Explore an innovative approach to interactive task learning (ITL) in this conference talk from CHI 2024. Discover how VAL, a new ITL system, integrates large language models (LLMs) with symbolic processing to overcome traditional limitations in natural language parsing. Learn about VAL's ability to incrementally acquire hierarchical task knowledge from human instruction, resulting in interpretable and generalizable knowledge. Examine the system's performance in a video game setting, where users successfully taught VAL using natural language. Gain insights into the potential of this LLM/symbolic integration philosophy for advancing human-AI interaction and knowledge acquisition in various domains.