Hi Joseph, happy new year! Thank you for supporting the manifesto and for your feedback. That’s a good point and actually came up when we wrote the manifesto; we opted to go with verifiable AI as it’s more broad (and open-source is one way to achieve this).
Hi everyone, happy new year and thank you for today’s call (2025.01.02). This is the generated summary (short version, please find the long version here ):
The working group identified key goals and topics of interest going into 2025. Plans include expanding collaborations with Web3 and AI ecosystems, improving developer resources through tutorials and bounties, and fostering cross-community engagement. Regular sessions and publications, such as a “State of DeAI” update, will showcase progress and ecosystem advancements. Notable discussions focused on integrating AR and AI into decentralized applications, addressing technical challenges, and highlighting opportunities for cross-ecosystem compatibility. Upcoming sessions aim to explore infrastructure, partnerships, on-chain LLMs, and the DeAI manifesto.
Hi everyone, thank you for today’s call (2025.01.09) and special thanks to @icarus for leading it. This is the generated summary (short version, please find the long version here ): AI4AI on the Internet Computer (IC) is a community-driven initiative aimed at developing scalable, decentralized AI infrastructure by fostering collaboration between DFINITY, developers, and node providers. It focuses on creating an open, community-maintained roadmap aligned with DFINITY’s goals, advancing hardware capabilities (e.g., GPU-enabled nodes), and exploring software solutions for efficient AI inference and specialized AI subnets. By adhering to principles of openness, privacy, and sustainability, the initiative envisions transforming the IC into a leading platform for decentralized AI, supporting real-time applications, model orchestration, and secure, on-chain AI workloads.
Hi everyone, this Thursday the DFINITY AI team will join our call (thank you for organizing this @jennifertran ) and we’ll take the opportunity to get to know each other please also bring any questions or topics you might have. See you all then
note: the call will take place in the new ICP Discord: ICP
Hi everyone, thank you for yesterday’s call (2024.01.16) and special thanks to the DFINITY AI team members who joined us! This is the generated summary (short version, please find the long version here ):
The DeAI Working Group call was a joined session with the DFINITY AI team. It focused on integrating AI and GPU capabilities into the Internet Computer Protocol (ICP), deploying small-scale AI applications using CPUs, optimizing matrix multiplication, and exploring future hardware upgrades like AMD Zen 5 CPUs. Collaboration opportunities with networks like Fetch.AI and Swarm were discussed, alongside the potential for decentralized MCP servers for diverse data sources. The group emphasized balancing flexibility and efficiency, promoting community demos, and creating resources to showcase current AI capabilities while advancing toward GPU-based solutions.
Hi everyone, this week we’ll have a demo session during our call and Viviana from FAI3 will present how they certify AI models on-chain
does anyone else want to share some of their latest work or anything else?
Thank you for today’s call (2024.01.23)! As an overview:
Viviana of FAI3Network presented her solution to certify AI models.
She demoed how to upload a model, how FAI3 assesses the models, and analyzes them based on various data point evaluations (ex. statistical parity difference, disparate impact).
You can upload different data sets to see how the models perform with them. Models do not need to be open-sourced.
Suggestions for improvement included:
Add documentation on FAI3 works and what academic literature it references
Use case: For a decentralized metaverse, in order to register an LLM into a metaverse, the model would certain language barriers.
Community ask: Please review FAI3 network and provide any feedback to Viviana through telegram: vsiless.
I presented the concept of Eliza, the AI framework, and the concept of plugins. I talked about the PickPump plugin. I demoed the example of an AI agent creating a token from a Telegram bot that the Pickpump team created.
@PickPump, would you like to come back and give an overview of the code and demo?
Hi everyone, in our call this Thursday, we’ll talk about the Proof-of-AI-Work Protocol, the icpp-pro Bootcamp and of course DeepSeek AI on-chain. @icpp and I will share some of our latest work there. Does anyone else want to share something or see any specific items on the agenda?
Hi everyone, thank you for yesterday’s call (2025.01.30). This is the generated summary (short version, please find the long version here ):
The onicai team presented their work on self-sovereign AI using ICP, focusing on deploying the Deep Seek AI model on-chain and optimizing its performance. Key challenges include high computational costs, instruction limits, and memory inefficiencies. Proposed solutions involve optimizing llama.cpp, introducing AI Workers with GPUs for scalable AI execution, and exploring TEEs for secure verification. Additionally, a C++ Canister Development Kit bootcamp aims to attract more developers. Dfinity is actively researching AI optimizations, and community feedback is sought to determine the best path forward for decentralized AI on ICP.
Hi everyone, during this week’s call we’ll have Zensh from the ICPanda DAO talking about the Anda agent framework (Introducing Anda)
Are there any additional items you’d like to see on the agenda? Please let us know
Hi everyone, thank you for today’s call (2025.02.06). Special thanks to @zensh and @zire for the presentation and discussion! This is the generated summary (short version, please find the long version here ): @zensh from IC Panda presented the Anda AI agent framework and its integration with DeepSeek and ICP. Anda, a Rust-based framework, focuses on decentralized AI agents with secure execution using trusted execution environments (TEEs) and encrypted memory on ICP. Community efforts include upcoming developer documentation, a hackathon, and grants. Future directions involve multi-agent coordination, improved infrastructure, and expanding real-world AI agent use cases on ICP.
Hi everyone, thank you for today’s call (2025.02.13) and special thanks to @icarus ! This is the generated summary (short version, please find the long version here ):
The discussion focused on upgrading Internet Computer (ICP) node hardware from Gen1 to Gen3, emphasizing faster AMD EPYC processors, PCIe 5, and improved memory bandwidth for enhanced AI processing. It covered GPU integration, including AMD Instinct and Nvidia options, and the concept of AI subnets for high-performance tasks. The impact of mixing hardware generations on ICP subnets was considered, along with an RFP announcement for AI agent frameworks. Other topics included historical blockchain data access, chain fusion, and community engagement through events and technical discussions. The call provided a deep technical dive into optimizing ICP’s infrastructure for AI advancements.
Hi everyone, on this week’s call we’ll talk about ETH Denver and the quarterly State of DeAI report we ideated before. For ETH Denver, we’ll connect all of you who are going and can discuss any opportunities to meet up, plus we had plans to come up with ideas how we may get external teams that you meet at ETH Denver involved in the DeAI group and what kind of collaboration opportunities there are. So we’ll make these plans concrete too. Are there any other items you’d like to see on the agenda?