Google DeepMind

Physical AI data company Maxinsights has made its first public announcement. Based in Silicon Valley, the team began operating in China in March and says it has delivered a cumulative 2 million hours of egocentric data, with 1.5 million hours of high-quality data in stock and monthly collection capacity of 450,000 hours. Its self-developed MaxEgo models handle hand and pose tracking, SLAM, and video understanding annotation. Customers include Google DeepMind, Figure, 1X Technologies, Genesis AI, and Dyna Robotics.

Noiz AI, with The Hong Kong University of Science and Technology, Tsinghua University, Carnegie Mellon University, and Google DeepMind, has released HelixWorld 1.0, a real-time interactive audio-visual world model. It generates 24 FPS video and 48 kHz stereo audio together through a native Transformer, supporting interactive world exploration rather than fixed clips. Model weights and code are to be fully open-sourced in the coming weeks, according to the release.