AI Series Part 1 : So what actually is it?

This is the first in a new series of articles about artificial intelligence — what it is, how it works, what it can and cannot do, and how to use it without accidentally giving away more than you intended. No technical knowledge required.

You have almost certainly heard the word artificial intelligence more times in the past two years than in the rest of your life combined. It is on the news, in your phone, in your workplace, in your children's school. Everyone seems to be talking about it.

But if someone asked you to explain, in plain English, what AI actually is — not what it does, but what it is — most people would find that surprisingly difficult. And that matters. Because the less you understand about what something is, the harder it is to make good decisions about when to trust it, when to be cautious, and when to think twice before handing it your photograph.

So let us start from the beginning.

What is artificial intelligence?

The honest answer is that artificial intelligence is not one thing. It is a broad term for computer systems that can do things which, until recently, only humans could do — things like understanding language, recognising faces, translating between languages, generating images, writing text, and making decisions based on complex information.

What all of these systems have in common is that they learn from examples rather than following a fixed set of rules. A traditional computer programme does exactly what it is told, step by step. If you ask it something it was not specifically programmed for, it cannot cope. AI systems work differently. They are trained on enormous amounts of data — millions of books, images, conversations, websites — and through that training they develop the ability to recognise patterns and generate responses that were never explicitly programmed.

Think of it this way. If you wanted to teach a child to recognise a cat, you would not write them a manual listing every possible cat — every colour, size, breed, and angle. You would show them hundreds of cats. After a while they would just know what a cat looks like, even if they had never seen that particular cat before. AI learns in a similar way, just with far more data and far more speed.

The type of AI most people encounter today — the kind that can hold a conversation, write an essay, or generate an image from a description — is called a large language model, or LLM. It is trained on a vast amount of written text and has learned, with extraordinary sophistication, how language works and how to produce responses that feel natural and useful. ChatGPT, made by OpenAI, is one. Claude, made by Anthropic, is another. Google's Gemini is a third. They are different products built on the same fundamental idea.

What AI is not

This is just as important as understanding what AI is — because a lot of anxiety, and a lot of misplaced trust, comes from misunderstanding this part.

AI does not think. It processes patterns in data and generates responses that fit those patterns. It has no opinions, no feelings, no awareness of itself, and no understanding of the world in the way that you do. When a chatbot says "I understand how frustrating that must be," it is not understanding anything. It is producing the kind of response that, based on its training, tends to follow that type of message.

AI does not know things the way you know things. A large language model does not look up facts in a reliable database. It generates text based on patterns learned during training. This means it can be confidently wrong — producing a plausible-sounding answer that is simply not true. This is called a hallucination, and every AI system does it to some degree. Never rely on an AI for something important without checking it.

AI does not remember you — unless it has been specifically designed to. Most AI chat tools start fresh every conversation. The AI you spoke to last week does not recall what you said. It has no ongoing relationship with you.

AI is not magic, and it is not a person. It is a very sophisticated pattern-matching system, trained by humans, run by companies, and subject to all the ordinary rules of business, privacy, and self-interest that apply to any service you use online.

We have had chatbots for years. Why is this different?

You have probably encountered chatbots before — the little pop-up windows on websites that say "Hi! How can I help you today?" and then prove completely useless the moment you ask anything beyond a simple question. So what makes the current generation of AI different?

The old chatbots were essentially decision trees. They recognised specific words and phrases and followed a pre-written script in response. If you typed something they were not programmed to handle, they failed immediately. They could not understand context, follow a conversation, or adapt to what you actually meant.

Modern AI models were trained on a scale that was simply not possible before — hundreds of billions of examples of human language — and using techniques that allow them to understand context, nuance, and even tone in a way the old systems never could. The difference between a 2015 chatbot and a 2025 AI assistant is roughly the difference between a vending machine and a well-read human who has absorbed everything ever written and can discuss any of it with you.

This is genuinely impressive. It is also genuinely new, which is why so many people are still working out how to use it wisely.

What is an AI agent?

You may have started hearing the phrase AI agent more recently. It is worth understanding what this means, because agents are where AI is heading next — and they raise their own questions about trust and oversight.

