AI stands for artificial intelligence. It means computer programs that solve tasks which would otherwise take human abilities such as learning, planning or language: writing a text, recognising an image, answering a question. Put simply: today's AI isn't programmed step by step, it learns patterns from a huge number of examples and applies them to new cases.
I have worked with Claude every day since February 2025, and I now run my business with it. My fourteen AI employees run on Claude Team, some also on open models or through the API. This post is the starting point of my basics: here we clarify the term, and the details follow in separate posts. All AI basics are on my AI Basics page.
What AI means, explained simply
The European Parliament describes artificial intelligence as the ability of a machine to display human-like capabilities such as reasoning, learning, planning and creativity. The key word is "human-like". An AI doesn't think like a person. It produces results that look as if someone had thought about them.
Google puts it even more soberly in its machine learning glossary: AI is a program or model that can solve sophisticated tasks. As examples, Google names a program that translates text and one that identifies diseases in radiological images. A model is what comes out of learning from examples: a huge collection of numbers the program uses to compute an output from an input. What exactly a model is and how it differs from the app, I explain in What Is an AI Model? Simply Explained.
The difference from ordinary software: a classic program follows rules a person wrote down beforehand, such as "if the amount is over 1,000 euros, request approval". An AI gets a huge number of examples instead and finds the rules itself. That's why it can also handle things nobody foresaw, and that's why it is sometimes wrong without anyone being able to say exactly why.
What AI stands for
AI is short for artificial intelligence. In German the same thing is called künstliche Intelligenz, or KI for short, which is why you'll see both abbreviations on German-language websites such as this one. Both mean exactly the same. The EU law on artificial intelligence, for example, is called the AI Act in English and the KI-Verordnung in German.
How the EU defines AI
For businesses the definition in the EU AI Act matters too, because obligations depend on it. I am a developer, not a lawyer, and this is not legal advice, just the operator's perspective. In Article 3, the regulation defines an AI system as "a machine-based system that is designed to operate with varying levels of autonomy" and that "infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions".
In plain words, that means:
- machine-based: it runs on computers.
- more or less on its own: it works without a person setting every step.
- infers: it works out from your input what it should produce.
- outputs: a prediction, a text or image, a recommendation or a decision.
The recitals, the explanations at the start of the regulation, state explicitly that systems working only by rules people defined in advance are not meant. A simple if-then formula in your spreadsheet therefore doesn't fall under it.
What kinds of AI there are
Lots of terms come up around AI. The most important ones fit together like this:
- AI is the umbrella term for everything described above.
- Machine learning, according to Google, is a sub-field of AI: a program learns from example data and then makes predictions for new data. More on this in my post on machine learning.
- Generative AI creates new content: text, images, music, videos or code. What it can do and where its limits are is in Generative AI Explained, with Examples.
- Language models, large language models or LLMs for short, are a form of generative AI that works with language. They are behind ChatGPT, Claude and Gemini. How they work, I explain in How Do LLMs Work? A Plain-English Guide.
- AI agents are AI systems that don't just answer but complete tasks themselves using tools. I explain the difference from a chat in What Is an AI Agent? Explained Simply.
Where you already use AI
Often you don't even notice AI is involved. The European Parliament names, among other things, recommendations when shopping online, search engines, digital assistants on your smartphone and automatic translation.
An invented but typical morning at a painting and decorating business: the spam filter sorts the advertising out of the inbox. The owner dictates a note into her phone, and the phone writes it down. A customer asks for a quote by email, and an AI chat writes a friendly draft reply, which the owner checks and sends. That's three different AI applications in one morning, and only one of them calls itself "AI".
What to watch out for
- AI imitates; it doesn't understand like a person. Results that sound convincing can still be wrong. Check numbers, names and dates.
- "With AI" on the label says little. For every tool, ask what exactly the AI in it does and with which data.
- Your data goes with it. Whatever you type into an AI service leaves your computer. Customer data belongs only in plans with a suitable contract.
- Start with a task, not with the tool. A recurring email is a better start than "let's see what AI can do".
Frequently asked questions
What does AI mean?
AI means artificial intelligence: computer programs that solve tasks we would otherwise associate with human abilities. Today that almost always means programs that have learned from a huge number of examples.
Is ChatGPT an AI?
Yes, more precisely an app in which you chat with an AI model from OpenAI. The app is the interface; the model behind it generates the answers. I explain the difference between model and app in What Is an AI Model? Simply Explained.
Can AI think?
Not the way a person does. An AI computes from what it has learned which result fits best. Some models work through intermediate steps before answering that look like thinking. Even the developers only partly understand what exactly happens inside, as Anthropic writes about its own models.
Is every computer program an AI?
No. A program that only follows fixed rules written by people is not AI, nor is it under the EU AI Act. What counts there is that the system itself infers from your input how to produce its output. According to the recitals, that can happen in two ways: through machine learning from data, or through logic- and knowledge-based approaches that draw conclusions from encoded knowledge. Whether a particular tool falls under it is a legal question. I am a developer, not a lawyer, and this is not legal advice.
When did artificial intelligence begin?
As a field of research, in the 1950s. On 31 August 1955, John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon wrote a proposal for a summer project on artificial intelligence, which took place at Dartmouth College in 1956. According to the European Parliament, more computing power, large amounts of data and new algorithms have led to the big breakthroughs in recent years.
Where to go from here
The next step on the learning path is the kind of AI you deal with most often: Generative AI Explained, with Examples. If you'd rather get started right away, read AI for Beginners: What to Know First. Terms you come across along the way are in the glossary of AI terms. How others take their first steps with AI is what we discuss in my German-speaking community Claude Practitioners.
