[4S1-OS-30a] OS-30

[4S1-OS-30a-04] A Study on Language Acquisition in Robots based on Large Language Models

〇Kazuki Miyazawa¹ (1. Osaka University)

Keywords:Robot, Large language Models, Language Acquisition

Recent advancements in deep learning have greatly enhanced large language models, enabling them to perform sophisticated language processing and be applied in various fields. Beyond processing language, these models are now learning from vast amounts of data, including images, sounds, and robot actions, to understand real-world information and make decisions. Unlike humans, who acquire language through personal experiences, these models learn from extensive data produced by various sources. This study explores how large language models, trained on the extensive linguistic experiences available on the web, can facilitate language acquisition in robots, offering a different perspective from human language learning. We discuss a novel approach to language acquisition for robots, using two mobile manipulators with the capability for movement and object manipulation, to investigate interactions with others and the environment, including language.