Embeddings convert text into vectors (lists of numbers) where semantically similar texts are close together in the vector space. The sentence 'the cat sat on the mat' and 'a feline rested on the rug' would have similar embeddings despite different words.
Embeddings power similarity search, retrieval-augmented generation (RAG), clustering, and classification. Models like text-embedding-ada-002 or Gemini's embedding model produce embeddings specifically optimized for these downstream tasks.