← Videos
Search By Meaning -  Embeddings

AI from the Ground Up8 Nov 2026 · 4:40

Search By Meaning - Embeddings

You search for “What computer should I take?” The note says “Bring a laptop.” How can search connect those phrases when the words differ?

From the episode description

You search for “What computer should I take?” The note says “Bring a laptop.” How can search connect those phrases when the words differ? This introduction to embeddings follows text into a numerical representation, compares a query with candidate passages, and brings the result back to its original source. A simple arrow diagram makes cosine similarity easier to see without requiring advanced maths. In this episode: • What an embedding and a vector represent. • How compatible query and passage representations can be compared. • What cosine similarity tells us in a small example. • Why a similar passage can still be outdated or unsuitable. • How keyword and vector search can work together. Workshop Notes is fictional. The two-dimensional vectors are hand-chosen to explain the calculation; they are not measured embeddings of the sentences or a model’s internal map of meaning. Try comparing paraphrased questions with an exact identifier in a search system you use. Inspect the returned passages and their versions. A useful search result gives you somewhere to look—it does not finish the evidence check. ResearchNode: AI from the ground up • Episode 05 of 10. Previous: Why AI Can Sound Confident and Still Be Wrong Next: How AI Uses Your Documents: RAG Explained Presented by Michelle, an AI-generated presenter.

Descriptions and chapter markers are supplied by the channel. They are not a transcript or an independent check of every claim in the video.