The Anatomy of Memory in Humans and AI Agents
summary
Argument: LLMs are stateless with fixed context windows (expanding them scales quadratically), and training a modern model would take on the order of 100 million years at 1 billion ops/sec; practical memory needs require external systems. Redis ships an open-source agent memory server that extracts facts and preferences, applies TTL and recency-based decay, auto-summarizes when capacity hits a threshold, and integrates with Spring AI — demonstrated by a stock agent that persistently stores holdings and correctly resolves pronouns.
AI-written summary · about
chapters (7)
- 0:00Memory in humans and AI
- 1:03How LLMs handle memory
- 15:03Human memory anatomy
- 33:29Human vs LLM memory
- 34:51Agentic memory challenges
- 40:06Redis agentic memory demo
- 44:58Summary
related talks
25mWhat does it take to build long-term state for AI agents?
54mThe Anatomy of Memory in Humans and AI Agents
15mMemory-Augmented AI Agents: Enhancing Long-Term Reasoning and Context Awareness
27mMemory Engineering: Going Beyond Context Engineering
23mYour Agent Has Amnesia: Why Mobile AI Agents Need Memory to Be Useful
18mArchitecting Agent Memory: Principles, Patterns, and Best Practices
47mAgents and Memory Architecture
28mAgentic Memory