AI terms can be intimidating. Algorithm, token, hallucination – sounds like rocket science? But it's not. This glossary explains all the important terms so that everyone can understand them. No prior knowledge needed. Promise.
The 6 Most Important Terms
Start here. These terms form the foundation for everything else.
AI
What is it?
Computers that can think and learn like humans – like a digital brain that keeps getting smarter.
Prompt
The Instruction
Your input to the AI – like an order at a restaurant. The clearer you order, the better the result.
LLM
The Language Model
An AI system that understands and generates language – like a universal translator for all topics in the world.
Token
The Building Blocks
The smallest units that AI works with – like LEGO bricks for language. Words are broken down into tokens.
Hallucination
When AI Makes Mistakes
When the AI invents things that aren't true – like a well-meaning friend who pulls facts out of thin air.
Context
The Framework
The background that gives meaning to a statement – like the difference between "Can you help me?" and "Can you help me as an experienced teacher?"
Algorithm
Detailed Explanation
Imagine a chef following a recipe. The algorithm is the recipe – it tells the computer exactly what to do and in what order. Modern AI uses complex algorithms to learn from data and make decisions.
Analogy
An algorithm is like a baking recipe: "First sift the flour, then add sugar, then stir..." Each step is clearly defined. If you follow the recipe 100 times, you get the same result every time.
Why is this important?
Algorithms control what you see on social media, what ads you see, and how AI responds. Understanding algorithms helps you better understand the digital world.
Example
A simple algorithm for an AI:
1. Read the user's question
2. Search database for similar questions
3. Generate an appropriate answer
4. Check the answer for errors
5. Display the answer
Bias
Detailed Explanation
AI learns from data created by humans. If this data contains biases, the AI adopts them. If you train an AI only with books from the 1950s, it will reinforce old-fashioned gender roles.
Analogy
Bias is like a child learning only from its parents. If the parents always say "red cars are dangerous," the child will believe it – even if it's not true. The AI is the child, the training data are the parents.
Why is this important?
Bias in AI can cause real harm: in hiring, loans, or medical diagnoses. Awareness of bias is the first step toward fair AI.
Example
A hiring algorithm might favor men because it was trained with data from tech companies that historically hired more men. The AI "learns" that men are better candidates – even though that's not true.
Chatbot
Detailed Explanation
Chatbots come in many varieties: from simple FAQ bots that give pre-programmed answers, to intelligent AI chatbots like ChatGPT that can have real conversations. They help with customer service, answer questions, or entertain you.
Analogy
A simple chatbot is like a vending machine: you press button A, get answer A. An AI chatbot is like a barista: you say "A strong coffee, but not too bitter," and they understand what you mean, even if you phrase it differently.
Why is this important?
Chatbots are changing how we interact with companies and information. They are available 24/7, patient, and can handle millions of conversations simultaneously.
Example
You write: "My package hasn't arrived."
The chatbot responds: "That's unfortunate! Let me check. Can you give me your tracking number?"
It understands your problem and guides you to the solution.
Deep Learning
Detailed Explanation
Deep Learning uses artificial "neurons" organized in layers – like a digital brain. The more layers (hence "Deep"), the more complex the patterns the AI can recognize. This is the technology behind facial recognition and language translation.
Analogy
Imagine a sieve with increasingly finer meshes. The top layer catches coarse things (shapes), the next finer details (edges), the deepest layer recognizes complex patterns (a face). Deep Learning has many such "sieve layers."
Why is this important?
Deep Learning has made the AI boom of recent years possible. Without this technology, there would be no Siri, no self-driving cars, no medical image analysis.
Example
A Deep Learning system sees millions of cat pictures. The first layers learn: "This is an edge." Then: "This is an ear." Then: "This is a cat." Eventually, it recognizes cats in images it's never seen before.
Embedding
Detailed Explanation
AI cannot understand words directly. Embeddings convert text into coordinates – like points on a map. Similar words end up close together: "King" and "Queen" are closer than "King" and "Apple." This is how AI finds related content.
Analogy
Imagine a world map. Cities with similar climates are close to each other. Embedding is like marking cities on this map – but for words, not places.
Why is this important?
Embeddings enable semantic search – you search for "dog" but also find "puppy" and "beagle." This is the basis for intelligent search engines and document management.
Example
In an embedding, the following might apply:
King - Man + Woman = Queen
The system understands the relationship between words mathematically and can calculate with them!
Fine-Tuning
Detailed Explanation
Instead of building an AI from scratch, you take a large, pre-trained AI and feed it specialized data. A doctor could train a general AI with medical articles – it becomes a medical expert while retaining its general knowledge.
