Einführung in LLMs und Prompting

Large Language Models (LLMs) are AI built to understand and generate human-like text. Picture super-smart text predictors trained on vast amounts of writing.

LLMs learn by spotting patterns, grammar, and facts across huge text datasets, then predicting the next likely word. They don't truly understand, but they mimic us remarkably well.

You've likely used some already: ChatGPT (OpenAI), Gemini (Google), and Llama (Meta). They differ in size, capability, and how you access them.

Prompting is simply how you talk to an LLM - the input you give it. It can be a question, a detailed request, or text to analyze. It tells the model what you want.

Output quality tracks prompt quality. A clear, well-structured prompt yields accurate, useful answers; a vague one invites generic or wrong results - like giving fuzzy directions.

LLMs are versatile: they generate text, summarize long documents, translate between languages, answer questions, brainstorm ideas, and even draft code.

LLMs have limits: hallucinations (confident but false claims), outdated knowledge past their cutoff, no real-world common sense, and training-data bias. Verify what matters.

Prompt engineering is the craft of writing prompts that get the best results. It means knowing how LLMs process input and steering their behavior - essential for AI apps.

For developers, prompt engineering is about control and predictability: consistent outputs your app can rely on, integrated cleanly into your software.

Which of the following best describes a Large Language Model (LLM)?

Recap: LLMs are pattern-driven text generators with real strengths and clear limits. Prompting is how you steer them, and prompt engineering is the dev skill. Next: prompt structure.