Machine Learning Street Talk (MLST) cover image

Jay Alammar on LLMs, RAG, and AI Engineering

Machine Learning Street Talk (MLST)

00:00

Mastering LLMs: Embrace Embeddings

To effectively leverage large language models (LLMs), one can choose to either consume commercial LLMs or build systems that utilize them. A key aspect for those building systems is understanding embeddings, as they provide a deeper engineering capability beyond basic prompt engineering. Embeddings are crucial for enabling reliable behaviors and driving semantic search, especially in the context of retrieval-augmented generation. Semantic search benefits from embeddings in two main ways: through dense retrieval, which allows for efficient searching of text archives based on proximity in the embedding space. The advancements in this area over the last few years, particularly with models like BERT, have significantly enhanced the performance of semantic search processes.

Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner
 Get the app