An embedding is a long series of numbers that represents the meaning of a text. Texts with related meanings get number series that lie close to each other. A vector database stores the number series so they can be searched quickly.
Embeddings make it possible to search by meaning rather than exact words. A search for 'terminating employees' can find a document about 'dismissals', even though the words differ.
Embeddings are the engine of RAG. The organisation's documents are split into sections, each section gets an embedding, and when an employee asks a question, the system finds the sections closest to it. Those are passed to the language model when it answers.
In practice
When a vendor offers an AI assistant that answers from your own documents, embeddings and a vector database are typically doing the work behind it. The quality depends on the documents. Outdated, duplicated or contradictory documents produce outdated, duplicated and contradictory answers.
The concepts in this glossary run through my AI courses and workshops, where we work with them on your own tasks.