[seminar] "Project RAISE, Real-World Adaptive Intelligence Generated through Soft Embodiment" by Prof. Koh Hosoda, Kyoto University

Prof. Koh Hosoda
Graduate School of Engineering, Kyoto University

Talk title:  Project RAISE, Real-World Adaptive Intelligence Generated through Soft Embodiment

Abstract:  Recent advances in AI are opening new possibilities for robotics. Technologies such as large language models, reinforcement learning, and imitation learning now allow robots to generate useful behaviors with less manual programming than ever before. However, these approaches usually depend on enormous amounts of computation, energy, and training data.

Living creatures show a very different kind of intelligence. Even with limited energy and computational resources, animals can quickly learn and adapt to changing environments. One reason for this ability is "soft embodied intelligence": soft bodies interact with the environment in ways that naturally support adaptive behavior.

Project RAISE aims to build a new foundation for robotics by combining this biological idea of soft embodiment with modern AI. Rather than relying on computation alone, the project explores how intelligence can emerge from the interaction between body, brain, and environment.

By fusing advanced AI with the inherent adaptability of soft bodies, RAISE seeks to realize robots that can learn on site, adapt continuously in the real world, and work more flexibly, robustly, and safely in complex environments.