Achieving No. 1 in Physical AI Through Localization of World Models: "If We Don't Act Now, China Will Take the Lead"

World Model and Data Are the Core of Physical AI
Establishing a Pan-Government Data Management System

"If we do not accelerate our efforts in physical artificial intelligence (AI) now, countries like China will inevitably take the lead. We will speed up the development of core technologies such as the Robot Foundation Model (RFM) and World Model to establish ourselves as the global leader in physical AI."

On July 1 in Seoul at the KT West Building in Gwanghwamun, the Ministry of Science and ICT announced its 'Strategy to Secure Core Competitiveness in Physical AI,' stating, "We will actively work to secure the three foundational technologies of physical AI: foundation models, world models, and computing platforms."

Lee Dokyu, policy chief at the Ministry of Science and ICT, said, "South Korea aims to be among the world's top three in AI, but with large language models (LLM), the fact is that as a latecomer, we are limited because these models require vast amounts of data for training. However, physical AI has not reached that stage yet, so now is the right time to act."

He explained that since physical AI is still in its early stages, the United States is focusing on AI full-stack technologies, while China is leveraging its hardware strengths in an intense competition for dominance. He added that if South Korea, with its world-leading manufacturing base, can bring its capabilities together, the country could surpass both the US and China to become the global leader.

Lee highlighted the development of the world model and data collection as the top priorities for achieving leadership in physical AI. "The world model predicts and simulates changes in the real world to support the learning and decision-making of physical AI," he said. "Given the urgency to take the lead in physical AI, we plan to accelerate the simultaneous development of the RFM and the world model." He stressed that while the RFM is a large AI model trained to perform various tasks, the world model is an AI that simulates the real world internally; combining these two technologies is vital to achieving a truly general-purpose physical AI.

In particular, the government aims to strengthen its competitiveness by developing a domestically produced world model. Park Taewan, Director of ICT Industry Policy at the Ministry of Science and ICT, pointed out, "If we simply use overseas world models as they are, data from domestic sites could be directly exported in the training process. In security and advanced industries such as nuclear power and semiconductors, failing to develop our own world model would only help grow overseas companies' data assets."

Urgent Need to Aggregate and Manage Dispersed Data Across Sectors

Building a system to collect and comprehensively manage data scattered across various sectors is also an urgent task. Lee said, "The key competitive edge in physical AI is data," noting that data collected in the field is not being fully utilized and that the challenge is to aggregate data dispersed across different ministries and sectors.

Accordingly, the government will consolidate sector- and ministry-specific data into a pan-government data library and establish a system for comprehensive management. The plan is to actively support the creation of a foundation where companies can freely access, train, and validate the general-purpose data they need.

Song Changjong, head of the Device AX Innovation Team at the Ministry of Science and ICT, explained, "China is building data factories at both the national and company levels to collect data from all directions. South Korea's Ministry of Science and ICT and the Ministry of Trade, Industry and Energy are also preparing a system to consolidate government-wide data." However, he noted that some conglomerates are reluctant to share their company-specific data, and efforts are underway to address these concerns.

Since there is currently no benchmark for evaluating the performance of physical AI and robots, the government also plans to develop one. Lee added, "We will strengthen cooperation with international organizations and take the lead in standardization by supporting specialized research labs. We will also establish a legal framework to comprehensively support technology development, validation, and commercialization."

This content was produced with the assistance of AI translation services.