How machines learn from data

Machine learning is a type of artificial intelligence where a computer program improves its performance by studying examples — called training data. Instead of a human writing every rule, the program figures out the rules itself by looking at thousands of examples and spotting patterns. The more examples it studies, the better it gets at making correct predictions.

How Machine Learning Works — Step by Step

1. Collect training data — gather thousands of labeled examples (e.g. photos labeled "cat" or "not cat").
2. Feed data to the model — the program studies the examples and looks for shared features.
3. Find patterns — it notes what cats look like: pointed ears, fur, whiskers, and so on.
4. Test the model — show it new photos it has never seen and check if it labels them correctly.
5. Improve — correct wrong answers and re-train the model so it does better next time.

Examples of Machine Learning

• An email app that learns which messages are spam after seeing many examples.
• A photo app that learns to tell dogs from cats by studying labeled photos.
• A streaming service that learns your taste in films by tracking what you watch and rate.
• A speech recogniser that improves by hearing thousands of different voices.
• A self-driving car that learns safe driving by analyzing millions of kilometers of road data.

What is NOT Machine Learning

• A calculator that always uses the same math rules — it does not learn from examples.
• A website that shows the same homepage to every visitor regardless of their behavior.
• A traffic light that changes on a fixed timer — no data, no learning.
• A spell-checker that only uses a fixed dictionary — no pattern learning from usage.