machine learning
Machine learning is how a computer gets better at a job by studying examples, instead of being handed a rulebook. Imagine teaching a child to recognize a cat: you don't recite a definition of "cat," you just point at hundreds of them until the pattern clicks. A machine learning program works the same way — show it thousands of photos labeled "cat" or "not cat," and it gradually tunes itself until it can spot a cat it has never seen before.
The shift is profound. In old-style programming, a human writes out every rule by hand: if the email contains "free money," mark it as spam. But spammers are clever, and the rules never end. Machine learning flips this around — you feed the program a pile of emails already sorted into spam and not-spam, and it discovers the telltale patterns on its own. That is why it now powers everything from your photo app to the suggestions that finish your sentences.
One common myth is that the machine "understands" what it learns. It doesn't — it has simply found statistical patterns that work. Feed it biased or sloppy examples and it will faithfully learn the bias too. The model is only ever as good as the data it learned from, which is why the examples matter far more than most people expect.
Don't write the rule "if pixels form whiskers and pointy ears, it's a cat." Show 10,000 labeled photos, and let the program find that rule itself.
Rules-out, examples-in: the core reversal at the heart of machine learning.
The term was coined in 1959 by Arthur Samuel, an IBM engineer whose checkers-playing program improved by playing against itself, tuning its own sense of which board positions were good or bad — learning from experience rather than from hand-written instructions.