Data → Machine Learning & AI
Embedding
A numeric representation of data that captures semantic similarity in a vector space.
Motivation
An embedding solves the problem of representing text, images, or other objects as vectors that preserve useful similarity.
Where it fits
It belongs to AI applications and search/retrieval systems.
Mental model
Similar things should land near each other in vector space, making semantic search possible.
Important details
Embeddings are used for retrieval, recommendations, clustering, deduplication, and RAG.
Common mistakes
- Assuming vector similarity always means factual relevance.
- Mixing embeddings from incompatible models.