Memoro: Using Large Language Models for Real-Time Memory Augmentation

Explore a 15-minute conference talk from CHI 2024 that introduces Memoro, a wearable audio-based memory assistant utilizing large language models for real-time memory augmentation. Discover how this innovative system infers users' memory needs in conversational contexts, performs semantic memory searches, and presents minimal suggestions through a concise interface. Learn about Memoro's two interaction modes: Query Mode for voiced queries and Queryless Mode for predictive assistance without explicit queries. Examine the results of a study involving 20 participants engaged in real-time conversations, which demonstrated reduced device interaction time, increased recall confidence, and preserved conversational quality. Gain insights into user preferences, experiences, and the potential of LLMs in designing minimally disruptive wearable memory augmentation systems.

Memoro: Using Large Language Models to Realize a Concise Interface for Real-Time Memory Augmentation