- Artificial intelligence (AI)
- Artificial intelligence refers to computer systems that perform tasks normally requiring human intelligence – such as understanding language, writing text, recognising images or making predictions. Modern AI learns from large amounts of data instead of being programmed with fixed rules.
- Machine learning
- Machine learning is a subfield of AI in which a system learns patterns from examples and improves without each rule being programmed individually. The more suitable the data, the better the results.
- Generative AI
- Generative AI creates new content such as text, images, code or music – unlike AI that only classifies or predicts. ChatGPT and Claude are well-known examples of generative AI.
- Large language model (LLM)
- A large language model (LLM) is an AI trained on enormous amounts of text that understands and generates language. It predicts the most likely next word and can answer questions, write text and summarise content.
- ChatGPT
- ChatGPT is an AI chatbot by OpenAI based on a large language model. You ask questions or give tasks in normal language, and ChatGPT responds with text – useful for writing, research, ideas and much more.
- Claude
- Claude is an AI assistant by Anthropic, comparable to ChatGPT. Claude is designed for helpful, safe and traceable answers and is also used at absofort to assess practical tasks.
- Prompt
- A prompt is the input or instruction you give to an AI – your question or task. The clearer and more precise the prompt, the better the result.
- Prompt engineering
- Prompt engineering is the skill of formulating prompts so that the AI delivers strong and relevant results. This includes clear goals, context, examples and step-by-step improvement of the instruction.
- Token
- A token is a text unit into which AI language models split text – usually a word or part of a word. The length of inputs and outputs, and the cost of many AI services, are measured in tokens.
- Hallucination
- A hallucination is an AI answer that sounds convincing but is factually wrong or invented. AI results should therefore always be checked critically, especially for figures, sources and names.
- AI agent
- An AI agent is an AI system that not only answers, but independently performs several steps to achieve a goal – for example using tools, searching for information and completing subtasks.
- Training & fine-tuning
- Training is the process by which an AI model learns from data. Fine-tuning means further adapting an already trained model with additional, specific data – for example to an industry or task.
- Bias
- Bias is a systematic distortion in AI results caused by one-sided training data. AI can therefore produce unfair or one-sided answers – an important reason to question results.
- Data protection & AI (FADP)
- When using AI, sensitive personal data must not be entered into tools without careful consideration. In Switzerland, the revised Data Protection Act (FADP) applies; companies must clarify which data may be sent to which AI services.