Yesterday we turned text into tokens. Today: how those tokens become meaning a computer can work with — embeddings. This one concept quietly powers search, RAG, recommendations, and clustering. Here's an interactive demo with a real "meaning map."
🧭 Explore the meaning map + do vector math: https://dev48v.infy.uk/ai/days/day13-embeddings.html
Meaning becomes distance
An embedding turns a word (or sentence, or image) into a list of numbers — a vector. They're learned so that similar things sit close together. On the demo's map, dog/puppy/cat/kitten cluster in one corner; car/truck in another.
You can do arithmetic on meaning
The famous one: king − man + woman ≈ queen. Because relationships are encoded as directions in the space, you can add and subtract concepts. The demo runs it and lands on "queen."
Measuring closeness
Cosine similarity scores how aligned two vectors are (1 = same direction, 0 = unrelated). "dog ↔ puppy" scores high; "dog ↔ car" near zero. That single number is what powers semantic search.
Where it shows up
Embeddings + a vector database (nearest-neighbour search) = the retrieval half of RAG, "related items," dedup, and clustering.
🔨 Build it (text → vectors → cosine similarity → nearest-neighbour search) on the page: https://dev48v.infy.uk/ai/days/day13-embeddings.html
Part of AIFromZero. 🌐 https://dev48v.infy.uk
Top comments (0)