Improving Retrieval Systems
Jason shares his journey from working on text embeddings in 2016 to addressing Gen AI challenges today. He highlights the limitations of relying solely on external embedding models, emphasizing the need for tailored embeddings that reflect user preferences more accurately. The conversation reveals a common misconception: tuning generation often overshadows the importance of providing the right context for large language models.In this clip
From this podcast
The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)
Why Your RAG Pipeline Is Broken, and How to Fix It with Jason Liu - 709
Related Questions
Is there anyone taking a different approach to prompt engineering for large language models that makes the process more accessible to a wider audience, as discussed in the episode Why Your RAG Pipeline Is Broken, and How to Fix It with Jason Liu - 709 and the clip Diagnosing AI Problems?
Do I get it right that a Retrieval Augmented Generation (RAG) system can retrieve data in addition to its training data as discussed in the episode Building LLM-Based Applications with Azure OpenAI with Jay Emery - 657 and the episode with Cohere co-founder Nick Frosst on building LLM apps for business?