethereum development and entrepreneurship? 👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events 🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus 📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

ETHGlobal · Oct 9, 2025

Ismael Hishon Rezaizadeh, co-founder of Lrange Labs, discusses the critical role of Ethereum's infrastructure and zero knowledge proofs (ZKPs) in ensuring trust, privacy, and safety in AI applications. He introduces zero knowledge machine learning (ZKML), which leverages ZKPs to verify AI model integrity, privacy, and fairness, especially in security-sensitive sectors like healthcare, defense, finance, and content moderation. Lrange’s Deep Prove library enables fast, scalable, and commercially viable ZKML, supporting large models and real-world deployments with partners like Intel, Nvidia, and AWS. Ismael emphasizes the need for verifiability and transparency in AI, moving beyond blind trust to cryptographic proofs, and highlights the growing adoption of ZKML in both Web2 and Web3. He also discusses future advancements, including proofs of training, fairness, and reasoning, and the societal impact of embedding verifiability and privacy as primitives in AI systems. 00:00 Introduction 00:36 Zero Knowledge Proofs & ZKML Overview 01:35 Challenges in AI Trust, Safety, and Bias 02:26 Importance of ZKPs for AI Verifiability 03:17 Lrange Labs & Deep Prove Introduction 04:29 How Deep Prove Works 05:56 Use Cases: Defense, Healthcare, Web2, Web3 08:05 Decentralization & Open Source Approach 09:07 Lrange Research & Industry Collaboration 10:44 Extending ZKML Beyond Crypto 11:29 Societal and Sectoral Impact of ZKML 13:14 Agentic Systems, Compliance, Content Moderation 15:02 DeFi, Governance, and Healthcare Applications 16:40 The Future: ZK as a Default Primitive 17:43 Q&A: ZK Computational Improvements & Future Directions 20:33 Conclusion _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🇫🇷 *Pragma Cannes* Pragma Cannes 2025 was held on July 3rd at the Carlton Cannes in Cannes, France and was an in-person summit for builders and leaders in the web3 ecosystem. Watch the full Pragma Cannes YouTube Playlist here: https://www.youtube.com/playlist?list=PLXzKMXK2aHh6jZYPY5-YIBzvMtUT3ajjI ETHGlobal's Pragma series takes place in cities around the world, and is designed to be a different kind of event. Pragma is a one-stage conference with founders-only on stage, bringing together a small group of curated attendees and speakers to discuss the future of web3 and reflect on the past. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ Follow Lagrange Labs X: https://x.com/lagrangelabs ✅ Follow ETHGlobal X: https://x.com/ETHGlobal​ Warpcast: https://warpcast.com/ethglobal Website: https://ethglobal.com YouTube: https://www.youtube.com/@UCfF9ZO8Ug4xk_AJd4aeT5HA _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Are you interested in Ethereum development and entrepreneurship? 👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events 🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus 📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