A standard AI assistant — like ChatGPT or Claude — responds to what you type. You ask a question, it answers. You ask it to write something, it writes it. The conversation is the limit of what it does.

An AI agent is different. An agent can be given a goal and then take steps to achieve it — by browsing the web, writing and running code, sending emails, booking things, managing files, and connecting to other services — all without you manually directing each step. You might tell an agent "plan a weekend trip to Edinburgh for three people and book the train tickets" and it will go away and actually do it, checking timetables, comparing options, and completing the booking.

This is genuinely useful. It is also a meaningful shift in how much autonomy you are giving to a piece of software. An agent that can send emails on your behalf, access your calendar, browse the web, and spend money is a very different thing from a tool that answers questions. The more capable the agent, the more important it is to understand what permissions you have given it and what it might do with them.

AI agents are already in use in workplaces and are coming to consumer products fast. Understanding what they are — and that they act, rather than just respond — is the first step to using them sensibly.

Why you should think before you type

Here is something most people do not think about when they open an AI chat window: what happens to what you type?

When you type something into an AI assistant, that message travels over the internet to a server — usually in another country — where it is processed by the AI model, which is run by a company. That company has terms and conditions. Those terms and conditions explain what they do with your conversations.

Most reputable AI services have clear policies about this, and many allow you to opt out of having your conversations used for training purposes. But the defaults vary. And people regularly type things into AI chat windows that they would not post publicly online — personal problems, confidential work details, medical concerns, the names and situations of people they know.

Before you use any AI service regularly, it is worth spending two minutes on the question: does this company store my conversations, and if so, for how long and for what purpose?

The image trend trap

Now for the part that catches a lot of people out — and is genuinely worth sharing with everyone you know who uses social media.

Every few months a new AI image trend sweeps the internet. One month it is turning your photo into a Studio Ghibli animation. Another month it is a vintage yearbook photo. Another is a fantasy portrait. They are fun, they look great, and tens of millions of people join in.

Here is what is worth knowing before you do.

When you upload a photograph of your face to an AI image generation service, you are handing that service a high-quality image of your actual face. The terms and conditions of many of these services — the long document that almost nobody reads — can grant the company a licence to use the images you upload, or the images generated from them, for their own purposes. In some cases, that includes using them as training data for future AI models. In some cases it includes using your generated likeness in ways you did not specifically consent to.

This matters for several reasons. Your face is biometric data — data about your physical body — and it is uniquely personal in a way that a password or an email address is not. A password can be changed. Your face cannot. Images of your face can be used to create deepfakes — realistic fake videos or images that show you doing or saying things you never did. They can be used to build fake social media profiles. They can potentially be used to bypass some forms of facial recognition.

None of this means every AI image tool is malicious. Many are run by reputable companies with responsible policies. But the question to ask before you upload a photo is simple: have I actually read what this company will do with my image? And if not, am I comfortable with the fact that I do not know?

In February 2026, 61 data protection authorities from around the world issued a joint statement specifically about this issue — raising concern about AI image generation services that collect, retain, and potentially misuse images of real individuals without meaningful consent. The concern is real, it is official, and it applies equally to a ten-year-old uploading a selfie and a seventy-year-old joining a fun social media trend.

What does this mean for me?

Treat AI chat like a public conversation, not a private one. Do not type personal details, confidential work information, or sensitive information about other people into AI chat tools unless you have checked the privacy policy and are comfortable with how your data is handled.

Before uploading your photo to any AI image trend, ask one question. What does this company's terms and conditions say about images I upload? If you cannot find the answer in under two minutes, that is itself an answer.

Understand that AI agents are a different kind of tool. They do not just answer questions — they take actions. Be thoughtful about what permissions you grant and what they can do on your behalf.

Remember what AI is and is not. It is a powerful and genuinely useful tool, trained on human knowledge, run by companies, and capable of impressive things. It is not a friend, not a person, and not infallible. It can be wrong, confidently and fluently. Treat its answers as a helpful starting point, not a final authority.

Talk to the children in your life about this. The generation growing up with AI available on their phones needs to understand these basics before they form habits that are hard to change. The earlier those conversations happen, the better.

🧠 The Human Factor

Next in the AI series: How to spot AI-generated content — and why it matters that you can.