Reducing LLM Hallucinations with Retrieval Augmented Generation: A Conversation with Ofer Mendelevitch
In this episode, Joshua Schoen of Work AI discusses the complexities and innovations in the field of Retrieval Augmented Generation (RAG) with Ofer Mendelevitch from Vectara.
They delve into topics such as the importance of accurate data feeding to Large Language Models (LLMs) to reduce hallucinations, the capabilities of vector databases, and how Vectara provides a comprehensive RAG-as-a-Service solution. Offer also shares his journey from working with early GPT models to co-founding Vectara. They explore various use cases for Vectara's technology in legal, healthcare, and other sectors, and end with a demo showcasing the practical applications of RAG in legal document management. 00:00 Introduction to AI and Hallucinations
01:29 Offer's Journey with GPT and
Syntegra
02:42 Joining Vectara and Its Mission
03:06 Vectara's Growth and Funding 03:31 What is Vectara? 06:54 RAG vs. Fine-Tuning
09:30 Reducing Hallucinations with RAG
12:39 Building a RAG Pipeline with Vectara
14:45 Customer Use Cases and Future of LLMs
18:51 Live Demo of Vectara's Capabilities
22:00 Conclusion and Final Thoughts
Share this post