May 15
Loading video player...
In this video, we break down one of the most important ideas behind modern AI: embeddings. Embeddings are how AI systems turn words, sentences, documents, images and other objects into vectors: lists of numbers that capture meaning. Once meaning becomes a vector, we can compare it, search it, cluster it, recommend with it and retrieve relevant information for LLMs. We’ll build the intuition visually, then use Python to create sentence embeddings, calculate cosine similarity and find the most similar sentence based on meaning instead of exact keywords. We will go through - What embeddings are - Why AI turns words into numbers - Meaning space intuition - Cosine similarity - Semantic search - Embeddings in RAG and LLM applications - Limitations of embeddings 💻 Code 👉 github repo: https://github.com/afterhoursml/Embeddings If you would like to dive deeper into related concepts - Principal Component Analysis (PCA): https://youtu.be/ejLauPnuK08 - Linear Regression: https://youtu.be/a5b1f77rXsk - K-Means: https://youtu.be/k-7B4tzX_h4 - KNN: https://youtu.be/EujLda19h9E - Why accuracy is not always the best metric: https://youtu.be/TSxtdKB2uH8 If you enjoy this kind of content, consider subscribing and leaving a like & comment. It really helps more curious minds discover the channel. Feel free to share in the comments what your favorite Machine Learning algorithm is. Thanks for watching! See you in the next one. 00:00 Intuition 00:52 Why it matters 01:37 Meaning space 02:53 Similarity 03:28 Python demo 04:42 Semantic search 05:27 PCA view 06:28 Applications 07:58 Limitations 09:11 Summary #machinelearning