Models & ArchitectureCore

Vector

Definition
An ordered list of numbers that represents data in a high-dimensional space. In AI, vectors (embeddings) encode semantic meaning, enabling similarity search, clustering, and retrieval-augmented generation.
Why it matters
Vectors are the mathematical representation that makes semantic AI applications possible. When you convert text, images, or other data into vectors, you can compute similarity between any two pieces of content using simple math (cosine similarity or dot product). This is the foundation of modern search, recommendations, and RAG systems. Understanding vectors helps you reason about what AI search can and cannot do: two documents can be semantically similar (close in vector space) even if they share no keywords, which is why vector search outperforms keyword search for conceptual queries. The dimensionality and quality of your vectors directly determine the quality of every downstream application.
In practice
Modern embedding models produce vectors with 768-3,072 dimensions. OpenAI's text-embedding-3-large uses 3,072 dimensions; Cohere's embed-v3 uses 1,024. Higher dimensions capture more nuance but require more storage and compute. A typical RAG system stores millions of document vectors in a vector database and retrieves the top-k most similar vectors for each query. Cosine similarity is the standard distance metric. Vector databases like Pinecone, Weaviate, and Chroma are optimized for fast approximate nearest-neighbor search at scale. The vector representation has extended beyond text to images (CLIP), audio (CLAP), and even code (CodeBERT), enabling cross-modal search and similarity.

We cover models & architecture every week.

Get the 5 AI stories that matter — free, every Friday.

Know the terms. Know the moves.

Get the 5 AI stories that matter every Friday — free.

Free forever. No spam.