Artificial intelligence
Artificial intelligence (AI) is changing the way we work. Find out how to use AI tools effectively and which ETH rules and guidelines you need to follow. The “AI compass” event series also provides up-to-date knowledge and practical examples.
The quality of the results generated by AI tools depends largely on your input – the prompt. The clearer and more precise your instructions, the more helpful the response.
The five elements of a good prompt
Clearly describe what you want the AI to do. Where possible, start with an action verb.
Example: “Create an outline for a guide on topic XY.”
Provide relevant background information. The context influences the quality of the result.
Examples:
- “The target audience is first-year students.”
- “The content is aimed at technical and administrative staff.”
- “The topic will be covered in an introductory session.”
- “The text should be summarised in a presentation.”
Important: Do not enter confidential or strictly confidential information or sensitive personal data (see Information Security and Artificial Intelligence (AI) – Guide for Users of AI at ETH Zurich).
Examples, templates or concepts help the AI understand the desired result more clearly.
Example: “Follow the ETH guidelines for ...”
Important: Do not enter confidential or strictly confidential information or sensitive personal data (see Information Security and Artificial Intelligence (AI) – Responsible Use of AI).
Specify how you would like the result to be presented, for example:
- bullet points
- a table
- paragraphs
- a presentation
If necessary, specify the file format you require, for example:
- Word document
Example: “Present the results in a table with the columns ‘Topic’, ‘Description’ and ‘Practical example’.”
Define the writing style you want, for example:
- factual
- academic
- motivating
- easy to understand for beginners
Example: “Use factual, easy-to-understand language.”
Important
Use AI services responsibly. Do not enter confidential or strictly confidential information. Follow ETH regulations.
Do not rely on AI-generated results without checking them. Review them critically.
- Recording: “KI-Kompass: Datensicherheit und Leitplanken in der KI-Nutzung” (German only)
- Recording: “KI-Kompass: Effizient arbeiten mit Microsoft Copilot” (German only)
- Course: “KI-Kompass: Was sind KI-Agenten? Ein Kompakt-Einstieg” (German only)
- Course: “KI-Kompass: Einsatz von KI-Agenten in der Praxis” (German only)
- Course: “KI-Kompass: KI-Werkstatt Modul 1: Erst denken, dann programmieren: Wo lohnt sich KI wirklich?” (German only)
- Course: “KI-Kompass: KI-Werkstatt Modul 2: Wie baue ich einen KI‑basierten Chatbot?” (German only)
- Course: “KI-Kompass: KI-Werkstatt Modul 3: Coaching Session – bringen Sie Ihren Chatbot ans Ziel?” (German only)
Browse recordings of past AI Compass events and discover courses and useful tools.
General introduction
Copywriter and Creative Director Albert Heiser shows how ChatGPT can help overcome the fear of the blank page. At the same time, he points out that AI-generated texts are often formulaic and lack originality. Carefully crafted prompts can significantly improve the quality of the results. Available only in German.
How can AI tools be used without compromising data protection and information security? Johannes Hadodo, Chief Cybersecurity Officer at ETH Zurich, explains which tools and data are permitted, what to look out for with AI agents and how to recognise risks such as phishing, deepfakes and fabricated AI responses.
- Classify data: Public and internal data (see classification system of ETH Zurich) may be used depending on the tool. Confidential data belongs only in approved services, and strictly confidential data generally must not go into the cloud. The internal environment Leonhard Med is available for certain use cases.
- Check tools: With the appropriate licence, Microsoft Copilot is also approved for confidential data. Tools that have not been assessed should be used only occasionally and with internal data at most.
- Prompt carefully: Enter only the data you need, disable history and model training where possible and critically check results, sources, and copyright.
- Secure AI agents: Do not run agents on production systems. Test in a separate environment with synthetic or public data and as few permissions as possible.
Michael Vonlanthen shows how Microsoft Copilot can support everyday work, from first prompts to practical applications. He demonstrates how to visualise conversations, adapt content for specific audiences and summarise tasks from emails, chats and calendars.
With the appropriate licence, Copilot is integrated into the Office applications and can, for example, consolidate and analyse data in Excel. Vonlanthen stresses that AI results must always be checked.
Immanol Schlag, a researcher at the ETH AI Center, explains ...
- Technical details of language models: How large language models (LLMs), machine learning and generative models work, and the role of training data.
- Post-training of language models: Fine-tuning and reinforcement learning from human feedback (RLHF) to make language models more useful and ethical.
- Swiss AI Initiative: A national research initiative led by ETH Zurich and EPFL that aims to develop a transparent and responsible Swiss LLM.
