Ten bite-size lessons for total beginners. Read for a minute, answer 3 quick questions, earn a badge. Finish all ten for your free certificate. No sign-up, and your progress stays on this device.
Artificial intelligence (AI) is software that does tasks we usually connect with human thinking: understanding language, recognising images, making predictions or creating text and pictures.
Most modern AI isn't programmed rule by rule. Instead it learns patterns from lots of examples. Show a system thousands of photos labelled "cat" or "dog" and it learns to tell them apart. This approach is called machine learning.
You already use AI every day: spam filters, map apps predicting traffic, phone cameras improving photos, and streaming apps suggesting what to watch next.
AI is powerful but it doesn't "understand" the world the way people do. It works out the most likely answer from patterns it has seen, which is why it can be brilliant at one moment and confidently wrong the next.
An AI model is the "brain file" that comes out of training. It's a huge set of numbers (called parameters) that capture the patterns found in the training data.
Training is the expensive, slow part: the model makes guesses on example data, gets told how wrong it was, and nudges its numbers to do better, millions of times over.
Once trained, the model is used for inference: you give it new input (a question, a photo) and it produces an output in seconds.
Large language models (LLMs) like the ones behind chatbots have billions of parameters and were trained on large amounts of text. A model only knows what was in its training data up to a cut-off date, unless the app also lets it search or read new information.
A prompt is the instruction or question you give an AI. The clearer the prompt, the better the answer, just like briefing a new helper.
Five building blocks make prompts stronger: Role ("You are a friendly teacher"), Context (who it's for and why), Constraints (length, tone, what to avoid), Format (list, table, email) and Examples (show what you like).
Compare "write a caption" with "You're a social media helper for my lemonade stand. Write 3 short, upbeat Instagram captions for families. No hashtags." The second gives the AI far more to work with.
Prompting is a conversation. If the first answer misses, say what to change: "shorter", "more playful", "use simpler words". Try our Prompt Coach to practise.
Text models (large language models) read and write words. They predict the next piece of text, one small chunk at a time, which lets them answer questions, summarise, translate and write code.
Image models create pictures from a text description. Many use a technique called diffusion: they start from random noise and step by step turn it into an image that matches the prompt.
Some models are multimodal: they can take in text and images (and sometimes audio) together, for example describing a photo or answering a question about a chart.
Prompting differs too. Text prompts benefit from role, context and format. Image prompts work best with concrete visual details: subject, style, lighting, colours and composition.
A hallucination is when an AI states something false as if it were true: a made-up fact, a fake quote, a book that doesn't exist or a wrong date.
It happens because text models predict likely-sounding words. They aren't looking facts up in a verified database (unless the app connects them to search or documents), so a fluent answer can still be wrong.
Hallucinations are more likely for niche topics, very recent events, exact numbers, citations and links.
How to protect yourself: check important facts against a trusted source, ask the AI for its sources and then open them, and tell it "say you don't know if you aren't sure". For health, legal or money decisions, talk to a qualified person.
AI learns from data made by people, so it can pick up the same gaps and unfair patterns that exist in that data. This is called bias.
Example: if an image model mostly saw photos of one kind of person labelled "doctor", it may keep drawing doctors that way. If a hiring tool learned from past decisions that weren't fair, it could repeat them.
Bias can come from who is in the data, who is missing, how the data was labelled, and how the system is tested and used.
What helps: diverse and well-checked data, testing results across different groups, being open about limits, and keeping a human in charge of important decisions. As a user, notice when outputs seem one-sided and ask for other perspectives.
Treat a chatbot like a helpful stranger: friendly and useful, but don't hand over secrets. Never paste passwords, bank or card numbers, ID numbers, or private details about other people.
Many AI services may store your chats and, depending on settings, use them to improve their models. Check the privacy settings and turn off chat history or training if you prefer.
At work, follow your company's rules. Don't paste confidential documents or customer data into tools your employer hasn't approved.
Watch for AI-powered scams: cloned voices on the phone, fake videos and very polished phishing emails. If a message is urgent and asks for money or codes, stop and verify through a channel you already trust.
For a small business, AI is like a fast first-draft assistant. Popular uses: writing product descriptions, social posts and emails; summarising reviews or meeting notes; brainstorming names, offers and ideas.
It also helps with customer service (drafting replies to common questions), simple images for posts, and organising spreadsheets or explaining formulas.
Start small: pick one repetitive task that eats your time, try a free tool for a week, and measure whether it actually saves time.
Keep a human check. Review everything before it goes out, especially prices, policies and claims. Don't paste customer data into unapproved tools, and be honest with customers when they're chatting with a bot.
AI images are getting very realistic, so no single trick is foolproof. Look at the details: hands and fingers, teeth, jewellery, text and signs (often garbled), patterns that don't repeat properly, and reflections or shadows that don't match.
Check the background: warped lines, objects melting into each other, or surfaces that are too smooth and glossy.
Think about the source: who posted it first? Does a trusted news outlet show the same scene? A reverse image search can reveal where a picture came from.
Some tools add labels or hidden watermarks and content credentials (such as C2PA) to AI images, but these aren't on every image and can be removed. When it matters, verify before you share. Test yourself with our Is it AI? quiz.
You don't need a PhD to work with AI. There are roles for many skill sets: machine learning engineers and data scientists build and test models; data analysts turn data into decisions; AI product managers decide what to build and why.
There are also roles in AI safety and ethics, data labelling and quality, UX and conversation design, technical writing, and training others to use AI.
And almost every job is becoming an "AI-assisted" job. Marketers, teachers, nurses, designers and shop owners who can use AI tools well are increasingly valued.
How to start: learn the basics (free courses exist), practise with real tasks, build small projects you can show, and learn a little Python and data skills if you're curious about the technical side. Stay curious and keep learning; the field changes fast.