Machine learning sounds complicated — but the idea behind it is actually very simple. Here's how to understand it, with examples from apps your kids use every day.

What is machine learning?

Machine learning is a type of AI where a computer learns from examples instead of being given exact instructions. In normal programming, a human writes rules: "if this happens, do that." In machine learning, instead of writing rules, you feed the computer thousands (or millions) of examples and let it figure out the patterns itself.

Think of it like teaching a dog to sit. You don't explain the physics of sitting — you just show the dog what "sit" looks like hundreds of times with treats as rewards, until it gets it. Machine learning works the same way: show the computer lots of examples, reward the right answers, and it learns.

Real examples kids already use

YouTube recommendations

When YouTube shows you a video in the "Up Next" sidebar, that's machine learning. YouTube's model has studied what millions of people watched before and after similar videos, and learned which videos go together. It predicts what you specifically will enjoy based on your history.

Netflix and Disney+

Same idea. Those personalised rows — "Because you watched..." — are a machine learning model that learned your taste from everything you've played, paused, and skipped.

Spam filters in email

Your email app has a machine learning model trained on millions of spam and non-spam emails. It learned what spam looks like — dodgy links, certain phrases, strange senders — and now automatically spots new spam it's never seen before.

Face unlock on your phone

When you set up face unlock, your phone takes lots of scans of your face and trains a small machine learning model to recognise you — even in different lighting or at different angles.

Machine learning vs regular programming

Regular ProgrammingMachine Learning
Humans write every ruleComputer finds rules from examples
Works for problems with clear rulesWorks for fuzzy problems (faces, language, taste)
Output is predictableOutput can be unpredictable
Doesn't improve with useCan improve as it sees more data

How does a machine learning model actually learn?

The process has three steps:

  1. Training data: You feed the model thousands of labelled examples. For an image classifier, that might be 100,000 photos labelled "cat" or "not cat."
  2. Training: The model makes guesses, checks if it's right, adjusts itself, and tries again — millions of times — until its guesses are accurate.
  3. Testing: You give the model new examples it's never seen to check how well it learned.

Why does this matter for kids?

Machine learning is already shaping what your children see online, what news they read, what music they hear, and what ads they're shown. Understanding how it works gives kids the power to notice when they're being influenced — and to question it. In 10 years, machine learning skills will be as sought-after as coding is today.

The best way for kids to actually understand machine learning? Build something with it. The AI Adventures course has hands-on projects where kids train their own simple AI models.