The full course in one video
Understanding AI – the basics (the full course)
The complete foundations course as one film: what AI is, how language models work, where their limits are and how answers become reliable on your own knowledge.
Understanding AI: from language models to the EU AI Act
25 videos · around 49 minutes
All areasThe largest area, and the only one that has nothing to do with operating the software. These videos explain how artificial intelligence arrives at its answers: what separates a language model from classic software, how it learns, what tokens and context windows are, why hallucinations happen and what grounding and RAG change about them. The whole course is also available as one continuous film.
The second half is about using AI in a company: what happens to your data, which jurisdiction applies, what a model looks like in operation, where bias comes from, what the EU AI Act requires and how to introduce AI responsibly. Written for everyone who has to take part in the decision without having studied computer science.
What is AI?
What sets artificial intelligence apart: recognising patterns from examples instead of fixed rules.
AI, ML and LLM
The map of terms: AI as the umbrella, machine learning as the usual method and the language model as one kind of it.
How an AI learns
How a model learns from large amounts of data in training – parameters instead of programming, and training versus use.
Language models
What a language model is: it predicts the next word, and from that apparent understanding emerges.
Model and tool
The difference between an AI model and an AI tool (GPT versus ChatGPT): the model is the engine, the tool builds interface, memory and rules around it.
The prompt
The input is the instruction: how a clear, precise question leads to a better answer.
Tokens and context window
How an AI reads text in chunks (tokens) and why its memory per request is limited (the context window).
Hallucinations
Why an AI can be convincingly wrong – and why you should not blindly trust answers but check the sources.
Prompt injection and guardrails
How hidden instructions can be smuggled in (even via documents or websites) and how safeguards contain them.
Grounding and RAG
Answers based on real company knowledge instead of guessing: the AI looks things up first and then backs every statement (retrieval augmented generation).
Semantic search
How an AI finds by meaning rather than just by keyword – the basis for finding the right knowledge for a question.
The limits of RAG
Why questions like "show me all ..." do not return a reliably complete list, and when a classic keyword search is the better choice.
AI and your data
What happens to your input: where the model runs, who has access and whether data is used for training.
Regulation and jurisdiction
The GDPR as the framework – and why it matters not only where the data sits but which law the provider is subject to.
Running an LLM
What running large and small models costs – memory, hardware and the trade-off between performance, cost and control.
Using AI responsibly
What AI is and is not good for – and why the human stays in control and checks the results.
AI agents and automation
AI that does not just answer but breaks a goal into steps, uses tools and works autonomously.
Multimodal AI
A model that understands not only text but also images, speech and documents.
Reasoning AI
Reasoning models take time for intermediate steps before answering – better on tricky tasks, but slower and more expensive.
Adapting AI to your knowledge
Prompting, RAG or fine-tuning: the three ways to teach an AI your knowledge – and when each one fits.
Open source or proprietary
A ready vendor service versus open weights: convenient and strong versus control and independence.
Bias and fairness
Why an AI picks up skews from its training data and is not neutral by default – and how to contain that.
The EU AI Act
The EU AI law in brief: regulation by risk and what matters for companies (not legal advice).
Introducing AI in your company
From idea to use: a clear use case, taking people along, clear rules and starting small.