Phase 1: The Big Picture Lesson 2 of 4

AI, Machine Learning
and Deep Learning

These three terms get used interchangeably in the media, in job adverts and in conversations at dinner parties. They are not the same thing. Here is what each one actually means.

You will learn
The real difference between AI, ML and DL
Why the confusion exists in the first place
Real products that show each type in action
How to use these terms correctly from now on

Why does the confusion exist?

Pick up any newspaper article about technology and you will find the terms AI, machine learning and deep learning scattered around as if they all mean the same thing. A company releases a new product and calls it "AI-powered." A researcher publishes a paper using "deep learning." A business analyst writes a report about "machine learning strategy." Same thing, right?

Not quite. These three terms describe real and distinct concepts. They are related, but understanding the difference between them is one of the most useful things you can do before going any deeper into this subject.

The good news is that once you see the relationship clearly, it is almost impossible to forget.

The simplest way to think about it

Imagine three circles, one inside the other. Deep learning sits inside machine learning. Machine learning sits inside artificial intelligence. Every deep learning system is a machine learning system. Every machine learning system is a form of AI. But not every AI system uses machine learning, and not every machine learning system uses deep learning.

The three layers, clearly defined

The broadest category
Artificial Intelligence
Any technique that allows a machine to mimic or replicate human intelligence. This includes chess engines, voice assistants, recommendation systems and anything else that makes a computer appear to think or make decisions. AI does not have to learn from data. A chess program that follows handwritten rules is AI.
A subset of AI
Machine Learning
AI systems that learn from data rather than following fixed rules written by a programmer. Instead of telling the machine exactly what to do in every situation, you show it thousands of examples and let it figure out the patterns. The machine improves as it sees more data.
A subset of machine learning
Deep Learning
Machine learning that uses layered networks loosely inspired by the human brain. These networks can learn extremely complex patterns from raw data such as images, audio and text. Most of the AI breakthroughs you hear about today, including ChatGPT, image generators and voice recognition, are built on deep learning.

The critical difference: rules vs learning

The single most important thing that separates old-school AI from machine learning is how the system gets its instructions.

In traditional AI, a programmer sits down and writes out every rule the system needs. For a spam filter, they might write: "If the email contains the words 'click here to claim your prize', mark it as spam." This works until spammers change their language. Then you rewrite the rules. Then they change again. It is a constant arms race between human programmers and the problem they are trying to solve.

In machine learning, nobody writes the rules. Instead, you show the system thousands of emails that humans have already labelled as spam or not spam. The system looks at those examples, finds the patterns that separate the two categories, and builds its own rules automatically. When spammers change tactics, you just feed in new examples and the system updates itself.

Machine learning is the science of getting computers to act without being explicitly programmed.

Andrew Ng, AI researcher and educator

So what makes deep learning different?

Standard machine learning works brilliantly when you can describe your data in a table. Rows of numbers, columns of features. Predicting house prices from square footage and number of bedrooms, for example, is a machine learning problem that works very well.

But what about a photograph? A photograph is not a table of numbers with clear features. It is millions of pixels, and the meaning of any single pixel depends entirely on what is around it. The same shade of grey means something completely different in a face versus in a sky versus in a car bumper.

Deep learning solves this by stacking multiple layers of processing on top of each other. The first layer might learn to spot edges. The next layer combines edges into shapes. The next layer combines shapes into objects. By the end, the system can recognise a cat in a photograph without a single human ever writing a rule about what a cat looks like.

Why the word "deep"?

The "deep" in deep learning refers to the depth of the network, meaning the number of layers stacked on top of each other. Early neural networks had one or two layers. Modern ones have dozens or even hundreds. The depth is what gives them their extraordinary capability.

Seeing it in the real world

The clearest way to cement this understanding is to look at products you already use and place them correctly into one of the three categories.

Artificial Intelligence
Google Maps route planning
Finds the fastest route using algorithms and real-time traffic rules. Intelligent behaviour, but largely rule-based. This is AI in the broadest sense.
Artificial Intelligence
Chess or Go engines
Uses search algorithms and evaluation functions to play at superhuman level. Older chess engines follow explicit rules. This is AI without machine learning.
Machine Learning
Spotify recommendations
Learns your listening patterns and compares them to millions of other users to predict what you will enjoy next. Classic ML using your behaviour as data.
Machine Learning
Credit card fraud detection
Trained on millions of past transactions labelled as fraudulent or legitimate. Learns the patterns that indicate something suspicious without being told what those patterns are.
Deep Learning
Face ID on your phone
Uses a deep neural network trained on facial images to recognise your face in varying light, angles and expressions. No human wrote rules about what your face looks like.
Deep Learning
ChatGPT and Claude
Large language models built on transformer-based deep learning networks trained on enormous amounts of text. The most visible example of deep learning in everyday life today.

A quick comparison side by side

Aspect AI Machine Learning Deep Learning
Learns from data? Not necessarily Yes, always Yes, always
Needs labelled examples? Sometimes Usually yes Often yes, sometimes no
Works well on images and audio? Rarely on its own Sometimes, with effort Exceptionally well
How explainable is it? Often very clear Reasonably clear Often hard to explain
How much data does it need? Varies Moderate amounts Large amounts

The one thing to take away

If someone says "we use AI in our product," that tells you almost nothing. It is like saying "we use technology." AI is the umbrella, not the description.

The more interesting questions are always: does it learn from data, or does it follow fixed rules? And if it learns, does it use deep learning or a simpler approach? Those two questions will tell you far more about what a system can and cannot do.

From this point on in the course, you will be able to hear those terms in any context and place them accurately. That alone puts you ahead of the majority of people who talk about AI confidently but have never stopped to separate the layers.

Lesson Activity · No tools needed
Sort the Products
Now that you know the three layers, let us test it with things you already use. This is not a trick exercise. There are no wrong answers as long as you can explain your reasoning.
01 Take these five products: Netflix recommendations, Google Translate, a traditional calculator, Siri or Alexa, and your email spam filter.
02 For each one, decide: is it AI only, machine learning, or deep learning? Write your reasoning in one sentence per product.
03 Bring your answers to the live session. We will go through each one together, and you will likely discover that some of these products use all three layers at once.
Your Notes
Studying independently? Write your thoughts or answers below. Notes save automatically to your browser.
Pause & Reflect

Check your understanding

Click any question to reveal a thinking prompt. There are no wrong answers.

Can you explain AI, Machine Learning, and Deep Learning to a 10-year-old using one everyday example for each?

Try analogies: AI is the broad idea of making machines smart. ML is teaching a machine by example, like training a dog. Deep Learning is a method inside ML that mirrors how the brain works in layers. A voice assistant, a spam filter, and face unlock each map to one of these levels.

Is all Deep Learning also Machine Learning? Is all Machine Learning also AI? Draw the relationship in your head.

Think of three concentric circles: AI is the largest. Inside it is Machine Learning. Inside ML is Deep Learning. Every deep learning system is an ML system, which is an AI system. But not every AI system uses ML, and not every ML system uses deep learning.

Your phone's autocorrect: would you call it AI, ML, or Deep Learning? What clues help you decide?

Traditional autocorrect used rule-based AI (dictionaries, grammar rules). Modern predictive keyboards use ML trained on typing patterns. The latest models use deep learning to predict whole phrases from context. The answer depends on which generation of the feature you are thinking about.

Progress
Done with this lesson?
Mark it complete to track your progress.