Embedding models
Embedding is about representing existing text, images or videos as numerical values.
These models helps to convert the text, images and videos into a numerical representation and store them in vector database.
Later, these numerical representations can be used to find similar text, images or videos. This is done by comparing the numerical values of the text, images or videos and finding the closest match. This is called as vector search.
This how RAG concept works where additional contextual data is retrieved from the vector database and passed to LLM to generate the final output.
The term "embedding" in AI comes from the idea of embedding something complex into a simpler, often lower-dimensional space while preserving its essential characteristics and relationships.
This means, placing the text, images and videos inside another representation space.