3.1. | What is an LLM?
What is this module about?
- For the first time, you get direct access to large language models like ChatGPT, Claude, or Gemini – via the xpand AI platform.
- You will learn how to navigate there (→ screenshot with the Plus area).
- You will learn what an LLM is – and how you can use it effectively.
What is a Large Language Model anyway?
A Large Language Model (LLM) is an AI model that has been trained with billions of text examples. It recognizes patterns in language and can therefore:
- Write or summarize texts
- Develop ideas
- Structure knowledge
- Answer questions
- Analyze processes and make suggestions
In short: It is a language-based dialogue partner – but with a vast knowledge base and pattern sensitivity.
How does an LLM work?
LLMs work on the principle of probability calculation. They have been trained with enormous amounts of text and have learned to recognize patterns. When you make a request, the following happens:
- Input Processing: Your text is broken down into small units (tokens) and translated into a machine-readable format.
- Pattern Matching: The model searches its „memory“ for similar patterns and contexts.
- Prediction: It calculates which words and sentences are most likely to follow as a meaningful answer.
- Output Generation: The model generates its response step by step, with each new word building on the previous one.
Important: LLMs do not understand like humans. They calculate linguistic probabilities. This is why they sometimes provide answers that sound logical but are factually incorrect (→ see Module 4.3).
Special features and limitations:
- LLMs have a knowledge cut-off date – they do not know information created after this date.
- They cannot access the internet (except for special versions with integrated search tools).
- They have no memory across different conversations.
- They can sometimes „hallucinate“ – that is, provide plausible-sounding but fabricated information.
Good to know: The LLM remembers what you said previously within a single conversation. If you start a new conversation, it starts from scratch.
How to find the area on the platform
- Open the xpandAI platform
- Log in (if not done automatically)
- Click on the „+“ in the top navigation
- Choose your desired model – e.g., GPT-4, Claude 3, or Gemini
- Start your first own prompt
Navigation to the LLM area on the xpandAI platform
xpand Tip
If you’re still unsure where to start, just begin with a real question from your daily life – just as you would ask a colleague. Observe what the model makes of it – and gradually adjust your prompt. This is the easiest way to learn how to use it.
Your next step
In the next module (3.2), we will look at which LLM is suitable for which task and what the specific features of the models are.
Your Takeaway
- An LLM is an AI model that has been trained with billions of text examples and recognizes patterns in language
- LLMs work through probability calculation, not by „understanding“
- They have a knowledge cut-off date and can sometimes „hallucinate“
- The easiest way to learn: Start with a real question and observe how the model reacts