AI vs Machine Learning vs Deep Learning vs ChatGPT

Artificial intelligence, machine learning, deep learning, and ChatGPT are related, but they describe different things.

In Chapter 1, you trained a model to predict whether a student would pass or fail using study hours and sleep hours. That model demonstrated machine learning—but it could not answer questions like ChatGPT.

This chapter explains why.

By the end, you will be able to:

• Explain the relationship between AI, machine learning, and deep learning.
• Distinguish a trained model from an application or product.
• Explain why more data does not automatically give a model a different capability.
• Draw a simple AI relationship map.
• Describe your first model using the correct terms.

No installation or coding is required for this chapter. You only need a notebook and a pen for the practical activity.

AI vs Machine Learning vs Deep Learning vs ChatGPT

The relationship is:

Artificial intelligence
└── Machine learning
    └── Deep learning

ChatGPT: a product using deep-learning models
and supporting software

AI, ML, and deep learning are related technical categories. ChatGPT is a product name.

What Is Artificial Intelligence?

Artificial intelligence is the broad field of building systems for tasks associated with intelligence.

These tasks include:

• Planning a sequence of actions.
• Reasoning using stored knowledge.
• Learning patterns from examples.
• Recognising images.
• Working with language.

AI includes different approaches. Some learn from data; others use knowledge, written rules, or search.

For example, an expert system could compare observations about a crop against guidance supplied by agricultural experts. It could use those stored rules without learning them from example records.

Automatic operation alone does not establish AI. A kettle switching off at a fixed temperature is ordinary automation. To understand a system, ask what it does and how it makes its decisions.

What Is Machine Learning?

Machine learning is an approach within AI in which a learning procedure finds useful patterns in data. The resulting model uses what it learned when given new inputs.

Imagine collecting records of Siddu ingredients, cooking times, and customer feedback. A learning procedure could use those examples to find patterns related to customer preferences.

Different ML models can:

• Predict categories, such as pass or fail.
• Estimate numbers, such as expected hotel bookings.
• Generate content, such as a written reply.

In Chapter 1, the decision-tree training procedure used study hours, sleep hours, and recorded outcomes to learn decision rules.

You chose the method and supplied examples. You did not manually write every final decision the trained tree would make.

What Is Deep Learning?

Deep learning is a kind of machine learning that uses neural networks with many layers.

A neural network contains connected mathematical units arranged in layers. Each layer combines numerical inputs and passes results to another layer. Training adjusts internal numerical values using examples.

For handwritten-digit recognition, training examples might include an image of a handwritten 7 paired with the correct label, seven. The network learns calculations that help predict which digit an image contains.

The layers perform numerical calculations. They are not little human brains or people inside a computer.

The decision tree from Chapter 1 was ML within AI, but not deep learning, because it did not use a neural network.

Where Does ChatGPT Fit?

ChatGPT is a conversational AI product. Deep-learning models provide its language capabilities, while its interface and supporting software help users access those capabilities.

This distinction matters:

• A model provides a particular capability.
• A product brings capabilities and supporting software together for users.

LLM stands for large language model. These models are trained to work with patterns across large amounts of text. We explore LLMs more fully in Chapter 5.

A product name alone does not tell you every model or method used inside it.

Remember It with the Solan Kitchen Analogy

Imagine a busy Solan kitchen:

The analogy helps you remember the relationships. It does not mean AI systems literally work like kitchens.

Whenever you hear an AI-related name, ask:

Am I naming a field, a learning approach, a particular model, or a product?

These terms describe different responsibilities.

1. Training: The Learning Process

In Chapter 1, a training procedure fitted a decision tree using student examples.

Training is the process of learning from examples.

For this lab, the process was:

Student records + decision-tree training procedure
                        ↓
                Trained decision tree

The tiny dataset demonstrated how training works. It did not establish that the model could make reliable predictions about real students.

2. Trained Model: The Result You Can Use

The trained model is the result of training. You can save it, load it later, and give it suitable inputs.

In the Chapter 1 lab, predict.py loaded the saved model and used it like this:

Study hours + sleep hours
            ↓
    Trained decision tree
            ↓
    Pass-or-fail prediction

A model that predicts a category is called a classifier. Here, the categories were pass and fail.

This classifier was fitted to predict a label from two numerical inputs. It was not fitted to accept an English question and generate an explanation.

Why Won’t More Records Make It Talk Like ChatGPT?

More student records might help you investigate or improve the same prediction task, depending on their quality and relevance.

They do not automatically change the classifier into a language-generating system.

The model’s inputs, outputs, and learning task matter. Generating language requires a model suited to that different task.

3. Application: The Software Around the Model

An application makes a capability usable.

For the student predictor, an application might:

1. Collect study hours and sleep hours through a form.
2. Pass those numbers to the trained model.
3. Display the predicted category.

The model predicts. The application collects inputs, connects components, and presents the result.

Adding a chat screen to the student predictor would not give it ChatGPT’s language capabilities.

Must You Train a Model to Build an AI Application?

No. You can build an application using an existing trained model.

