AI has taken the world by storm and has become a household name. It is now important for everyone to learn about AI so that it is not just a "black box," or, as we call it in our book, Magic Box. However, AI is full of buzzwords, and we need an easier way to explain how AI works so both adults and kids can understand it correctly, and use it safely. Every child growing up as an AI native should learn how AI works before other competencies, such as evaluation of AI's strengths and weaknesses, explainability, human role, and ethics, can be cultivated properly. This illustrated activity book walks young readers through three steps of the AI process: Collect, Train, Decide. By following the story of Connor and Atena, two siblings who are trying to convince their mother to get them a puppy, readers will discover the foundation of AI algorithms known as artificial neuron networks or deep learning. Big concepts like "machine learning", "AI models", and "model training" become much easier to understand as readers complete our step-by-step activities in tandem with our story and characters. Readers will learn a math-based approach to artificial neural networks and deep learning. We designed this book for readers aged 9-15 who can do simple arithmetic (multiplication and addition in sequence) and algebraic formulation. Each chapter ends with an activity section where readers can practice the skills they have just learned. A final challenge project is provided in the end of the book with a thoroughly explained example with fully worked out solution. Answer key to all activities is provided in the back of the book. This book is meant to be written in! Use a pencil to annotate key concepts and new skills in the text, as well as to work out the problems in the activity sections. We hope this book will answer many questions about how AI works, and give you the knowledge and skills to use, evaluate, and investigate AI in your everyday life! Expected learning outcomes for AI literacy: Readers will be able to Construct a data table for training a classification model, with features, targets, and labels properly one-hot encoded as needed. Explain the mathematical steps of training an artificial neuron, especially how weighted sum and step function work, errors are defined, and weights are updated. Explain how to use a trained neuron model to make predictions on new data records. Praise for AI The Magic Box: "This book introduces core AI concepts like training, weights, and supervised learning through an engaging story of siblings learning from their data scientist mom. As a developmental psychologist, I find that the combination of storytelling, drawings, and exercises makes AI less intimidating and more approachable, and fun for young readers." - Dr. Pilyoung Kim, Director of Center for the Brain, AI, and Child, University of Denver "AI: The Magic Box is an engaging resource for young readers and beginners. It provides a solid foundation in AI education and serves as an excellent resource for introductory programming and AI classes. The characters' journey to adopt a hypoallergenic dog adds a practical and enjoyable element to learning complex AI topics, such as classification algorithms and model training while maintaining scientific accuracy." - Dr. Hyesung Park, Associate Professor of Information Technology, Georgia Gwinnett College "AI The Magic Box serves a need with a printed resource that teachers can use without computers, while also meeting standards related to ELA and informational texts. The AI theory is presented in a relatable way and helps unpack the "magic box" behind many of our current technologies." - Charlotte Dungan, Chief Learning Officer, Mark Cuban Foundation