Amy Hodler and Dave Bechberger dive into the crucial role of memory in advanced AI systems, especially at the intersection of graphs, knowledge graphs, and generative AI.Dave Bechberger, currently focusing on MCP servers, agentic memory, and semantic data layers, explains that memory is fundamental because standard LLM calls are atomic and lack recollection of prior interactions. An agent without memory lacks continuity for complex user interactions.The discussion breaks down three key types of memory and how graphs apply:Episodic Memory: Transactional details are directly integrated into the context.Short-Term Memory: Session-based interactions that require compaction or summarization.Long-Term Memory: For extracting and storing patterns, trends, and preferences across multiple interactions.

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related