ShengShu Technology’s Motubrain Team Opens Physical AI Training to the Public

What happened

ShengShu Technology’s world model team began a public livestream of the full Motubrain 2.0 training process on October 21, planning to continue for half a month. The livestream will cover pretraining, post-training, agent task execution, and research into recursive self-improvement (RSI), showing how the embodied model is trained, tested, and iterated. Motubrain 2.0 is set to be open-sourced; the team says the research remains exploratory and its results need validation through further experiments and evaluations.

The story so far

ShengShu Technology’s world model team has begun a public livestream of Motubrain 2.0’s full training process ahead of its planned open-source release, allowing viewers to follow the embodied model’s training, testing, and iteration. The team plans to run the broadcast for half a month. The livestream will cover pretraining, post-training, agent task execution, and research into recursive self-improvement (RSI). The team says the research remains exploratory and its results require validation through further experiments and evaluations.

Keep reading

Trending now

1. 1Mistral and NVIDIA Announce Industrial Engineering AI Collaboration and Open Benchmarks
2. 2Mistral Releases Voxtral 3 Speech Models, Voice Agents API and LiveKit-Based SDK
3. 3DeepSeek Open-Sources DeepInfer and API Components, With Model and Hardware Limits
4. 4OpenAI Acquires Orchestra Bio to Add Drug Development Operations to Its Life Sciences Tools