AI From Zero · ChatGPT and AI Assistants

What Is ChatGPT?

Understand what ChatGPT is, how it uses language-model technology to generate responses, how conversation context works, and what its limitations are.

Estimated learning time: 15 minutes

What You'll Learn

By the end of this lesson, you will be able to:

- Explain what ChatGPT is.
- Explain the relationship between ChatGPT and a language model.
- Understand the basic process involved when you send a message to ChatGPT.
- Explain why conversation context matters.
- Identify common tasks that ChatGPT can help with.
- Understand why ChatGPT can sometimes produce incorrect information.
- Distinguish between ChatGPT and a traditional search engine.
- Use ChatGPT more effectively by providing clear instructions and context.

What Is ChatGPT?

ChatGPT is an AI assistant that allows people to interact with artificial intelligence using natural language.

Instead of learning a programming language or using a complicated technical interface, you can type a question or instruction in ordinary language and receive a response.

For example, you might ask:

"Explain machine learning to me as if I am a beginner."

ChatGPT can generate an explanation based on the request and the context available to it.

It can also help with many other tasks, including writing, summarizing, brainstorming, explaining concepts, analyzing information, generating ideas, and working with computer code.

ChatGPT Is an AI Assistant

It is useful to distinguish between an AI model and an AI assistant.

A language model is a machine learning model that processes language and generates text.

ChatGPT is an application and service built around AI models that allows people to interact with those capabilities through a conversational interface.

This means that ChatGPT is more than simply a box containing a language model.

An AI assistant can include several components working together, such as:

  • A user interface
  • An underlying AI model
  • Conversation and context management
  • Safety and policy systems
  • Tools and other capabilities

The exact capabilities available can depend on the version of the service, the user account, and the tools available in a particular environment.

Why Is It Called ChatGPT?

The name contains two useful ideas.

Chat refers to the conversational way people interact with the system.

GPT refers to a family of generative pretrained transformer-based language models.

You do not need to understand the technical meaning of every word in that description yet.

For now, the important idea is that ChatGPT uses language-model technology to process your instructions and generate responses.

How Does ChatGPT Generate a Response?

Let's look at a simplified example.

Suppose you type:

"What is artificial intelligence?"

A simplified view of what happens is:

  1. Your message is received by the ChatGPT service.
  2. The text is processed and represented in a form the AI system can work with.
  3. The relevant conversation context may be considered.
  4. The underlying model processes the input.
  5. The model generates a response sequentially.
  6. The resulting response is presented to you through the ChatGPT interface.

The actual systems involved are considerably more sophisticated than this simplified description.

However, the basic idea connects directly to what we learned in the previous module about language models and next-token prediction.

ChatGPT and Next-Token Prediction

In the previous lessons, we learned that language models can generate text by predicting what tokens are likely to come next.

ChatGPT uses language-model technology to generate responses in a similar fundamental way.

Imagine that the beginning of a response is:

"Artificial intelligence is"

The model evaluates possible continuations based on the input and context.

Once a token is generated, it becomes part of the growing sequence, and the system continues generating.

Repeated many times, this process can produce a complete answer, explanation, story, program, or other form of text.

Why Does ChatGPT Appear to Understand You?

When you communicate with ChatGPT, it can appear as though you are having a conversation with another person.

For example:

User: "I am learning about artificial intelligence."

User: "What should I learn next?"

The second question can be interpreted using the earlier message as context.

This ability to use conversational context makes an AI assistant much more useful than simply entering isolated questions.

However, conversational behavior should not be confused with human understanding.

The system generates responses using learned patterns and the information available to it. It does not necessarily understand information in exactly the same way that a human does.

What Can ChatGPT Do?

ChatGPT can assist with many different kinds of tasks.

Learning and Explanation

You can ask ChatGPT to explain difficult concepts in simpler language.

For example:

"Explain neural networks to me using a simple everyday example."

You can also ask follow-up questions and request a different explanation if the first one is difficult to understand.

Writing

ChatGPT can help draft many forms of written content.

For example, you might ask it to help create:

  • An email
  • An outline
  • A report
  • A summary
  • A presentation structure
  • A piece of creative writing

You should still review generated writing before using it, particularly when accuracy or professional judgment matters.

Brainstorming

AI assistants can also help generate ideas.

For example:

"Give me ten ideas for a small business that could use AI."

The value is not necessarily that every suggestion will be good. Instead, the assistant can help you explore possibilities that you can then evaluate yourself.

Programming

ChatGPT can help people understand and write computer code.

You might ask:

"Explain this PHP function."

or:

"Show me how to read a CSV file using Python."

Generated code should always be reviewed and tested rather than assumed to be correct.

Summarization

When appropriate information is provided to it, ChatGPT can summarize longer material.

For example, you might ask it to turn a long explanation into five key points.

A good summary request can also specify the intended audience or the level of detail required.

ChatGPT Is Not the Same as a Search Engine

It is important not to think of ChatGPT as simply another search engine.

