What is machine learning?
Machine learning is a way of programming a computer with examples instead of rules. Try writing rules for “this is a cat” and you’ll never finish. Show a program thousands of labeled photos, though, and it can find the pattern itself, then use it to guess about photos it has never seen. It’s behind the spam filter in your email, the words above your phone’s keyboard and the feed you scroll.
Rules vs. examples
An ordinary program follows rules somebody wrote down. That works when the world is simple, like adding sales tax. It breaks when nobody can say the rule, like recognizing a face or a voice.
Machine learning flips it around. You collect examples, the program searches for the pattern that fits them best, and that pattern becomes the program.
Features and labels
The clues a model gets to look at are called features. The answer you want it to predict is the label. To guess whether a bus will run late, the features might be the time of day, the weather and the route, and the label is how many minutes late it was.
Picking good features takes knowing the subject, not just the math. And every feature has to exist at the moment you make the guess, or the model is cheating.
Three ways machines learn
Supervised learning
Learning from examples that come with answers. Most of what you meet is this kind: spam filters, the app that reads a zip code, price estimates.
Unsupervised learning
Finding patterns when no answers are given, like a news app sorting stories into piles nobody labeled.
Reinforcement learning
Learning from rewards by trying things, like a traffic light that learns which timing keeps cars moving.
How it goes wrong
A model is only as good as its examples. Train it on data that leaves some people out and it will get those people wrong more often. A model can also memorize its examples instead of learning the pattern, then fall apart on anything new.
That’s why the honest step is grading it on examples it has never seen before anyone trusts it.
Questions people ask
Is machine learning the same as AI?
Machine learning is the part of AI that learns from data, and it’s behind most of what people call AI today, chatbots included. AI is the broader goal of getting computers to do things that seem to take intelligence.
What is an example of machine learning?
Your email’s spam filter. It learned from messages people marked as spam or not spam, and it uses that pattern on new mail. The suggested words above your phone’s keyboard work the same way.
Do you need math to understand machine learning?
Not to understand what it does or how it goes wrong. The math matters when you build one. The idea itself, examples in, pattern out, graded on new cases, takes a few minutes to get.
Learn it by doing it
Module 14, “Teaching With Examples,” opens Part V of the course, and the five modules after it cover predicting numbers, sorting, finding groups, memorizing versus learning, and bias.
Data to AI is five-minute lessons on how AI works, starting with the data on your phone, for curious teens, parents and grandparents. No math needed. The first lesson runs right here in your browser.