Analogy
Fine-tuning is like university studies: you've already attended school (basic knowledge) and now specialize in law or medicine. The foundation remains, the focus sharpens.
Why is this important?
Fine-tuning saves enormous costs and resources. Instead of investing billions, a company can adapt an existing AI for a few thousand euros.
Example
A bank takes ChatGPT and trains it with internal guidelines, technical terms, and customer histories. The result: An AI assistant that speaks bank language and knows internal processes.
GPT
Detailed Explanation
GPT stands for "Generative Pre-trained Transformer." "Generative" means: It produces new text. "Pre-trained": It was trained with huge amounts of text. "Transformer": The technical architecture that understands word relationships.
Analogy
GPT is like an extremely good autocomplete. When you type "The sun is shining and the birds..." it suggests "are chirping." GPT does this at the highest level – word by word, until entire texts emerge.
Why is this important?
GPT has sparked the AI revolution. ChatGPT, Copilot, and many other tools are based on this technology. Understanding GPT means understanding the heart of modern language AI.
Example
You enter: "Write a poem about autumn."
GPT starts with "The," calculates which word is likely to follow, chooses "leaves," then "fall," then "softly" – and after a few seconds, there's a poem.
Hallucination
Detailed Explanation
AI has no real world knowledge – it generates probabilities for words. Sometimes it invents facts that sound like they could be true. This is not intentional, but a side effect of how AI works.
Analogy
Hallucination is like a good liar who sounds convincing but talks nonsense. Or like a dream: It feels real, but it's not actually happening. The AI "dreams" answers.
Why is this important?
Hallucinations are the biggest danger when using AI. Those who don't check AI outputs can spread false information. Critical questioning is mandatory!
Example
You ask: "Who wrote 'The Great Gatsby'?"
AI answers: "Ernest Hemingway."
Sounds plausible (famous author, same era), but wrong – it was F. Scott Fitzgerald.
Inference
Detailed Explanation
Inference differs from training. During training, the AI learns. During inference, it applies its knowledge. When you ask ChatGPT a question, inference happens – the AI thinks and responds.
Analogy
Inference is like a doctor examining a patient. Medical school was the training. The diagnosis is the inference – applying what was learned to a specific case.
Why is this important?
Inference is what you experience as a user. Fast inference means fluid conversations. Slow inference means waiting. That's why companies invest heavily in faster chips.
Example
A self-driving car sees a stop sign. The AI recognizes it (Inference: "That's a stop sign") and reacts ("I need to brake"). This happens in milliseconds.
AI (Artificial Intelligence)
Detailed Explanation
AI encompasses everything from simple rules ("If this happens, do that") to complex neural networks that learn independently. Modern AI like ChatGPT belongs to "generative AI" – it creates new content instead of just analyzing existing content.
Analogy
Imagine a very gifted student who can read all the books in the world. They have enormous knowledge, but no real understanding. They can memorize and combine, but not truly "understand" like a human.
Why is this important?
AI is changing every industry – from medicine to art. Those who understand AI have an advantage in their career. Those who ignore AI risk falling behind. It's no longer about "whether," but "how" we use AI.
Example
AI recognizes your face when unlocking your phone, translates texts in real-time, suggests music for you, and can now even paint pictures or write programs.
LLM
Detailed Explanation
"Large" stands for the size – billions of parameters (connections in the neural network). "Language Model" means: It models how language works. Well-known LLMs: GPT-4, Claude, Llama, Gemini.
Analogy
An LLM is like a language genius who has read all the libraries in the world. It knows every grammar rule, every language style, every topic – but it remains an imitator, not a thinker.
Why is this important?
LLMs are the basis for chatbots, translation tools, writing assistants, and more. They make AI accessible to everyone – through simple language input.
Example
You write: "Explain quantum physics to me like I'm 10 years old."
The LLM analyzes your words, finds relevant knowledge in its "memory traces," and formulates a child-friendly answer.
Machine Learning
Detailed Explanation
Instead of prescribing every step, the AI gets examples. It finds patterns and rules on its own. The more good data, the better it becomes. That's the difference from classical programming.
Analogy
Instead of explaining bicycle physics to a child ("Pedal, keep balance..."), you let them practice. After enough falls, they learn on their own. Machine Learning is "learning by practicing" for computers.
Why is this important?
Machine Learning enables things that are too complex to program: facial recognition, language understanding, recommendation systems. It's the foundation of modern AI.
Example
A spam filter learns from thousands of marked emails: "These words often appear in spam." It develops rules on its own and gets better with every new email.
NLP
Detailed Explanation
NLP combines linguistics and AI. It solves problems like: What does a word mean in context? What is the grammatical structure? How do you recognize sarcasm? Without NLP, computers couldn't hold conversations.