ETHGlobal · Oct 9, 2024

Harry Grieve, co-founder of Gensyn, discusses decentralizing machine learning compute at Pragma London 2024. Learn how Gensyn's protocol creates a global supercluster of ML compute resources, enabling permissionless access to GPU power at lower costs than centralized providers. Discover the technical challenges of verifying distributed ML work, ensuring reproducibility across hardware, and optimizing model sharding. Explore the future intersection of crypto and AI, including machine-to-machine economic interactions and the coming wave of decentralized artificial intelligence. Essential viewing for ML engineers, blockchain developers, and anyone interested in the decentralized future of AI infrastructure. 00:00 - Introduction to Gensyn and Decentralized ML Compute 00:50 - The Growing Intersection of Crypto and AI 01:20 - Gensyn's Mission: Internet-Scale ML Training 01:57 - Founding Story: The GPU Access Problem 02:29 - Rising Compute Costs and GPT-2's Impact 03:02 - Academic vs Industry ML Resource Disparity 03:30 - Vision for Internet-Scale Machine Learning 04:01 - ML as the Next Industrial Revolution 04:31 - Transition to Dynamic AI-Powered Internet 05:02 - Controlling the Future of Online Information 05:32 - Training Models Across Any Compatible Device 06:03 - Five Key Research Challenges in Decentralized ML 06:33 - Reproducibility Across Different Hardware 06:58 - Optimal Model Sharding and Distribution 07:26 - Blockchain and Data Availability Considerations 07:51 - Current Status: Internal Network Testing 08:27 - Implementation Progress and Security Focus 08:55 - Data Availability Layer Requirements 09:52 - Four Key DA Needs for ML Protocols 10:22 - Temporary vs Permanent Storage Solutions 11:15 - Decentralized Model Training Research 11:45 - Physical Limits of Centralized Compute 12:18 - Breaking the Data Center Scaling Barrier 12:46 - Three Core Focus Areas for Gensyn 13:16 - Permissionless Network Principles 13:49 - Machine-to-Machine Compute Demand 14:23 - Autonomous AI Economic Agents 14:56 - The Liberty Argument for Decentralized AI 15:26 - Owning Your Artificial Neurons 15:51 - Crypto's Role in AI's Future 16:24 - Three-Layer Model for Crypto+AI Integration 16:50 - ML Ops: Attribution and Guardrails 17:23 - AGI on Cryptographic Substrates 17:49 - Four-Year Prediction for Permissionless AGI 18:23 - AI-Physical World Interface Challenges 18:55 - Gensyn's Roadmap: Private to Public Testing 19:27 - Research Opportunities in Distributed ML 19:56 - Need for External Data Solutions 20:23 - Closing Thoughts on Decentralized AI Future _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 🇬🇧 *Pragma London* Pragma London took place at Christchurch Spitalfields in London on March 14, 2024, and was an in-person summit for builders and leaders in the web3 ecosystem. Watch the full Pragma London playlist here: https://youtube.com/playlist?list=PLXzKMXK2aHh7rvaHBM14peT1TwI3QX38p&si=MzGrEYwYjxSQFiFM ETHGlobal's Pragma series takes place in cities around the world, and is designed to be a different kind of event. Pragma is a one-stage conference with founders-only on stage, bringing together a small group of curated attendees and speakers to discuss the future of web3 and reflect on the past. Ever since we first introduced our new event series Pragma in Tokyo of the spring of 2023, we have been fostering a curated, global community of individuals who deeply care about web3. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ✅ *Follow Harry Grieve* X: https://x.com/_grieve ✅ *Follow Gensyn* X: https://x.com/gensynai ✅ *Follow ETHGlobal* X: https://x.com/ETHGlobal​ Warpcast: https://warpcast.com/ethglobal Website: https://ethglobal.com YouTube: https://www.youtube.com/c/ETHGlobal _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _Are you interested in Ethereum development and entrepreneurship?_ 👉 Sign up for the next ETHGlobal event: https://ethglobal.com/events 🎁 Get exclusive access and perks with ETHGlobal Plus! https://ethglobal.com/plus 📣 Want us to throw an event in your city? Tell us where! https://ethglobal.com/city _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

ETHTokyo · Sep 21, 2024

This talk explores how Satoshi Nakamoto's invention of Bitcoin sparked a decentralization movement that is reshaping human organization. It highlights the shift from traditional, centralized models to decentralized networks where AI and individuals collaboratively manage infrastructure and economic activity on a global scale. The video features a speaker discussing the legacy of Satoshi Nakamoto, the creator of Bitcoin, and predicting that it will not be Bitcoin itself but rather the foundation for decentralized machine perception and AI interaction with the physical world. The speaker draws parallels between rapid urban development and AI growth, suggesting that machine-to-machine transactions will dominate future economies. They argue that AIs will need to buy collaborative spatial computing resources to navigate and understand the physical world. The speaker also touches upon how current tech giants are investing in spatial computing, which is crucial for AR devices and autonomous machines but raises privacy concerns due to potential surveillance capitalism. They introduce an alternative concept called Deepin (decentralized physical infrastructure networks), which allows people to deploy large-scale infrastructure projects collaboratively. Specifically, they present their project as a decentralized network for machine perception called PostMesh, which aims to provide an external sense of space for machines. The project is already being piloted in retail environments to improve inventory management and communication through spatial computing. The speaker announces the launch of their token and a grant program to encourage developers to contribute to this decentralized protocol, aiming to prevent a future where centralized entities like Optimus robots become the primary source of data collection. For more information or involvement, viewers are directed to their website and Twitter page.

