AI & LLMs
How computers find things by meaning
Old-school search matched words. Type "dog" and it found the pages containing "dog." But "how do I look after my new dog" and "puppy care tips" share almost no words while meaning nearly the same thing. Modern search bridges that gap by turning meaning into location.
Turning meaning into coordinates
An embedding model reads a piece of text and outputs a list of numbers — a vector. You can think of that vector as coordinates for a point in space. The trick the model is trained to pull off: texts that mean similar things land close together, and unrelated things land far apart.
Searching by distance
Once everything lives in this space, search becomes geometry. Embed the user's query into the same space, then find the nearest points — typically by cosine similarity, a measure of how closely two vectors point in the same direction. The closest neighbours are your most relevant results, even if they don't share a single word with the query. That's semantic search.
Embeddings turn meaning into a location. "Related" becomes "nearby." Search becomes: put the question on the map, grab whatever sits closest.
Where you meet it every day
- Search and recommendations that understand intent, not just keywords.
- Deduplication — spotting that two support tickets are really the same issue.
- RAG (retrieval-augmented generation) — the step where an AI assistant fetches the most relevant documents before answering, so it's grounded in your data rather than guessing. The "retrieval" there is exactly this nearest-neighbour search.
It's a deceptively simple idea with enormous reach: teach a model to place meaning on a map, and suddenly "find me things like this" is just "find me what's close."