Give Your LLMs a Left Brain
summary
Claim: combining knowledge graphs with LLMs via retrieval-augmented generation materially reduces hallucinations and improves explainability. Demonstrations use Neo4j + LangChain, a public LM-graph-builder that ingests docs (SPDX example), builds nodes/relations and embeddings, and answers queries accurately. Architecture uses vector search with graph context to fetch graph subgraphs then prime the LLM. Data scale examples: Panama Papers 11.5M docs (2.6TB) and model exposures ~500B words; vendor research and Gartner cite big reliability gains, estimated 50% of gen initiatives by 2025.
AI-written summary · about
chapters (15)
- 0:00Welcome
- 1:24LLMs reshaping development
- 3:32LLMs reasoning limits
- 7:06LLM hallucinations demo
- 14:13How LLMs process language
- 15:17Word vectors and Transformers
- 19:55Knowledge graphs for LLMs
- 22:00Fraud detection case study
- 26:20Graphs + LLMs reduce hallucinations
- 29:29Graph RAG architecture
- 31:35Graph RAG implementation patterns
- 33:42Governance with knowledge graphs
- 35:48Knowledge Graph Builder demo
- 40:25Future of LLMs with graphs
- 42:53Q&A
related talks
52mGive Your LLMs a Left Brain
41mEnhancing LLMs with Graph Technology
47mGive Your LLMs a Left Brain
28mEnhancing LLMs with Graph Technology
28mGrounded Intelligence Boosting LLMs with Real-Time Graph Context
23mKnowledge Graphs and LLMs in Action
52mAnatomy of an LLM
28mThe Present and Future of AI-Enabled Knowledge Graph Generation