Physical AI
AI systems that interact with and operate in physical 3D environments, typically through robotics and embodied agents. Leverages World Foundation Models trained in simulation to understand and manipulate real-world spaces.
Key Developments
- NVIDIA Cosmos 3: An “Omnimodal World Model” representing a significant leap in physical AI capabilities. See: NVIDIA Cosmos 3: Omnimodal World Model for Physical AI and Robotics
- Spatial Intelligence & Data Scarcity: Fei-Fei Li, CEO of World AI, highlights how spatial intelligence addresses the critical bottleneck of robotics data scarcity. This approach was further solidified by World AI’s acquisition of Scenix (co-founded by Yunzhu Li), combining visions for scalable physical AI. See: Fei-Fei Li: Spatial Intelligence Solves Robotics Data Scarcity
Industry Trends (IBM 2026)
According to IBM’s 2026 AI trend predictions:
- Multi-agent Orchestration: Team of agents (Planner, Worker, Critic) collaborating under an “Orchestrator” layer. [[concepts/mu