What AI actually is (and isn't)
Understand in plain words what today's AI is, what it can do for you, and what it can't.
AI, short for artificial intelligence, is software that has learned from a huge number of examples instead of being given step-by-step rules.
The kind most people mean today is the AI assistant: a chat box where you type (or say) what you want in normal words, and it writes back. ChatGPT, Claude, Gemini and Microsoft Copilot are all examples.
Think of it like a very well-read helper. It has read more than any person ever could, so it can explain things, draft emails, summarise long documents, suggest ideas and answer questions.
But it isn't a person, it doesn't "know" things the way you do, and it can make mistakes while sounding confident. It also doesn't automatically know about your life, your work or today's news unless you tell it or it looks it up.
The simple rule: AI is great for a first draft, an explanation or a fresh idea. You stay in charge of checking it and deciding what to use.
A simple definition
Artificial intelligence (AI) is a broad name for computer systems that do things we'd normally expect to need human thinking, like understanding language, recognising a face in a photo or spotting a pattern in numbers.
Older software followed rules written by programmers: "if this, then that". Modern AI is different. It learns patterns from examples. Show it millions of photos labelled "cat" and it learns what cats tend to look like. Show it a huge library of text and it learns how language works.
The AI most people use today
When people talk about AI in 2026 they usually mean an AI assistant (sometimes called a ). You type or speak a request in everyday language and it replies, also in everyday language. Popular examples include ChatGPT (from OpenAI), Claude (from Anthropic), Gemini (from Google) and Microsoft Copilot. Most have a free version.
Under the hood these assistants run on a large language model: the "engine" that has learned from vast amounts of writing. You'll learn more about that in the next lesson.
What it's genuinely good at
- Explaining things at whatever level you ask for ("explain my energy bill like I'm 12").
- Writing and rewriting: emails, letters, CVs, social posts, a polite way to say no.
- Summarising long documents, articles or meeting notes.
- Brainstorming: gift ideas, names, a plan for a party or a project.
- Organising: turning messy notes into a list, a timetable or a table.
What it isn't
- It isn't always right. It predicts a likely answer, which is usually good but sometimes confidently wrong. Lesson 5 shows you how to spot this.
- It isn't a person. It has no feelings, opinions of its own or memory of you unless the app is set up to remember.
- It isn't automatically up to date. Some assistants can search the web; others only know what they learned up to a certain date.
- It isn't a replacement for your judgement, especially for health, money or legal decisions.
Why it matters to you
You don't need to understand the technology to benefit. Like a calculator or a sat nav, the skill is knowing what to ask and when to double-check. That's exactly what this Learn section teaches.
Where the term comes from
"Artificial intelligence" was coined in 1956 for a summer research project at Dartmouth College in the US. For decades, most AI meant hand-written rules: experts tried to write down everything a computer needed to know. That worked for narrow jobs like chess but broke down in the messy real world, because nobody can write rules for every situation.
The shift to learning from examples
From the 2010s, a different approach took over: machine learning, where the system works out the rules itself by studying examples. A particular style called deep learning, loosely inspired by networks of brain cells, turned out to be very good at this when given lots of data and lots of computing power. It powered big jumps in speech recognition, translation and photo recognition. Your phone recognising your face is deep learning.
Why AI suddenly felt different in 2022
In late 2022 ChatGPT made a new kind of system available to everyone: a large language model (LLM) wrapped in a chat window. LLMs learn from a vast amount of text to predict what words should come next. At enough scale, that simple-sounding skill produces something that can explain, write, translate, summarise and reason through problems in ordinary language. Since then, every big tech company has released its own, and they now also handle pictures, voice, documents and, increasingly, tasks done on your behalf.
Narrow, general and the hype in between
Every AI system available today is, strictly speaking, narrow: built and trained for particular kinds of tasks, even if language models cover a surprisingly wide range. Artificial general intelligence (AGI) means a system that could match people at almost any thinking task. Some AI companies say they are working towards it; many researchers disagree about whether or when it will arrive. When you read headlines, it helps to separate what tools can do today (which you can test yourself) from predictions about the future (which no one can test yet).
How to think about AI's strengths and weaknesses
A useful mental model is "a very well-read intern with no common sense guarantee":
- Well-read: it has absorbed patterns from more writing than any person could read in many lifetimes, so it's fluent in almost any topic.
- Intern: it works quickly and willingly, but its work needs checking, especially on facts, numbers and anything high-stakes.
- No guarantee: it doesn't check its answers against reality unless it's using a tool such as web search, so it can "fill in" plausible-sounding details. This is called hallucination.
The honest debates
AI raises real questions you'll meet in the news: how it was trained (including on copyrighted work), its energy use, bias inherited from its data, the effect on jobs, and how it should be regulated. Lesson 20 looks at these fairly. None of them stop you using AI sensibly today, but they're worth understanding.
What this means for you
You don't need any of the history to use AI well. But knowing that it predicts rather than knows explains almost every tip you'll learn: give it context, ask follow-up questions, check important facts, and keep private information out.
Try this prompt
I've never really used AI before. In five short bullet points, explain what you can and can't help me with, using examples from everyday life.
Replace the [brackets] with your own details
On haystak
Quick check · 1 of 3
What makes modern AI different from older software?
Next step
Chatbots, assistants and models: who's who
Make sense of the names you keep hearing, so you know what you're choosing between.