What Is Embedding?

Embedding — a numerical vector representation of data — such as a word, sentence, image, or user — learned so that items with similar meaning sit close together in a high-dimensional space.

Embeddings are how modern AI turns messy real-world data into something a model can compute over, and they power semantic search and retrieval-augmented generation. They carry governance implications: embeddings can encode bias, and research shows sensitive information can sometimes be reconstructed from them, so they are not automatically a privacy-safe representation.

Source: Machine-learning literature