Lies, Damn Lies, and AIs
summary
Argues that pairing deterministic Knowledge Graph stores with LLMs via RAG cuts hallucinations and adds explainability, citing a paper where SQL+RAG lifts accuracy 16% and a graph-backed RAG boosts it 54% accuracy uplift. Describes production patterns: vector embeddings to resolve graph nodes, graph neighborhood validation, Microsoft’s graph-RAG pipeline, and Yin-Choy’s symbolic knowledge distillation that extracts contradictions into a curated graph to train smaller, more accurate models. Practical demos and tool mentions (LangChain, Docker Gen stack, Neo4j) are in play.
AI-written summary · about
chapters (8)
- 0:00Introduction and setup
- 1:22Context and human bias
- 6:17Limitations of LLMs
- 10:50Knowledge graphs explained
- 16:31Solving hallucinations with RAG
- 20:23LLMs and knowledge graphs together
- 22:32Future of AI with knowledge graphs
- 26:21Closing and Q&A
related talks
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs
36mAgentic GraphRAG: AI’s Logical Edge
26mPractical GraphRAG Making LLMs smarter with Knowledge Graphs
38mLeveraging Knowledge Graphs for RAG: A Smarter Approach to Contextual AI Applications
42mMaking OSS AI Real with Knowledge Graphs
28mEnhancing LLMs with Graph Technology
20mPractical GraphRAG: Making LLMs smarter with Knowledge Graphs
49mPower up LLMs to be Factual Ninjas using GraphRAG