Machine learning is a way of building AI systems that learn patterns from examples instead of being explicitly programmed with every rule.
Rather than telling a computer exactly how to recognise every spam email, predict every price or identify every cat, we can show it lots of examples and let it learn useful patterns.
The more accurate version is a little more nuanced than “the machine learns like a human” — but the core idea is surprisingly easy to understand.
Let's break it down without drowning in equations.
Traditional programming vs machine learning
Traditional software usually follows instructions written explicitly by humans.
💻 Traditional programming
Rules + Data → Answer
A programmer defines the logic and the computer follows it.
🧠 Machine learning
Data + Examples → Learned model
The system finds useful patterns that can later help it make predictions or decisions.
Writing a rule that identifies every possible spam email would be extremely difficult. But showing a system many examples of spam and non-spam messages can help it learn patterns associated with each.
So how does a machine “learn”?
Not by thinking, understanding or experiencing the world like a human.
In machine learning, “learning” usually means adjusting a mathematical model so that it becomes better at finding patterns or making predictions based on data.
📚 Give the system data
Examples, measurements, images, text or other relevant information.
🔍 Find patterns
The training process looks for relationships within that data.
🔧 Adjust the model
The system improves its ability to produce useful predictions.
🎯 Test on new data
We check whether what it learned works beyond the examples it trained on.
A simple real-world example: predicting house prices
Imagine you have information about thousands of houses:
📍 Location
Which suburb or area is the house in?
📐 Size
How large is the property?
🛏️ Features
How many bedrooms, bathrooms and other characteristics?
💰 Sale price
What did similar houses actually sell for?
A machine-learning model can analyse these examples and learn relationships between the features and historical sale prices.
Later, when shown a new house, it can use the patterns it learned to estimate a possible price.
It is not memorising one simple rule.
The model can combine many patterns and relationships. That is what makes machine learning useful for problems where writing every rule by hand would be difficult.
Why data matters so much
A machine-learning model learns from the information it is given. That means poor, limited or biased data can produce poor or biased results.
📉 Bad data in
If the examples are inaccurate or incomplete, the model may learn misleading patterns.
⚖️ Biased examples
If some groups or situations are poorly represented, the results may not work equally well for everyone.
📏 Wrong signal
Data can contain patterns that look useful during training but fail in the real world.
🕒 Changing world
A model trained on old patterns may become less useful when reality changes.
The three big ways machines learn
There are several approaches to machine learning, but beginners can start with three broad ideas.
Learning from labelled examples
You provide examples along with the correct answer. For instance: thousands of emails labelled “spam” or “not spam”.
Finding patterns without being told every answer
The system looks for groups, structures or relationships in data without a predefined label for every example.
Learning through feedback
A system receives feedback based on its actions and gradually learns which choices lead to better outcomes.
What is “training”?
Training is the process of adjusting a machine-learning model using data and feedback.
A model makes an attempt, its output is compared with a target or evaluated using a training objective, and the model's internal parameters are adjusted. This process can repeat many times.
The goal is to build a model that can perform well not only on the examples it has already seen, but also on new examples.
The big challenge: learning patterns without memorising everything
Imagine a student who memorises the answers to last year's exam paper but cannot solve a new question.
Machine-learning models can face a similar problem called overfitting: they may perform extremely well on training data but poorly on new, unfamiliar data.
❌ Overfitting
The model learns the training examples too specifically and struggles to generalise.
✅ Generalisation
The model learns useful patterns that also work reasonably well on new examples.
Where do you encounter machine learning every day?
📺 Recommendations
Streaming and shopping services can use patterns to suggest things you may like.
📧 Spam filters
Email systems can learn patterns associated with unwanted messages.
📷 Image recognition
Apps can identify objects, faces or visual patterns.
🗣️ Language tools
Translation, speech recognition and modern AI assistants all rely heavily on machine-learning techniques.
AI vs machine learning: what's the difference?
Artificial Intelligence is the bigger umbrella. Machine learning is one important way of building AI systems.
Not every AI technique is machine learning. But machine learning has become one of the most important approaches behind modern AI.
And where does deep learning fit in?
Deep learning is a specialised area within machine learning that uses multi-layer neural networks.
You can picture the relationship like this:
Artificial Intelligence
↳ Machine Learning
↳ Deep Learning
We will explore deep learning separately because it deserves its own plain-English explanation.
Machine learning: Myth vs reality
“Machine learning means computers think like humans.”
Machine learning usually means finding patterns in data and improving performance on a defined task.
“More data automatically means a better model.”
Quality, relevance and representation matter alongside quantity.
“Once trained, a model is permanently correct.”
Models can make mistakes and may become less effective as real-world conditions change.
🔥 Key takeaways
- Machine learning helps computers learn useful patterns from data rather than relying only on explicitly written rules.
- Training adjusts a model so it can perform a particular task more effectively.
- Data quality and representation have a major impact on the quality of the results.
- Supervised, unsupervised and reinforcement learning are three useful beginner-level categories.
- A good model should generalise to new data rather than simply memorising its training examples.
- Machine learning is a major part of modern AI, and deep learning is one specialised branch within it.
Next up: What Is Deep Learning?
Now that we understand how machines learn patterns, let's go one layer deeper into the technology behind many of today's biggest AI breakthroughs.
Explore Evergreen AI →