Explain What Machine Learning Actually Is to Someone Who Has Read the Headlines

You have read enough AI news to know what machine learning is called, but the explanation is always either too basic or too technical. This gives you the real version: what it actually does differently from traditional programming, what training a model means, the three types of machine learning and one concrete example of each, and the things these systems cannot do that the headlines suggest they can. No robots, no singularity.

Explain what machine learning actually is for someone who has read a lot of AI news but does not have a technical background and wants a grounded understanding rather than another article about robots taking over. Cover: the basic idea of how machine learning differs from traditional programming, using one concrete analogy; what a model is and what it means to train one; the difference between supervised, unsupervised, and reinforcement learning in plain terms, with one real-world example of each; what these systems actually cannot do that the headlines imply they can; and the three things most people who have not studied machine learning believe about it that are not quite right.