A Fun & Absurd Introduction to Vector Databases
summary
Shows a hands‑on stack: a Unity front end, Kotlin/Spring Boot backend, and vv8 via Docker running offline; vectors use cosine distance and HNSW indexing and embeddings from ResNet50 and CLIP (text+image) to index ~50 weapons, ~700 Pokémon icons and D&D spells. Demonstrations reveal reliable semantic clusters for text and icons, brittle results for abstract images unless you add angles or more references, and failing audio matching when using OpenAI Whisper on guttural metal. Performance is millisecond nearest‑neighbor lookup; insertion is relatively slow and needs schema‑consistent dimensionality.
AI-written summary · about
chapters (14)
- 0:00Intro
- 1:11Goal of this talk
- 2:19Vectors
- 5:00Machine learning
- 6:35Vector databases
- 12:52Demo
- 17:32Implementing our own search
- 21:14Demo: HNSW
- 24:24Demo: Matrix
- 28:40Demo: D&D
- 33:52Demo: Pokemon
- 38:51Demo: D&D
- 43:24Demo: Duncan Hills
- 47:43Outro
related talks
46mA fun and absurd introduction to Vector Databases
48mA fun and absurd introduction to Vector Databases
47mA fun and absurd introduction to Vector Databases
46mA Fun & Absurd Introduction to Vector Databases
50mA fun and absurd introduction to Vector Databases
01h 23mVector Databases Demystified: Discover how they work
39mWhat's in a Vector Database?
41mGetting Started With Vector Databases