teaching

AI Safety Seminar

Seminar | MA-INF 4116 | Summer 2026 | 4 ECTS

A seminar introducing students to the technical foundations of AI safety, covering alignment and value specification, control and autonomy of advanced AI systems, and systemic risks. Compared to its predecessor A(G)I Ethics, the seminar concentrates on technical AI safety research rather than the broader ethical and policy debate. Students engage with recent research papers, give a presentation, and write an essay.

Learning Outcomes:

• Understand alignment and value specification challenges
• Analyze technical approaches to overseeing and controlling advanced AI
• Critically assess recent AI safety research papers
• Present research findings and write a scientific essay

Prerequisites: Basic computer science background; machine learning experience helpful but not required

Instructors: Dr. Florian Mai

Schedule: Wednesday 2:15-3:45 PM, Room B-IT 2.113a

AI Alignment

Seminar & Lab | Winter 2025/2026 | 9 ECTS

A two-part course, where students first learn about AI alignment techniques through weekly readings and class discussions. AI in weekly discussions. In the second part, students will work on a project relating to technical AI alignment issues and hand in a 6-page report.

Learning Outcomes:

• Understand state-of-the-art AI alignment techniques
• Analyze recent research developments in the field of AI alignment and AI safety
• Evaluate limitations of current approaches
• Design and implement AI alignment techniques

Prerequisites: One of the following: Introduction to Natural Language Processing, Reinforcement Learning, Technical Neural Nets

Instructors: Dr. Florian Mai

A(G)I Ethics

Seminar | MA-INF 4116 | Summer 2025 | 4 ECTS

A seminar introducing students to both philosophical and technical aspects of artificial general intelligence, covering AGI basics, alignment and value specification, control and autonomy, systemic risks, and policy governance. Students develop skills in assessing AI systems and reasoning through ethical issues.

Learning Outcomes:

• Assess AI systems and identify ethical dilemmas
• Understand alignment and value specification challenges
• Analyze systemic risks from advanced AI
• Evaluate policy approaches for AI governance

Prerequisites: Basic computer science background; ML/robotics experience helpful but not required

Instructors: Dr. Florian Mai

Schedule: Wednesday 2:15-3:45 PM, Room B-IT 2.113

Large Language Models

Seminar | MA-INF 4332 | Winter 2024/2025 | 4 ECTS

A comprehensive seminar exploring cutting-edge research in large language models, covering architectures, training methods, capabilities, and applications. Students engage with recent research papers and present findings on topics including model scaling, alignment, reasoning, and societal impacts.

Learning Outcomes:

• Understand state-of-the-art LLM architectures and training methods
• Analyze recent research developments in the field
• Evaluate capabilities and limitations of current models
• Assess societal and ethical implications of LLMs

Prerequisites: Basic knowledge of machine learning and natural language processing

Instructors: Dr. Florian Mai