Published: October 26, 2026
Last updated: October 26, 2026
The basic idea
What is generative AI?
Generative artificial intelligence refers to AI systems designed to produce new content in response to input. Depending on the system, that content can include text, images, audio, video, computer code or combinations of different media.
These systems learn patterns and relationships from training data. They then use what was learned to generate outputs that resemble patterns found in that data without simply following a fixed list of manually written responses.
Generative AI does not guarantee that the content it creates is true, original, appropriate or useful. The output still needs to be evaluated in the context in which it will be used.
Different outputs
What can generative AI create?
Generative AI is not limited to chatbots. Different models are designed to work with different kinds of information.
Text
Generative AI can draft, rewrite, summarise, translate and answer questions using natural language.
Images
Some models can generate or edit images from written instructions, reference images or combinations of both.
Audio
Generative systems can create speech, music, sound effects and other forms of synthetic audio.
Video
Some AI systems can generate or transform video, although quality, consistency and capabilities vary between models.
Software code
AI assistants can generate, explain and modify computer code, but generated code still needs testing and review.
Multimodal content
Some modern AI systems can work with several types of information, such as text, images, audio and video.
Simple example
Think of a language model as a powerful pattern predictor
Imagine beginning a sentence with: “The sun rises in the...” A language model can analyse the text and estimate which pieces of language are likely to come next based on patterns learned during training.
Modern large language models perform this kind of prediction on a much more sophisticated scale. They repeatedly predict and generate tokens to build responses that can include paragraphs, explanations, conversations, summaries and computer code.
This ability can produce remarkably useful language, but predicting plausible text is not the same as independently verifying that every statement is factual.
Large language models
What is an LLM?
LLM stands for large language model. It is a type of AI model trained to learn statistical patterns and relationships in language from large amounts of data.
Instead of processing language only as complete words, many language models divide input into smaller units called tokens. A token might represent a whole word, part of a word, punctuation or another piece of text depending on the model.
When generating a response, the model processes the available context and repeatedly predicts tokens that can follow. This process allows the system to generate long and apparently coherent responses.
Input
You provide a prompt, question, document, image or other supported information to the AI system.
Processing
The model uses its learned parameters and available context to calculate likely patterns for the response.
Generation
Tokens are generated in sequence to form the response you see, such as text, code or another supported output.
Prompts
What is a prompt?
A prompt is the information or instruction you provide to a generative AI system. It can be a simple question, a detailed set of instructions, a document to analyse or another supported type of input.
Good prompting is mainly about communicating your goal clearly. More words do not automatically create a better prompt. Useful context, clear requirements and a well-defined task are usually more important.
State the task clearly
Tell the AI what you want it to do instead of relying on it to guess your goal.
Provide useful context
Relevant background information can help the system produce a response that better fits your situation.
Specify the format
You can request a table, explanation, checklist, email, summary or another useful structure.
Add important constraints
Mention requirements such as length, audience, tone or information that should be included or excluded.
Review and refine
If the first response is not useful enough, clarify your request, correct misunderstandings and ask follow-up questions.
A major limitation
Why can an AI give a confident but incorrect answer?
A language model is designed to generate useful and plausible language, but that does not mean every generated statement has been independently checked against a reliable source.
A model can sometimes produce information that sounds convincing but is inaccurate, unsupported or fabricated. This is one reason important claims should be verified using trustworthy sources.
Verification becomes especially important when information could affect decisions involving money, health, safety, law, security or other serious consequences.
Important
A well-written AI answer is not automatically a verified answer. Check important information before relying on it.
Privacy & security
Be careful what you put into an AI system
Before entering information into an AI service, consider whether that information is private, confidential, commercially sensitive or protected by an agreement or policy.
Avoid casually entering passwords, authentication codes, banking credentials, confidential customer information, private business documents or other sensitive data into an AI service.
Organisations should also follow their own security, privacy and data-handling policies when deciding which AI tools may be used for work.
Key takeaways
What you should remember from Lesson 02
Generative AI creates new outputs based on patterns learned during training.
Large language models are designed to work with language and generate text by predicting sequences of tokens.
A prompt is the input or instruction you give an AI system.
Clear context and instructions can improve an AI response, but they cannot guarantee accuracy.
A fluent AI response can still contain incorrect or fabricated information.
Private, confidential or sensitive information should be handled carefully when using AI services.
Continue learning
Next: Using AI Safely & Responsibly
In Lesson 03, we will look more closely at privacy, verification, responsible AI use, sensitive information and situations where human expertise remains especially important.
Return to AI learning path