What is an Embedding? Vectors Explained

Preporato | AI for Engineers · Beginner ·🧠 Large Language Models ·1mo ago

About this lesson

Embeddings Explained: what they are, how they're trained, and why they power semantic search, RAG, and every large language model. Try it yourself — hands-on embedding & RAG labs in your browser: https://preporato.com/labs CHAPTERS 0:00 What is an embedding? 0:14 Why embeddings exist (king vs monarch) 0:32 A list of numbers (king → vector) 0:58 Same trick for images, code, audio (CLIP) 1:22 How they learn — the guessing game 1:52 Contrastive learning (modern embedding models) 2:23 Now the fun part 2:33 How many numbers? (Word2Vec 300 → text-embedding-3-large) 2:53 Cosine similarity (the angle = the meaning) 3:14 Vector arithmetic (king − man + woman = queen) 3:34 Embeddings inside every LLM 3:56 Where you use them — semantic search, recs, anomaly, RAG 5:02 Recap — turn anything into a vector, meaning becomes math

Original Description

Embeddings Explained: what they are, how they're trained, and why they power semantic search, RAG, and every large language model. Try it yourself — hands-on embedding & RAG labs in your browser: https://preporato.com/labs CHAPTERS 0:00 What is an embedding? 0:14 Why embeddings exist (king vs monarch) 0:32 A list of numbers (king → vector) 0:58 Same trick for images, code, audio (CLIP) 1:22 How they learn — the guessing game 1:52 Contrastive learning (modern embedding models) 2:23 Now the fun part 2:33 How many numbers? (Word2Vec 300 → text-embedding-3-large) 2:53 Cosine similarity (the angle = the meaning) 3:14 Vector arithmetic (king − man + woman = queen) 3:34 Embeddings inside every LLM 3:56 Where you use them — semantic search, recs, anomaly, RAG 5:02 Recap — turn anything into a vector, meaning becomes math

Watch on YouTube ↗ (saves to browser)

Sign in to unlock AI tutor explanation · ⚡30

Related Reads

Claude Sonnet 5 vs Fable 5: What I’d Actually Use in Production

Compare Claude Sonnet 5 and Fable 5 for production use, understanding their strengths and weaknesses

Dev.to · Marco Reyes

How I ended up contributing five ways to shape an LLM's output

Learn how to shape an LLM's output into structured data with five practical methods

Dev.to · Yatin Davra

10 AI Prompts That Every Professional Should Know

Learn 10 essential AI prompts to boost productivity and efficiency in the workplace

Medium · ChatGPT

DPO vs SFT vs RLHF: Which Training Method Does Your Model Actually Need?

Learn when to use DPO, SFT, or RLHF for fine-tuning your LLMs and understand the complexity of each method

Chapters (13)

What is an embedding?

0:14 Why embeddings exist (king vs monarch)

0:32 A list of numbers (king → vector)

0:58 Same trick for images, code, audio (CLIP)

1:22 How they learn — the guessing game

1:52 Contrastive learning (modern embedding models)

2:23 Now the fun part

2:33 How many numbers? (Word2Vec 300 → text-embedding-3-large)

2:53 Cosine similarity (the angle = the meaning)

3:14 Vector arithmetic (king − man + woman = queen)

3:34 Embeddings inside every LLM

3:56 Where you use them — semantic search, recs, anomaly, RAG

5:02 Recap — turn anything into a vector, meaning becomes math

5 Levels of AI Agents - From Simple LLM Calls to Multi-Agent Systems

Dave Ebbelaar (LLM Eng)