Embeddings are a powerful technology that can be used to solve various problems. They involve converting content into arrays of numbers, which represent the semantic meaning of the content. Embeddings can be used to find related content, perform searches, analyze code, and more. The author discusses their experiences using embeddings, such as building a “related content” feature for their blog, calculating cosine similarity distances, and exploring different embedding models like Word2Vec and CLIP. They also talk about their own tools, LLM and Symbex, which facilitate working with embeddings. The author emphasizes the usefulness of embeddings and encourages caution when relying on proprietary models.
https://simonwillison.net/2023/Oct/23/embeddings/