Machine Learning

The one-sentence difference from normal programming

In traditional programming, a human writes explicit rules, and the computer applies those rules to data to produce answers. Machine learning flips that completely: you give the computer data along with the correct answers, and it works out the rules itself. That reversal is the entire idea. It matters because some problems — recognising a face, translating a sentence, predicting what someone might want to watch next — are far too complicated for a human to write exact rules for by hand, but a model can find workable patterns from enough examples.

Three fundamentally different ways a model can learn

Almost every machine learning system falls into one of three learning styles. Supervised learning trains on examples that already have the correct answer attached, so the model learns to predict that answer for new, unseen examples. Unsupervised learning gets data with no answers attached at all, and instead looks for structure — grouping similar things together or finding patterns nobody labelled in advance. Reinforcement learning is different again: an agent takes actions in an environment and learns purely from reward and punishment signals over many attempts, the way a game-playing AI improves by playing itself millions of times.

Why this order matters

Nearly every advanced AI topic — neural networks, generative AI, large language models — is built on top of these same fundamentals, just at much greater scale. Getting the basic shape right here is what makes the more advanced modules below click instead of feeling like memorised jargon.