"Embeddings - the lesser known Hero of AI"
summary
Explains embeddings as 1,536‑dim vectors (uses text-embedding-ada-002 and text-embedding-3-small), compares them with cosine distance, and shows concrete code: OpenAI PHP client, Symfony cache, and saving vectors to Redis/Redisearch or vector stores (Pinecone, PGVector, MySQL 8 / MariaDB preview). Demonstrations include Monarch + woman ≈ Queen (cosine ≈ 0.08), using <0.1 as “same”, translation quality checks and sentiment/nearest‑neighbor searches with exact distance thresholds and working PHP examples.
AI-written summary · about
chapters (18)
- 0:00Introduction to embeddings
- 1:47What is an embedding
- 4:18Math with embeddings
- 8:17Generating embeddings in PHP
- 12:07Embedding arithmetic demo
- 13:32Translation quality check
- 16:49Normalizing unstructured data
- 19:26Testing bias in embeddings
- 23:04Bias in AI models
- 26:16Finding nearest neighbors
- 27:22Sentiment analysis with embeddings
- 29:55Applications in Zoom calls
- 31:00Finding similar products
- 31:42Search by intention
- 32:23Storing embeddings
- 34:53Alternatives to OpenAI embeddings
- 37:05History of embeddings
- 38:06Acknowledgments and Q&A
related talks
23mEmbeddings: Core Concepts for Developers
27mData with Purpose: The Power of Embeddings
28m"Embeddings: How Computers Learned to Read"
48mUnderstanding Embedding Vectors and the RAG Pattern in AI Solutions
48mEnter the Brave New World of GenAI with Vector Search
16mEmpower AI: Building Smarter Solutions with Embeddings
01h 46mVector Databases (for GenAI & beyond)
33mEnter the Brave New World of GenAI with Vector Search