AI, ML, and deep learning

Three words, one nesting

You hear "AI", "machine learning" and "deep learning" used as if they were the same thing, or as if you had to choose between them. They are not rivals. They fit inside one another like a set of nested boxes: the biggest box is AI, machine learning sits inside it, and deep learning sits inside that. Understanding the nesting is enough to follow almost any conversation about the technology.

AI: the big idea

Artificial intelligence (AI) is the broad goal of getting computers to do things that normally need human intelligence: recognising speech, translating languages, spotting patterns in images, making a recommendation. It is an umbrella term, not a single product. When a phone unlocks by looking at your face or an app predicts your commute, that is AI at work, quietly, in the background.

Note that "intelligence" here does not mean the machine understands or thinks the way you do. It means the machine produces a useful result on a task that used to require a person.

Machine learning: learning from examples

Machine learning (ML) is the method that powers most of today's AI, and it sits one box inside. In traditional software, a programmer writes exact rules and the computer follows them step by step. Machine learning flips that around: instead of being given the rules, the system is shown many examples and works out the patterns for itself.

A spam filter is the classic case. Nobody writes a rule for every possible junk email. Instead the system studies thousands of emails already labelled "spam" or "not spam" and learns the signals that tell them apart. Show it more examples and it gets better. The core idea is simple: it learns from data rather than from hand written instructions.

Deep learning: the innermost box

Deep learning is a particular, powerful style of machine learning, and it is the smallest box in the set. It uses structures loosely inspired by the brain, called neural networks, built in many stacked layers. That depth is where the name comes from.

Those layers let the system pick up very complex patterns, which is why deep learning drives the breakthroughs you hear about: understanding natural language, recognising faces in photos, generating images and text. It needs a great deal of data and computing power, but it can learn things simpler methods cannot.

Why the nesting matters

Keeping the boxes straight saves you from a lot of confusion. All deep learning is machine learning, and all machine learning is AI, but not the other way round. So when a product is described as "AI powered", the useful question is not whether it is impressive, but what it actually does: is it following fixed rules, learning from data, or using deep networks? The answer tells you far more than the label. And whichever box it sits in, the theme of the next lessons holds: these systems are tools that still need a human to check their answers.