A traditional search engine primarily helps you find information and webpages from an indexed collection of online resources.

A conversational AI assistant generates a response based on the model and the information available to it.

Depending on the tools and configuration available, an AI assistant may also be able to search for or retrieve information.

However, the underlying concepts are different.

A search engine may show you a list of sources.

An AI assistant may instead produce a synthesized response.

For important information, it is often useful to check reliable sources rather than accepting an AI-generated answer automatically.

Why Can ChatGPT Make Mistakes?

One of the most important things to understand about AI assistants is that they can produce incorrect information.

A response can sound confident and well written while still containing an error.

This can happen for several reasons.

  • The model may generate an incorrect prediction.
  • The available information may be incomplete.
  • The question may be ambiguous.
  • The model may misunderstand the user's intent.
  • The information may require verification from an external source.

This behavior is sometimes described as an AI system hallucinating.

The important lesson is simple:

A fluent answer is not automatically a correct answer.

How Should You Use ChatGPT?

The best way to use an AI assistant is not to treat it as an unquestionable authority.

Instead, think of it as a tool that can help you think, learn, create, analyze, and explore.

You remain responsible for evaluating the output.

For example, if you are using ChatGPT to learn a new subject, you can ask it to explain a concept and then verify important facts using reliable educational or primary sources.

Giving ChatGPT Better Instructions

The quality of an AI assistant's response can often be improved by providing a clear request.

Compare these two prompts:

"Tell me about AI."

and:

"Explain artificial intelligence to a complete beginner in 300 words. Use three everyday examples and avoid technical terminology."

The second request gives the assistant considerably more information about what you want.

You have specified:

  • The subject
  • The audience
  • The approximate length
  • The examples required
  • The desired level of technical detail

This is the beginning of prompt engineering, which we will explore in much greater detail later in this course.

ChatGPT and Human Judgment

AI assistants can be extremely useful, but they do not eliminate the need for human judgment.

For simple tasks, an AI-generated answer may be enough to get you started.

For important decisions, you should evaluate the information carefully.

This is particularly important for areas such as medicine, law, finance, security, scientific research, and other situations where an incorrect answer could have serious consequences.

AI can assist with these types of tasks in appropriate settings, but important decisions should not be based solely on an unverified AI response.

ChatGPT as a Learning Partner

One particularly useful way to think about ChatGPT is as an interactive learning partner.

Instead of asking only for an answer, you can ask the assistant to help you understand why an answer is correct.

For example:

"Teach me what a neural network is. First explain it simply, then ask me three questions to test my understanding."

You can then continue the conversation based on what you find difficult.

This makes the interaction more active than simply reading a definition.

The Big Picture

We can now connect ChatGPT to the concepts we have already learned:

Training Data → Language Model → Learned Parameters → Input + Context → Inference → Generated Tokens → Response

ChatGPT provides a user-friendly interface for interacting with AI capabilities built around language models and other supporting systems.

The conversational interface makes powerful AI technology accessible to people who do not need to understand all of the underlying mathematics or engineering.

What You Should Remember

  • ChatGPT is an AI assistant that allows people to interact with AI using natural language.
  • It uses language-model technology to generate responses.
  • Conversation context can help it handle follow-up questions.
  • ChatGPT can help with learning, writing, brainstorming, coding, summarization, and many other tasks.
  • ChatGPT is not simply a traditional search engine.
  • AI-generated answers can contain mistakes.
  • A confident or fluent response is not automatically correct.
  • Clear instructions and useful context can improve the quality of responses.
  • Human judgment and verification remain important.
  • Prompt engineering will allow us to explore how to communicate more effectively with AI systems.

What Comes Next?

We now know what ChatGPT is and how it connects to the language-model concepts we learned earlier.

But there is a more practical question:

What actually happens when you type a message into an AI assistant and receive a response?

In the next lesson, we will look more closely at how ChatGPT generates an answer and follow the journey from your message to the final response.

Key Takeaways

  • ChatGPT is an AI assistant that allows people to interact with AI using natural language.
  • It uses language-model technology to generate responses.
  • Conversation context helps it handle follow-up questions.
  • It can assist with learning, writing, brainstorming, coding and summarization.
  • AI-generated answers can contain mistakes and should be evaluated.
  • Clear instructions and useful context can improve responses.
  • Human judgment remains important when using AI.

Try It Yourself

Have your first structured conversation with an AI assistant.

Ask:

"Explain artificial intelligence to me as if I am completely new to the subject. Use one everyday example, keep the explanation under 200 words, and then give me three questions to test my understanding."

Read the response carefully.

Then ask a follow-up question about something you did not understand.

Finally, compare the result with this question:

"Tell me about AI."

Think about why the first instruction is likely to produce a more useful response.

Test Your Knowledge

You've reached the end of this lesson.

Test what you've learned with the Quiz: What Is ChatGPT?.

Take the Quiz
How ChatGPT Generates an Answer →
Back to Course