Developer Voices Semantic Search: A Deep Dive Into Vector Databases (with Zain Hasan)
10 snips
Oct 18, 2023 
Zain Hasan, a semantic search and augmented LLMs expert, joins the podcast to discuss the challenges of teaching large language models. They explore the concept of vector databases and their role in enhancing chat bots. The episode delves into optimizing search in a fictional service, the size and storage of indexes in vector databases, and the concept of multi-modality in vector search. The discussion also includes implementing semantic search at home using Weaviate and a conversation on Weaviate, an open-source database with managed instances.
AI Snips
Chapters
Transcript
Episode notes
Vector Databases Augment LLMs
- ChatGPT is a fine-tuned base model acting as a sentence completion tool using probabilities of word co-occurrence.
- It lacks real-time knowledge beyond training data, so vector databases augment it by providing updated, private context.
Retrieval Augmented Generation
- Retrieval augmented generation involves querying a vector database for relevant documents before prompting the LLM.
- This improves answer relevance and scales better than manually copy-pasting context with large document sets.
Vectors Capture Semantic Meaning
- Vector databases query numeric vector representations of data capturing semantic meaning.
- These vectors are black-box barcodes with latent space dimensions optimized for language prediction, not explicitly understandable features.