Machine Learning (ML)
A way of building software that learns patterns from examples rather than following rules a programmer wrote by hand. Show it enough labelled photos of cats and it works out for itself what a cat tends to look like.
Most traditional software does exactly what a programmer told it to do, step by step. Machine learning flips that around. Instead of writing the rules, you give a system lots of examples and let it work out the rules itself.
Say you want to tell apart photos of cats and dogs. Writing rules by hand would be hopeless: how do you describe, in code, the difference between a cat’s face and a dog’s? With machine learning you show the system thousands of labelled photos, and it gradually adjusts its internal settings until it can tell the two apart, often on animals it has never seen before. That process of adjustment is called training.
The catch is that a machine learning system is only ever as good as the examples it learns from. Feed it biased or patchy data and it will happily learn the wrong lesson, which is why so much of the real work is in gathering and cleaning the data rather than the maths itself.
Almost everything people mean by “AI” today, from the models behind chatbots to the vision systems in self-driving cars, is built on machine learning.