Talk at Voxel51: Reducing Hallucinations in ChatGPT Using Knowledge Graphs
Gave a talk at the Voxel51 AI, Machine Learning and Computer Vision Meetup on how knowledge graphs can reduce hallucinations in ChatGPT and similar AI systems.
The talk demonstrated using LangChain and Neo4j to ground LLM outputs in structured knowledge, showing how graph-based retrieval produces more accurate and contextually grounded responses compared to vanilla RAG approaches.
Alongside great talks from Harpreet Sahota (Data-Centric AI Competition on Hugging Face) and Tim Spann (Sensor Data in RAG with Milvus).