At a symposium held in Seoul, South Korea, Yann LeCun, the founder and chairman of AMI Labs, put forward the development direction for the next generation of AI. He is of the opinion that AI development is poised to transition from merely focusing on language learning to comprehending the physical world. He suggests that countries should harness world models capable of predicting behavioral outcomes and devising plans, in order to surmount the limitations that AI faces in industrial environments.

He highlighted that current AI systems exhibit shortcomings in understanding and controlling the physical environment. Relying solely on vast amounts of data for imitation learning, these systems struggle to guarantee reliability in fields such as robotics and autonomous driving. To tackle this issue, he introduced the Joint Embedding Predictive Architecture (JEPA). AMI Labs is currently developing an AI system based on this technology, which can autonomously formulate work plans. The aim is to empower AI to solve new tasks by leveraging previously learned content, eliminating the need for separate learning of operation methods for each individual task.

Yann LeCun also announced that he is set to expand AI development collaboration with the South Korean government and institutions. He emphasized that South Korea boasts exceptional talent and a solid foundation in high-quality technological education. Moreover, he underscored the importance of creating a research environment that can attract top global talent.