Machine Learning

The discipline of making machines improve at a task through exposure to data rather than through explicit programming. Arthur Samuel’s 1959 definition still holds: the field of study that gives computers the ability to learn without being explicitly programmed.

At its core it is function approximation under uncertainty: given examples, infer a mapping that generalizes beyond them. Everything else — gradient descent, backpropagation, attention mechanisms — is machinery in service of that single idea.

In this garden

• Artificial intelligence — the broader ambition that ML currently powers
• Data — the raw material; a model is only as good as its distribution
• Statistics — the mathematical backbone ML never really escaped