📊 Machine Learning: Learning From Data

Machine Learning is a specific approach to AI where instead of giving the computer explicit rules for every situation, you give it large amounts of data and let it discover the patterns itself. The computer "learns" by finding relationships in the data that allow it to make predictions or decisions about new data it has never seen before.

Here is the core difference in approach:

Traditional programming: A human writes rules. "If the email contains the word 'lottery' AND comes from an unknown sender AND has a link, mark it as spam." The programmer must think of every rule. If spammers change their tactics, the programmer must write new rules.

Machine Learning: You give the computer 10,000 emails labelled "spam" and 10,000 emails labelled "not spam." The ML algorithm analyses these examples and discovers its own patterns — patterns that might be too subtle or too numerous for a human to identify. When a new email arrives, the model uses the patterns it learned to predict whether it is spam or not. As spammers change tactics, you feed the new emails into the model and it adapts automatically.

Three Types of Machine Learning

Supervised learning is the most common type. The training data includes both the input and the correct answer (the "label"). For example: thousands of photos of cats and dogs, each labelled "cat" or "dog." The algorithm learns the visual patterns that distinguish cats from dogs. When shown a new photo, it predicts the label. Email spam filtering, medical image diagnosis, and credit card fraud detection all use supervised learning.

Unsupervised learning works with data that has no labels. The algorithm finds hidden patterns or groupings on its own. For example: given purchase data for 100,000 customers (with no labels), the algorithm might discover natural groupings — "budget shoppers," "premium buyers," "occasional splurgers" — that marketers can then target differently. Customer segmentation, anomaly detection, and recommendation systems use unsupervised learning.

Reinforcement learning learns by trial and error. The algorithm takes actions in an environment and receives rewards or penalties based on the outcomes. Over thousands of trials, it learns which actions lead to the best results. This is how DeepMind's AlphaGo learned to play Go (by playing millions of games against itself), how self-driving cars learn to navigate, and how robots learn to walk.

💡 Everyday ML You Already Use

Netflix recommendations: "Because you watched Stranger Things..." — ML analyses your viewing history and finds patterns shared with millions of other users to predict what you will enjoy next.

Spotify Discover Weekly: Analyses your listening habits, compares them to similar listeners, and generates a personalised playlist every Monday.

Instagram Explore: ML analyses which posts you like, comment on, and spend time viewing, then shows similar content from accounts you do not follow.

Google Search autocomplete: Predicts what you are typing based on billions of previous searches by other users.

Face ID on your phone: A neural network (deep learning) trained on millions of faces recognises your face in varying lighting, angles, and even with a mask.