- Examples of applications: Examples of how AI can be used in teaching, such as the virtual tutor “Apple”, which supports students.
Available only in German.
AI tools can be used not only for emails and other texts, but also for presentations and learning. This webinar provides an overview of tools such as Microsoft Copilot, PowerPoint, Gamma, ChatGPT, ScienceOS, Perplexity and NotebookLM, along with important information on data security. Available only in German.
Mike Schaffner shows how AI and prompt engineering can be used to develop ideas and drive innovation – from iterative idea loops and stakeholder personas to simulated panel discussions. Practical examples include White Coding and creating presentations, followed by a Q&A on tools, reasoning modes and ethics. Available only in German.
Leadership
The Leadership Companion, an AI system developed by ETH Zurich, provides personalised support in your day-to-day work and in your development as a leader.
Professor Silke Schönert from Rheinische Hochschule Köln provides insights into the use of AI in change and management processes. She shows how AI can support leaders in their day-to-day work. Available only in German.
Teaching
ETH Zurich takes a proactive approach to the use of generative AI (GenAI) in education and emphasises its responsible use by students and lecturers.
Dr Gerd Kortemeyer presents the work of the “Generative AI for Education” group at the ETH AI Center and shows how the Ethel platform uses AI to support teaching, assessment and feedback. His talk covers the underlying technology, practical applications in day-to-day teaching and learning, and questions of governance, equitable access and the future role of AI in higher education. Available only in German.
How is AI changing teaching and academic writing at ETH Zurich? Daniel Flück, a specialist in innovation management at the Unit for Teaching and Learning (UTL), and Caroline Welte from ETH Library discuss the opportunities and challenges of using AI in an academic context and explore practical approaches. They focus on ETH Zurich’s proactive approach, which encourages the responsible, transparent and equitable use of AI and fosters critical thinking. Available only in German.
Academic work
The ETH Library offers a self-study course on Moodle for anyone wishing to explore generative AI and AI-based tools in greater depth.
ETH Library courses help you develop your research and writing skills, deepen your knowledge of reference management and learn more about open access.
ETH rules and guidelines – using AI responsibly
Generative AI offers many opportunities. At the same time, it also involves risks – for example in relation to data protection, privacy and the reliability of content.
ETH has therefore developed guidelines for the use and dissemination of AI-generated content. Follow the applicable regulations:
Important: Confidential or strictly confidential information and sensitive personal data must not be entered into public AI systems. Always follow the classification system of ETH Zurich and ensure that data is handled in accordance with the required level of protection.
How can artificial intelligence (AI) help create a fair future, and what risks does it pose? Melinda Rieder from ETH Diversity and media artist and researcher Fernando Obieta discuss the impact of AI on diversity, inclusion and equal opportunities.
The discussion focuses on systematic biases, preconceptions and the concept of situatedness in feminist theory, as well as the technical foundations and limitations of AI models. Rieder and Obieta connect ethical questions with practical applications in research and art. Available only in German.
Is AI killing creativity – or enabling new forms of it? Drawing on examples from art, music and literature, Adrian Notz shows how AI is already being used creatively without being creative itself.
The webinar explores the role of personality, intention and expression, as well as the societal significance of creativity as a “dispositif”. Notz also examines artistic strategies for engaging with AI, critical perspectives on its impact and the ways it could transform the art world. Available only in German.
Nora Al-Badri asks what kind of AI we actually want – and for whom. Combining art, technology and social criticism, she shows how training data can lead AI to reproduce existing power structures, colonial patterns and one-sided perspectives on knowledge.
Using specific examples, she reveals biases in the representation of knowledge and the invisible labour behind AI systems. She also discusses risks such as linguistic homogenisation and anti-democratic tendencies. Available only in German.
Robin Staab, a doctoral candidate at ETH Zurich, offers insights into the challenges involved in evaluating AI systems: how can we determine whether they are fair, transparent and legally compliant? He explains why the EU AI Act is relevant to Switzerland and examines key topics such as data protection, environmental impact and the detection of AI-generated content. Staab also shows how research can contribute to the responsible development of AI. Available only in German.
A workshop on the practical use of AI tools for text, presentations, coaching and data analysis, with structured exercises based on specific use cases.
Number of participants: 4–12
Duration: 3.5 hours
Prior knowledge: Basic knowledge of AI tools and how to use them.
This workshop helps teams reflect on how they currently use AI and how they may use it in the future. Following a clear process, you identify specific use cases and define responsibilities for further integrating AI into day-to-day work.
Number of participants: 4–12
Duration: 2 hours
Prior knowledge: Basic knowledge of AI tools and how to use them.