Technical Working Group DeAI

Developers
264

Hi @MalithHatananchchige , that sounds great, thank you. Your work looks super interesting, so it’d be great to put a presentation (intro, overview and maybe also demo) on the agenda of one of the upcoming calls. Would you be open to sharing more on your work in this way with the group?

1 Like
265

Would be happy to join in and be apart of this.

1 Like
266

the discussions here are interesting, would be nice to see some updates by foundation about Caffeine.

1 Like
267

Hi everyone, in this week’s call, we’ll dive into several recent trends and developments in the AI space; Active Inference, Federated Learning, DeepSeek Technical Report, and Exploratory AI horizons. Our goal is to discuss state-of-the-art and next-gen AI in the group and stay up-to-date with the fast moving AI industry. Looking forward to seeing you all then :+1:

6 Likes
268

Hey all, fun little AI project I am working on: Consciousness Orchestration Layer

Evolves emotionally through interactions
Develops a unique personality over time
Is owned/tradeable as an NFT with distinctive traits

4 Likes
269

@patnorris, when is the call again weekly?

270

Hi @MalithHatananchchige , the calls are each Thursday at 6pm CET in the ICP Discord voice channel. This should take you to the event on Discord (and also show the time in your timezone): ICP

2 Likes
271

please demo it! is the code open source?

1 Like
272

Hi everyone, thank you for today’s call (2025.04.03). Special thanks to Tim , Kawa and @MalithHatananchchige for sharing their research and work! This is the generated summary (short version, please find the long version here ): In this DeAI Working Group call, participants explored decentralized AI training via federated learning, emphasizing data ownership, on-chain verifiability, and incentive mechanisms. The discussion then shifted to future AI paradigms like active inference, reservoir computing, neuromorphic chips, and quantum machine learning—highlighting their potential for continuous learning, energy efficiency, and decentralization. A notable idea was using the Internet Computer as a distributed runtime for factor-graph-based inference. The session concluded with plans for upcoming demos and deep dives into model context protocol (MCP), AI hardware, and on-chain agents, reflecting growing interest in alternative AI architectures aligned with decentralized principles.

Links shared during the call - Part 1:

Links shared during the call - Part 2:

Please find Tim’s prepared content in this document (“Active Inference, and Exploratory AI horizons”): DeAIWorkingGroupInternetComputer/WorkingGroupMeetings/2025.04.03/Active Inference, and Exploratory AI horizons.pdf at main · DeAIWorkingGroupInternetComputer/DeAIWorkingGroupInternetComputer · GitHub

4 Likes
273

Upcoming Calls & Topics

Date Topic With
Apr 10 Model Context Protocol (MCP) – First ICP Demo @baolongt
Apr 17 AI4AI – Accelerated AI Infrastructure @icarus
Apr 24 IConfucius – Fully On-Chain AI Agent on ICP @icpp (onicai)
May (tbd) Session on Efficient LLMs and DeepSeek Open
Future Expert session on Active Inference implementations Invite TBD
2 Likes
274

@patnorris In the call, there was a discussion on expert model training; a faster way I mentioned which can be Medus heads Link

1 Like
275

Hi everyone, this week we’ll focus on the Model Context Protocol (MCP) during our call :thumbsup: We’ll have a demo by @baolongt on the first ICP MCP server (as interface to SNS) that he built. In addition, it’d be great to discuss which applications of MCP on ICP are interesting and how we can help to establish MCP servers on ICP. Looking forward to the call and seeing you all then :muscle:

A few links:
MCP: https://www.anthropic.com/news/model-context-protocol
@baolongt SNS MCP server: GitHub - baolongt/sns-mcp-server: mcp server for sns
Forum discussion on MCP: Coordinated plan on supporting MCP for Vibe Coding and AI Agents

6 Likes
276

Hi everyone, thank you for today’s call (2025.04.10). Special thanks to @baolongt for sharing his work and expertise on MCP! This is the generated summary (short version, please find the long version here ): In this session, the group explored recent developments in open-source AI and emerging agent protocols, in particular the Model Context Protocol (MCP). The highlight was a demo by Long, showcasing the first integration of MCP with the Internet Computer (ICP), enabling AI agents to interact with blockchain tools like the NNS and perform actions such as proposal voting and wallet authentication. Discussions covered technical challenges (e.g., overlapping endpoints, authentication security, lack of full server-sent event support), the potential of self-sovereign AI agents, and next steps including developer tooling and upcoming sessions on AI infrastructure and on-chain agents.

Links shared during the call:

4 Likes
277

:spiral_calendar: Upcoming Calls & Topics

Date Topic Presenter
Apr 17 AI4AI – Accelerated AI Infrastructure for ICP @icarus
Apr 24 IConfucius – Fully On-Chain AI Agent on ICP @icpp (onicai)
May ANIMA demo (link) @swift
May (tbd) Session on Efficient LLMs and DeepSeek
May (tbd) Session on instruction limit
Future Expert session on Active Inference implementations Invite TBD
278

Hi everyone, this week @icarus will lead our next session on Accelerated Infrastructure for AI (AI4AI) to focus on hardware and infrastructure to support AI on ICP :+1:
This is the event for the call in the ICP Discord: ICP

Looking forward to the call and seeing you then

1 Like
279

Hi

Could you please bear in mind my suggestions for hardware.

1 Like
280

Hi everyone, thank you for today’s call (2025.04.17). Specials thanks to @icarus for leading the call and discuss, and to @MalithHatananchchige for sharing insights. This is the generated summary (short version, please find the long version here ): Today’s DeAI Working Group call focused on TensTorrent’s new Black Hole ASIC cards, which offer significant performance upgrades from their previous generation by moving from 12nm to 6nm lithography, using cost-effective GDDR6 memory instead of NVIDIA’s HBM. The group discussed practical implications of model sizes and agreed on the need for standardized open-source models suitable for agentic workflows on the Internet Computer. Malith shared insights on performance bottlenecks in multi-GPU setups and highlighted advantages of NVIDIA’s advanced memory interfaces like NVLink. Lastly, the group acknowledged the necessity of evaluating emerging inference software such as Modular’s inference server as potential improvements over existing solutions.

Links shared during the call:

2 Likes
281

:spiral_calendar: Upcoming Calls & Topics

Date Topic Presenter
Apr 24 IConfucius – Fully On-Chain AI Agent on ICP icpp (onicai)
May 1 ANIMA demo (link) swift
May 8 MCP & Agent2Agent Protocol
May (tbd) Session on Efficient LLMs and DeepSeek
May (tbd) Session on instruction limit
Future Expert session on Active Inference implementations Invite TBD
3 Likes