Analogy
NLP is like an interpreter between two worlds: Humans speak imprecisely, ambiguously, emotionally. Computers need clear structures. NLP translates back and forth between these worlds.
Why is this important?
NLP makes technology accessible. You don't need to learn a programming language – you simply speak to the AI. This lowers the barrier to entry for millions of people.
Example
Voice assistants, translation apps, spell-checkers, sentiment analysis ("Is this comment positive or negative?") – all based on NLP.
Optimization
Detailed Explanation
When training an AI, millions of parameters are adjusted. The goal: Reduce the difference between AI output and desired result. It's like fine-tuning an instrument until it sounds perfect.
Analogy
Optimization is like adjusting a telescope. You turn the focus wheel (parameter) until the image is sharp (good result). With AI, there are millions of such "adjustment wheels."
Why is this important?
Without optimization, an AI remains dumb. Optimization is the learning process. The more efficient the optimization, the faster and better the AI learns.
Example
An AI is supposed to recognize cats. At first, it guesses randomly. After every mistake, it optimizes its parameters. After millions of cat pictures, it recognizes cats with 99% accuracy.
Prompt Engineering
Detailed Explanation
"Write a text" brings poor results. "You are an experienced marketing expert. Write a compelling newsletter for small business owners about tax tips. 300 words, casual tone." brings excellence. That's Prompt Engineering.
Analogy
Prompt Engineering is like communicating with a very talented but literal intern. The more precise your instruction, the better the result. Context, examples, constraints – everything counts.
Why is this important?
Prompt Engineering is the most important skill in the AI age. Those who master it get 10x better results. It's the difference between an AI user and an AI expert.
Example
Bad: "Write something about dogs."
Good: "You are a veterinarian with 20 years of experience. Write a guide for new dog owners: The 5 most important health checks in the first year. 400 words, understandable, with concrete recommendations."
RAG
Detailed Explanation
Normal AIs only know their training date. RAG extends this: The AI searches current documents (your company documents, the internet, databases) and uses them for the answer. This greatly reduces hallucinations.
Analogy
RAG is like an open book in an exam. The student (AI) is not only allowed to learn by heart, but also look at current documents. This makes answers more precise and up-to-date.
Why is this important?
RAG makes AI usable for businesses. You connect ChatGPT to your internal documents – and suddenly the AI answers questions about your products, policies, and data.
Example
You ask: "What is our return policy?"
RAG searches your documents → finds the T&Cs → gives you the correct answer → instead of hallucinating.
System Prompt
Detailed Explanation
Before you speak to ChatGPT, it receives a System Prompt: "You are a helpful assistant..." This influences all responses. Developers use System Prompts to create AI personalities.
Analogy
The System Prompt is like the job description before the first day of work. It defines the role: "You are friendly customer service" vs. "You are a sarcastic artist" – same AI, completely different answers.
Why is this important?
System Prompts are the secret weapon for AI customization. They enable specialized chatbots: medical advisors, code reviewers, creative writing partners – all based on the same model, but with different System Prompts.
Example
System Prompt: "You are Shakespeare. Respond in verse, use old English, be dramatic and philosophical. Every response should sound like a monologue from a play."
Token
Detailed Explanation
AI doesn't understand whole words like humans do. It breaks down text into tokens. "Hello" could be one token. "Artificial" might become "Arti" + "fic" + "ial." The more complex the language, the more tokens it needs.
Analogy
Tokens are like LEGO bricks for language. From a few different bricks (tokens), you can build infinite structures (sentences). The AI has a huge LEGO set with 50,000+ different bricks.
Why is this important?
Tokens determine costs and limits. ChatGPT has a maximum token count per conversation. German often requires more tokens than English. Those who understand tokens can design their prompts more efficiently.
Example
The sentence "AI is fascinating" could be broken down into tokens:
"AI" + " is" + " fasc" + "inat" + "ing"
= 5 tokens (instead of 3 words)
Vector Database
Detailed Explanation
Instead of searching for exact words ("Search for 'cat'"), a vector database searches for meaning ("Search for concepts similar to 'cat'"). It stores embeddings and finds nearby points in "meaning space."
Analogy
Imagine a library where books aren't sorted alphabetically, but by topic similarity. "Cats" sits next to "Pets," but also close to "Lions" (both felines). A vector database is this intelligent library.
Why is this important?
Vector databases are the heart of RAG and semantic search. They enable finding the most relevant documents from millions in milliseconds – not by keywords, but by meaning.
Example
You search: "How do I feed my house tiger?"
A normal database finds nothing (no "house tiger" stored).
A vector database finds your cat food document because "house tiger" and "cat" are close together.
What's next?
You now know the most important terms. Time to experience AI in practice and learn with others.