The Evolution of Memory in Humans and AI Agents

Memory shapes how we think, learn, and create. Humans have built memory systems for thousands of years: from oral traditions to memory palaces. Today, AI agents also need memory to be useful and reliable. They need to recall past events, connect knowledge, retrieve information quickly, and even forget the right things.

In this talk, we will look at the history of human memory and see what lessons it offers for AI. We will explore different types of memory:

• Episodic memory for storing events and experiences
• Semantic memory for facts and knowledge
• Working memory for short-term reasoning
• Vector or embedding memory for fast retrieval across large contexts

We will then see how each of these can be implemented in AI agents, what challenges arise, and what trade-offs we need to consider. The goal is to give you both inspiration from human history and concrete tools for building better agents with memory.