13.1 Episodic vs Semantic Memory

The Two Core Components of an Agent's Mind

For an AI agent to exhibit sophisticated intelligence, it needs a robust memory system. Inspired by human cognition, we can divide an agent's long-term memory into two fundamental types: Semantic Memory and Episodic Memory. Understanding the distinction is key to building agents that can both recall facts and remember personal experiences.

Semantic Memory: The Library of Facts

This is the agent's knowledge base of general, context-independent facts about the world. It's like an encyclopedia. It contains information such as "The sky is blue," "Paris is the capital of France," or technical data learned from documents.

• Nature: Structured, factual, objective.
• Implementation: Often stored in vector databases for fast, semantic searching of documents or knowledge graphs.

Episodic Memory: The Personal Diary

This memory stores specific events and experiences from the agent's past interactions, anchored in time and context. It's the agent's personal history, containing records like "At 10:30 AM, the user asked me about the weather in London," or "In our last session, we discussed project Alpha."

• Nature: Sequential, autobiographical, subjective.
• Implementation: Can be a chronological log of conversations, a timeline of events, or summaries of past interactions stored with timestamps.