[IEEE UKI-RAS]-[Prestigious Talk] Prof. Jun Tani --- From Embodied Language Grounding to Systematic Generalization: Developmental Origins of Compositionality in Robots

#artificial-intelligence #robotics #HRI #RAS #robots #learning

Prof. Jun Tani  (Okinawa Institute of Science and Technology, Japan)

Bio:   Jun Tani received the D.Eng. degree from Sophia University, Tokyo in 1995. He started his research career with Sony Computer Science Lab. in 1993. He became a PI in RIKEN Brain Science Institute in 2001. He became a tenured Professor at KAIST, South Korea in 2012. He is currently a full Professor at OIST. He is also a visiting professor at The Technical University of Munich. His current research interests include cognitive neuroscience, developmental psychology, phenomenology, complex adaptive systems, and robotics. He is an author of “Exploring Robotic Minds: Actions, Symbols, and Consciousness as Self-Organizing Dynamic Phenomena." published from Oxford Univ. Press in 2016.

Abstract: How can infants develop skills for action and language from relatively limited experience through interaction with the world? This seminar presents a developmental robotics perspective on this question by investigating how compositionality—the capacity to systematically recombine known concepts in novel situations—can emerge through embodied interaction. First, we introduce a brain-inspired neural model based on predictive coding, active inference, and the free-energy principle, in which language, vision, and action are learned jointly [1]. Robotic experiments demonstrate that compositional linguistic representations emerge through sensorimotor experience, enabling generalization to previously unseen language-action combinations. Second, we discuss how self-exploration and intrinsic motivation can further support the development of systematic generalization under sparse learning conditions [2]. By integrating active inference with reinforcement learning, robots learn not only to achieve goals but also to seek information and behavioral novelty. The results suggest that richer compositional experiences promote the emergence of structured internal representations that support generalization beyond training examples. Together, these studies suggest that compositional intelligence can emerge from embodied predictive learning and active exploration, offering a possible developmental account of how systematic generalization can arise from relatively limited experience.

 Prof. Jun Tani from the Okinawa Institute of Science and Technology, Japan,  will present a Prestigious Lecture on "From Embodied Language Grounding to Systematic Generalization: Developmental Origins of Compositionality in Robots" at 12:00 (UTC+1) on Oct 21st 2026. 

The talk will last 60 minutes distributed between the Prestigious Lecture, followed by questions & answers.