Robotic Foundation Models: Key Challenges in Physical AI

Session Overview

Korea and Japan are among the world’s top manufacturing powerhouses. However, both are facing rapid demographic decline and severe labor shortages. In this context, Physical AI—the fusion of AI and robotics—has emerged as a new paradigm, with Robotic Foundation Models (RFM) at its core.

Going beyond language-based LLMs, RFMs present a new foundation for replacing physical labor across a wide range of industries through manufacturing, logistics, and services. The key differentiating points of RFM can be summarized in three areas:

• Architectural Design: RFM requires a new architecture that goes beyond traditional approaches.
• Industrial Data Utilization: Effective collection and utilization of large-scale real-world industrial data is critical.
• Generalization & Adaptability: Ensuring that robots can be adapted across various platforms and use cases is essential for real-world scalability.

With this foundation, RFMs can expand into manufacturing, operations, and services, evolving from specialized models to versatile, general-purpose robotic solutions.

To achieve this, RLWRLD is building an ecosystem that connects top-tier research, global corporations, and innovative startups to advance Physical AI.

The vision is not merely about replacing human labor with robots—but about creating a future where robots free up human time for more meaningful and creative work, enabling Korea and Japan’s manufacturing industries to reinvent themselves for the next generation.