ZuBerlin · Jun 19, 2024

Clip The video features a discussion on synthetic data and its growing importance in training AI models, with predictions that 60% of training data will be synthetic by the end of the year. The speaker emphasizes the benefits of using synthetic data, such as domain specificity, cost reduction, bias control, and consistent labeling. They explore different research papers on generating question-answer pairs for training AI models and discuss various methods and their effectiveness. The talk also covers the limitations of language models in generating unique and domain-specific queries, the superiority of human-generated data in specific domains, and different approaches to improve data generation. The speaker suggests starting with synthetic data generation and shares insights into adversarial example generation as well as the potential of decision-making algorithms. Questions from the audience address concerns about bias amplification, data quality trade-offs, and the efficiency of synthetic versus human-generated data.

ZuBerlin · Jun 19, 2024

Clip The video features a speaker discussing the importance of model composability and composable systems in the context of agentic AI and blockchain infrastructure. The speaker emphasizes that while many people view AI models as simple query-response tools, real-world applications often involve more complex systems composed of specialized models. These models can be combined to perform complex functions and improve accuracy through ensemble techniques. The talk also covers the inefficiencies of running computations on blockchains due to their replicated state machine nature. To address this, a novel approach called Validation Computation Separation (VCS) is introduced, involving inference nodes that perform stateless computations and transmit proofs to the network for validation, balancing scalability with decentralization. The speaker further explores the use of machine learning in quoting dynamic fees for automated market maker (AMM) swaps, potentially reducing losses for liquidity providers (LPs). They highlight how traditional finance uses various risk and forecasting models that are not yet common in Web3 but could become foundational with further development. Looking ahead, the focus will be on creating a collaborative research environment to develop models specifically designed for Web3 protocols. The infrastructure will support various model architectures and hardware acceleration for inference and proof generation. Additionally, an ERC standard is being developed to define agent interactions on EVM-based chains. Lastly, questions

ZuBerlin · Jun 19, 2024

Clip The video features a presentation on behalf of Gnosis Labs by one of its AI team members. The presenter announces the discovery of lost wallets and then proceeds to discuss the team's work on integrating AI with prediction markets, specifically the IOMEN platform. They delve into how prediction markets function, the role of AI agents in making market predictions, and the monetary incentives involved. The speaker introduces a hackathon challenge aimed at improving AI agent predictions within these markets. The team has developed a demo agent called Prediction Profet, which analyzes complex questions to make predictions. However, they acknowledge its limitations and invite participants to innovate better methods for information aggregation and result prediction. Another advanced agent, Tinker-Roughy, is mentioned as an example that generates related prediction markets for more comprehensive analysis. Participants are encouraged to join a Telegram group for support and use GitHub resources provided by Gnosis Labs to develop their solutions. The hackathon involves modifying a predict function within a Python script to improve accuracy beyond the current 75% achieved by their demo agent. The presentation concludes with details on how submissions will be tested and evaluated, including benchmarks against human predictions on another platform called Manifold. The hackathon offers prizes for winning solutions and aims to push forward the capabilities of AI agents in prediction markets over a

ZuBerlin · Jun 18, 2024

Clip The video features Jiahao Sun, the founder and CEO of Flock, discussing the company's mission and technology. Flock is a pun for federative learning blockchain, an idea stemming from a paper published in NeurIPS 2022 and further detailed in the journal "Transitioning AI." Jiahao's background includes working in traditional finance as head of AI at Eurobank Canada and as a researcher at Imperial College London. Flock aims to solve data silo issues in industries like finance and healthcare by combining federated learning with blockchain technology. Federated learning occurs on devices like smartphones, where data stays local while only model weight changes are sent to improve collective models. However, there's a risk of centralized control and data misuse, which Flock addresses using blockchain for transparency and security. The platform introduces an incentive mechanism where good behaviors are rewarded, and bad actors are penalized, ensuring honest participation. The system uses random selection for proposers who contribute to model training and voters who validate results. Malicious attempts are thwarted through this process, maintaining high accuracy even with a significant presence of bad actors. Jiahao also discusses experiments with the Kaggle Learning Club dataset that demonstrate the effectiveness of Flock's approach compared to traditional federated learning under malicious conditions.