You still need to check whether that model suits your task, accepts the required inputs, and produces useful results.

Also, not every AI approach involves data-driven training. AI includes approaches based on stored knowledge, rules, and search.

How do models, interfaces, and supporting software come together in an application?

Consider Subra Space AI, a product from the creator of this course. It provides a practical example of the distinction between an AI model and the product built around it.

Think of its parts separately:

The product name does not identify every model or method inside it. Similarly, downloading or using an AI product does not mean you trained its models.

Try This Observation Activity

Visit subraspaceai.com to explore its features and download options. The website currently advertises a free Personal Space, alongside paid options for additional functionality.

As you explore, ask:

1. What task does this feature help the user complete?
2. What information does the user provide?
3. What result does the application display?
4. Which parts appear to involve AI capabilities?
5. Which parts handle ordinary application responsibilities, such as navigation, storage, or displaying results?

You do not need to install the app to complete this activity; you can begin with the demonstrations on its website.

The learning goal is to distinguish the complete product from the models and software that make it work.

“AI, ML, and Deep Learning Are Competing Choices”

They are nested categories. Deep learning belongs within ML, and ML belongs within AI.

Studying deep learning means studying one part of machine learning.

“AI Must Be Able to Chat”

An AI system can perform a useful task without holding a conversation.

For example, a model could estimate how many rooms a Solan hotel will book next weekend. Its useful output could simply be a number.

“An Image Task Must Use Deep Learning”

Deep learning is one possible approach to recognising handwritten digits. Seeing a system recognise handwriting does not prove which method it uses.

The task and the method used to perform it are different things.

“Building an AI App Means Training a Model”

An application can use an existing trained model. Building the application and training the model are separate responsibilities.

“Deep Learning Is Always the Best Choice”

Choose a method because it fits the problem.

Compare suitable methods using the task, available data, computing resources, and measured results. Deep learning is not automatically the best option.

“A Product and Its Model Are the Same Thing”

A product combines several parts.

In a learning assistant, for example, learning materials provide content, a model provides capabilities, and the interface allows interaction. The product brings those parts together.

You need: a notebook and a pen. No installation or account is required.

Step 1: Draw Three Nested Circles

Draw:

1. A large circle labelled AI.
2. A smaller circle inside it labelled Machine Learning.
3. Another circle inside ML labelled Deep Learning.

The circles show inclusion:

• Every deep-learning method belongs within ML.
• Every ML method belongs within AI.

Step 2: Place the Student Decision Tree

Inside the ML circle, but outside the deep-learning circle, write:

“Our trained student decision tree”

It belongs there because it learned from data but did not use a neural network.

The label names a particular trained model. Its position shows the technical categories it belongs to.

Step 3: Add ChatGPT

Beside the circles, write:

“ChatGPT: conversational product”

Draw an arrow from Deep Learning to ChatGPT. Label it:

“provides models used by”

Placing ChatGPT beside the circles distinguishes a product from the technical categories. The arrow shows how the technology supports the product.

Step 4: Check Your Understanding

Complete these sentences before reading the answers:

1. AI is the broad ______.
2. ML learns patterns from ______.
3. Deep learning uses neural networks with many ______.
4. Our student decision tree was ML, but not ______.
5. ChatGPT is a ______ using AI models.

Stop here and write all five answers first.

Answers

1. AI is the broad field.
2. ML learns patterns from data.
3. Deep learning uses neural networks with many layers.
4. Our student decision tree was ML, but not deep learning.
5. ChatGPT is a product using AI models.

Now explain why the decision tree sits outside deep learning and why ChatGPT has an arrow rather than another nested circle.

Mini Project: Explain Your First Model

Using your map, write four short sentences explaining:

1. What inputs the student model received.
2. What it learned to predict.
3. Why it belongs within ML and AI.
4. Why adding records would not turn it into ChatGPT.

Try writing your explanation before reading this example:

Our model received study hours and sleep hours. It learned from examples to predict a pass-or-fail label. Learning from data makes it ML, which belongs within AI. More records do not change its task into generating language.

Explain it to someone who has not seen the lab. Use your map, but try not to read your sentences.

• AI is the broad field and includes several approaches.
• Machine learning learns patterns from data and belongs within AI.
• Deep learning uses neural networks with many layers and belongs within ML.
• ChatGPT is a conversational product using AI models and supporting software.
• Training is a learning process; a trained model is its result.
• An application connects capabilities with inputs, interfaces, and other software.
• More data alone does not give a model a different kind of capability.
• Building an AI application does not necessarily mean training a model yourself.

Final Challenge

A friend says:

“I built an AI app, so I trained a deep-learning model.”

Does that conclusion follow?

Think about your answer before continuing.

It does not. The application might use an existing trained model. Even if your friend trained a model, it might use an ML method other than deep learning.

Continue Learning

Previous — Chapter 1: Revisit the student predictor and identify its inputs, training examples, and predicted categories.

Next — Chapter 3: ChatGPT can produce a convincing explanation—but does that mean it is always correct? Explore why it can make up information and why checking its answers matters.