A Case-Based Persistent Memory for a Large Language Model

TL;DR

This paper advocates a case-based persistent memory architecture for large language models, arguing that CBR can benefit from DL/LLM advances to provide scalable, persistent memory across conversations. It surveys opportunities to model similarity with DL, scale case-bases to petabytes, and couple vector databases with LLMs to enable DeepCBR with ANNS. It reviews prior work integrating CBR and DL, and proposes a practical DeepCBR architecture that leverages FAISS/ANNS for efficient retrieval. The work emphasizes memory maintenance and explainability as key benefits toward AGI, aligning with industry trends toward persistent AI memory.

Abstract

Case-based reasoning (CBR) as a methodology for problem-solving can use any appropriate computational technique. This position paper argues that CBR researchers have somewhat overlooked recent developments in deep learning and large language models (LLMs). The underlying technical developments that have enabled the recent breakthroughs in AI have strong synergies with CBR and could be used to provide a persistent memory for LLMs to make progress towards Artificial General Intelligence.

Figures (1)

• Figure 1: Architecture of GPT-3.5 With External Memory Using a Vector Database, after Jia [20]