Salomon Kabongo

Hello World to AI | Machine Learning From Scratch: Teach a Computer to Add

@DrKabongo·août 2026

What does it actually mean for a computer to learn? In this one-hour introduction to Artificial Intelligence and Machine Learning, we start with one of the simplest things a computer can do: addition. But instead of programming the rule of addition directly, we take it away. The computer is not told that 2 + 3 = 5. Instead, it receives examples and must gradually rediscover the mathematical relationship on its own. And we build the entire learning process from scratch. ❌ No PyTorch. ❌ No TensorFlow. ❌ No scikit-learn. We derive the calculations by hand, implement the learning process in pure Python, and follow the model from randomly initialized parameters to a function that can approximate addition. This is the “Hello World” of AI: a simple problem used to understand the fundamental ideas behind much more sophisticated systems. *What You Will Learn* * Traditional programming vs. machine learning * Training examples, inputs, and targets * Parameters, weights, and bias * Error, loss, and why we need a loss function * Derivatives and partial derivatives * How gradients tell us which direction to move * Gradient descent & the role of the learning rate * Updating model parameters manually * How these ideas connect to neural networks and modern AI The objective isn't simply to write code that produces the correct answer—it's to understand *why* learning works. If terms like weights, bias, loss, derivatives, gradients, gradient descent, and optimization have ever seemed mysterious, this tutorial builds them from first principles so they feel less like magic and more like math. *Video Chapters* 00:00 — Introduction to AI: Teaching a Computer to Add 01:07 — Traditional Programming vs. Machine Learning 23:40 — How Does AI Actually Learn? 25:53 — Understanding Gradient Descent 31:31 — Math Refresher: Chain Rule & Power Rule 36:34 — The Forward and Backward Pass 42:09 — Intuition: The Blindfolded Hiker 43:35 — Building the Addition Dataset 45:33 — Overfitting, Underfitting & Data Distribution 47:17 — Initializing a Neural Network in Pure Python 48:32 — Training the Model From Scratch 55:43 — Model Inference: Testing on New Data 58:17 — Conclusion: Understanding Model Parameters *Join the Dr. Kabongo Community* Have a question about the lesson? Want to discuss AI, Machine Learning, mathematics, programming, or future tutorials? 💬 Discord Community: https://discord.gg/PUbkyrUcaa 📱 WhatsApp Channel: https://whatsapp.com/channel/0029VbDR33v6hENldqLbRe24 🌐 Website: https://www.drkabongo.com This is the beginning of the Hello World to AI series, where we build AI and Machine Learning concepts from first principles before moving to increasingly advanced systems like PyTorch, deep learning, LLMs, NLP, Computer Vision, RAG, and AI agents. If you find this approach useful, subscribe so you do not miss the next lesson! #ArtificialIntelligence #MachineLearning #AI #MachineLearningFromScratch #GradientDescent #NeuralNetworks #DeepLearning #LearnAI #Programming