ZuBerlin · Jun 18, 2024

Clip The video features a speaker with a background in AI and crypto, discussing the current landscape at the intersection of these two fields. The speaker, who has experience both at Google and in crypto mining, shares insights gained from meeting with numerous companies in the crypto space. The talk covers a range of topics including decentralized GPUs, inference networks, edge computing, data generation and labeling, ZKML (Zero-Knowledge Machine Learning), fully homomorphic encryption, and AI agents. The speaker expresses skepticism about certain areas like AI agents due to scalability issues and the premature nature of these technologies. They also highlight an experiment where AI was used to categorize companies based on Twitter data, discussing both successes and shortcomings of this approach. Additionally, they touch upon the potential for crypto to address core issues in AI such as copyright, privacy, and user ownership. Throughout the video, various subcategories within each main topic are discussed along with specific companies that are making strides in those areas. The speaker is keen on creating a dynamic market map that reflects current trends and is open to community contributions. They also emphasize their excitement about data generation categories due to their potential high utility in crypto. The session concludes with questions from the audience about specific aspects of the market landscape and an acknowledgment of ongoing developments that might change existing

ZuBerlin · Jun 18, 2024

Clip In this video, Anna, the co-founder and CEO of Vana, discusses her background in machine learning research and crypto, her journey from mining Ethereum to dropping out of school to start a machine learning company, and her work at Vana focusing on self-sovereign technologies for AI and data. She explores the concept of decentralized AI, emphasizing that it doesn't necessarily mean putting everything on the blockchain but can be achieved through local open-source models. Anna contrasts crypto's early fringe use cases with those of AI, such as generating explicit images using private local models. She explains the process of training diffusion models, tweaking model parameters for desired outcomes, and the importance of open-source models for maintaining privacy and control over AI. Anna highlights community-driven projects where contributors iteratively improve models, like in open-source software development. She also touches on funding challenges faced by open-source AI due to difficulties in monetizing freely available models and how this reliance has shifted towards big tech companies. Addressing data privacy concerns, she discusses Vana's approach to secure data contribution without risking leaks. The video concludes with a Q&A session where Anna addresses various topics including the potential risks and benefits of open-source AI models, decentralized approaches to model funding and training, the importance of keeping frontier models open source for

ZuBerlin · Jun 18, 2024

Clip The video features a speaker, Michael Friedrich, discussing various topics related to generative AI and its implications in the crypto and technical world. He begins by engaging the audience with questions about their use of AI tools like ChatGPT and its competitors. Michael introduces himself, mentioning his seven-year tenure at Adobe, his work on AI and ML platforms, and personal interests including surfing and FPV drone racing. He then delves into the definition of generative AI, illustrating its evolution from text-to-image applications to multimodal ones that include speech output. The conversation touches on the hype around generative AI in business marketing and its potential trajectory towards achieving human-level intelligence. Michael skips over basic explanations of transformer architecture due to the audience's familiarity, moving on to discuss large language models (LLMs), their structure, function, and how they learn from data inputs to predict outputs. He highlights the astonishing capabilities of LLMs like GPT-3 and GPT-4 in understanding language and generating coherent responses. Questions from the audience lead to discussions about unpredictability in AI outputs, the concept of creativity in models, and how randomness plays a role in generating new content. The speaker also addresses concerns about whether AIs can be considered conscious beings. Further into the talk, Michael explores the

European Blockchain Convention · Oct 9, 2023

Blockchain is revolutionizing the financial markets, but infrastructure and regulations are not there yet. How can we protect our on-chain operations and prevent fraud and money laundering? Listen to AnChain.AI presentation by Michael Fasanello at the European Blockchain Convention 2023. --------- The European Blockchain Convention is the most influential blockchain event in Europe. Follow us to find out more: Linkedin - https://www.linkedin.com/company/euro… Instagram - https://www.instagram.com/eblockchain… Twitter - https://twitter.com/EBlockchainCon