a visual course

How Machines
Understand Meaning

(Embeddings)

How AI turns words into points in space — the idea behind search, recommendations and RAG.

A free, visual course on embeddings — how machines represent the meaning of words, sentences and images as vectors in a space where closeness means similarity. The foundation of semantic search, recommendations and retrieval-augmented generation.

in the real world

Embeddings quietly power semantic search, recommendations, and the retrieval that lets AI chatbots cite real documents (RAG) — the workhorse under modern AI.

Start with a single word →

Part 1 — Meaning as numbers

1

Meaning as a place in space

Turn a word into a list of numbers — a point in a space where location means meaning.

2

Closeness is meaning

Related words sit near each other — and cosine similarity measures exactly how near.

Part 2 — Building the space

3

Learned from the company they keep

Embeddings aren't hand-coded — a model learns them by predicting words from their neighbors.

4

Doing math on meaning

king − man + woman ≈ queen: analogies fall out as simple vector arithmetic.

Part 3 — Embeddings everywhere

5

Embeddings for everything

Sentences, images, products, songs — anything can become a vector in the same kind of space.

6

Finding things by meaning

Nearest-neighbor search over embeddings powers semantic search, recommendations and RAG.

Part 4 — Out in the world

7

Where this shows up

Semantic search, recommendations, RAG, image search and dedup — the workhorse in action.