/
2025-04-22

2025-04-22

Agenda

  1. Pathway to humanoid robotics, @Sandor Felber, 15 mins

  2. Leverage MLComms benchmark for AI Security, @Victor Lu , 10 mins

  3. Release 1.2 Review on Work stream 2, Edge Data Agent, @Qi Tang , 15 mins

 

 

Participants

@Wilson Wang @Chun Liu @Yike Gong @Mengyu Hang @Ning Jiang @Jianwen Pi @Jiayi Song @Lance Lim @Qingkun Li @Mo Li @Noe Otero @Tina Tsou @Wenhui Zhang @Hao Xie @Xun Chen @Ye Xu @Ying Wang @Yizheng Jiao @Zixiong Zhang @Akram Sheriff @anfernee.guan @Ashwanth Gnanavelu @Yona @HaruhisaFukano @Jenny Yang @李博睿 @刘秉伟 @LU CHENG @Rachel Roumeliotis @Ruoyu Ying @Sam Han @Shunli @Suhaas Teja Vijjagiri @Sumeet Solanki @TangQi @Tina Tsou @TinaTsou @Tom Qin @Victor Lu @万里 @Wilson Wang @Yin Li @Yu Wang - AIG

 

Recording

https://bytedance.us.larkoffice.com/minutes/obus2edi29fw15g1sp8m926x

 

Summary

The meeting discussed the progress and plans in multiple work streams within the project. The main contents included:

  • Pathway to Humanize Robotics: Sender shared about developing a reliable talent operation framework, resolving calibration issues and looking to expand with a new hire.

  • Work Stream 6 and Physical AI: Details of physical AI projects and relevant teams working on robotics and agentic AI.

  • EMEA Comments and Benchmarking: Victor presented EMEA comment's AI safety benchmark and Intel's keyword spotting benchmark.

  • Work Stream 2 Release 1.2: Qi discussed work stream 2 data and algorithms, including code generation on the edge and improving local knowledge infusion.

  • Next Steps and Actions: Discussions on creating wiki pages, evaluating agent performance and future work in different work streams.

 

Chapters

02:40 Infinite Hai Community Meeting: Updates on Robotics, AI Security and Network Reliability

This section begins with Tina Tsou introducing the meeting agenda. Sender shares progress on humanizing robotics, including resolving calibration issues and upcoming plans. There's discussion about network reliability for robotics, with various participants sharing insights. Sender also promotes Nairo robotics. Then, Tina Tsou mentions the Lark and Slack groups. Finally, she invites Victor Lu to discuss leveraging the benchmark for infinite AI, and Victor starts to share the relevant website.

26:27 Discussion on AI Safety, Benchmarks, Hardware Platforms and Related Projects

This section discusses various topics. It starts with AI safety benchmarks, suggesting relevant ones for edge use cases. There's talk about evaluating hardware platforms like Google Coral and Adam. The speaker shares their work on AI and the law, and an ML ops platform. They also discuss robot safety, liability, and events like robot half Marathons. Additionally, they mention a framework by NYU Professor Justin for software updates in devices.

41:20 Discussion on Action Items, Hardware for Robots and Work Stream Presentation

This section first discusses action items. Tina Tsou asks Victor about creating something on the wiki page. Victor isn't an expert but mentions collaboration with Jeff regarding Harvard-related hardware testing. Jeff needs time to think. They also talk about form factors of different dev boards and workstations for robots. Then, Tina Tsou suggests giving time for G to present work stream 2 for release 1.2. G tries to share screen but gets disconnected. Someone mentions going to MLC's and recalls a successful collab in UC Berkeley.

45:33 Presentation on Work Stream No. 2 and Related Discussions

This section focuses on work stream No. 2 about data and algorithms. The speaker reviews its background, goal of pushing LMS capabilities with data and algorithms. ETA's ambitious target is mentioned. Challenges like edge scenarios are discussed. Release 1.2 is presented, including its single agent framework and multi-step self-reflection mechanism. Future work on evaluation is proposed. Also, there are discussions on wiki page content, TSC voting, and unrelated comments about HP tools and a meeting with 16Z.