An LG Electronics-led consortium has won a significant government project to build Korea's own foundational AI for robots.
This project, worth about $35 million in public and private funding, aims to create a World Foundation Model (WFM) and an open Robot Foundation Model (RFM). Think of these as a central "brain" or operating system that can power various robots, allowing them to understand and interact with the physical world much more effectively. The goal is to develop a domestic technology stack for indoor mobile robots, reducing reliance on foreign systems.
This initiative is Korea's strategic answer to a global race in physical AI. First, companies like NVIDIA have set a high bar with their GR00T project, a foundation model for humanoid robots. By developing its own models, Korea aims to achieve 'AI sovereignty'—ensuring it isn't dependent on foreign technology for critical infrastructure in manufacturing and logistics.
Second, the timing is perfect due to strong domestic drivers. Korea has the world's highest robot density, especially in manufacturing. A standardized AI model that can upgrade this massive installed base offers a clear path to boosting productivity. Third, the new AI Basic Act, which took effect this year, provides the legal and safety framework needed to deploy these advanced robots in real-world factories and public spaces.
The winning consortium is well-equipped for the task. LG brings a whole ecosystem of hardware and its EXAONE AI, while KT provides its "K-RaaS" robot platform, a commercial-ready system to deploy the new AI models. This combination of industrial muscle and technical readiness transforms the project from a simple research grant into a credible path toward commercialization.
In essence, this project is a cornerstone of Korea's industrial strategy. It's a calculated move to build a self-reliant physical AI ecosystem, leveraging the country's unique strengths to compete on the global stage.
- Physical AI: Artificial intelligence designed to understand and interact with the physical world, primarily through robotics.
- Foundation Model: A large-scale AI model trained on a vast amount of data, designed to be adapted for a wide range of tasks. GR00T and GPT-4 are examples.
- Sim2Real: A technique in robotics where AI models are trained in a virtual simulation and then transferred to a physical robot to perform tasks in the real world.