World Labs

Fei-Fei Li proposes a functional taxonomy of world models, categorizing them into renderers and simulators based on the POMDP framework to resolve conceptual confusion in AI, emphasizing that spatial intelligence requires learning statistical structures of space-time physics rather than text alone.

VAST secured nearly $200M in new funding and officially disclosed its world model roadmap, Project Eden, pioneering a decoupled architecture of state evolution and visual rendering to enable persistent multi-user interaction, modular reuse, and linearly scalable compute for AI-native sandboxes and embodied intelligence simulation.

空间智能与机器人技术融合将重塑AI未来，World Labs收购SceniX推动物理世界理解突破。

Fei-Fei Li defines the three functions of the world model: rendering, simulation, and planning.

VAST decoupled world state from visual rendering to build Project Eden — the world’s first deterministic model with independent state maintenance and multi-user interaction, breaking through limitations of video generation and static 3D reconstruction for true dynamic AI environments.

JunoBrain proposes a cognitive world model based on cognitive science to achieve higher-level embodied intelligence and has secured billion-level funding.

世界模型（World Model）被过度使用，其实际包含三种类型：渲染器、规划器和模拟器，各有不同功能。

Fei-Fei Li介绍SceniX与World Labs合作开发R2S2R引擎，通过模拟器加速机器人训练。

生成式世界模型构建的R2S2R模拟引擎可将物理任务转化为可控虚拟环境，加速机器人训练并降低试错成本。

文章探讨了sim-to-real对齐模拟在机器人训练中的应用，但信息密度较低，缺乏具体技术细节。

AI应改变教育方式，强调人类资本的重要性，并推动教育系统改革。

推文宣布SceniX加入World Labs，但未提供技术细节或工程价值。

文章信息密度不足，未详细说明R2S2R技术细节，仅提及项目愿景和初步成果。

Fei-Fei Li宣布SceniX加入World Labs，推动空间智能发展，团队在高保真模拟中训练机器人并实现实际部署。