Language models & chatbots
An embedding is a list of numbers that represents the meaning of a piece of text, image, or other data, arranged so similar meanings produce similar numbers.
Computers are good at numbers and bad at meaning. An embedding model bridges the gap. Feed it a word, a sentence, or a photo, and it returns a fixed-length list of numbers, often several hundred long. The clever part is the arrangement: items with similar meaning get lists that are close together, and unrelated items land far apart.
You can picture each list as coordinates for a point in a space with hundreds of dimensions. 'Puppy' sits near 'dog.' A review saying 'loved it' sits near one saying 'fantastic.' That closeness can be measured, which turns fuzzy questions like 'what does this mean' into precise ones like 'how far apart are these points.'
Embeddings power semantic search, recommendations, clustering, and RAG. Inside a language model, every token is first turned into an embedding before anything else happens. Different models produce different embeddings, so you cannot mix numbers from two models.
Two support tickets, 'my screen is frozen' and 'the display will not respond,' share no key words but get nearly identical embeddings, so they are grouped together.