Session 1: Introduction to course idea¶
Date: 2026-03-16
Plan¶
- We introduce the course idea and structure, and get to know each other.
- We discuss the course topics and the students' expectations and goals for the course.
- We have a look at some examples of generative AI applications and discuss the potential and challenges of these technologies.
Schedule¶
| Activity | Duration | Ressources |
|---|---|---|
| Introduction and getting to know each other | 10 min | - |
| Course overview and structure | 20 min | Slides: KIM-01-Intro |
| What is generative AI? | 15 min | KIM-01-Tasks Task 1 |
| State of the art in generative AI | 30 min | Slides: KIM-02-Examples |
| Clarifying expectations and goals | 30 min | KIM-01-Tasks Task 2 |
| Discussing the mindmaps | 30 min | KIM-01-Tasks Task 3 |
| Presenting key insights and questions | 45 min | KIM-01-Tasks Task 4 |
Homework¶
- Refine your mindmap and hand it in via FELIX until the deadline.
- Peer-review the mindmaps of your classmates in FELIX and reflect on the diversity of knowledge and interests in the class. Consider how your own goals for using generative AI might evolve based on what you learn from others.
Materials¶
- Slides:
- Task description: KIM-01-Tasks
Grading¶
- You find the grading criteria for the seminar under Grading & Deliverables
- You find the grading criteria for the colloquium under Colloquium
Video and audio samples for the course¶
The video above has been generated with Synthesia, a text-to-video AI tool: Syntheisa
https://hkchengrex.com/MMAudio/
Spotify Advertising generator: Spotify Creative Lab
Sources for shown image and video samples¶
- Person image: This Person Does Not Exist
- Midjourney house draft animation: Midjourney Explore
- VEO3 racecar video: Google AI Studio
- Common Sense Machine texture: Blog with other examples, shown example was published only on Discord.
Results¶
KIM-01-Tasks: Mindmap of course topics¶
General notes¶
Check out OpenClaw.
Things students know already¶
- Text generation basics: LLMs based on GPT, transformers, attention, tokenization, ...
- Image generation basics: Diffusion models, GANs, ...
- GenAI tools like: ChatGPT, Sora, veo, DeepSeek, LMStudio, Automatic1111, NotebookLM, ...
- Problems of GenAI: hallucinations, bias, energy consumption, deep fakes, ...
Things students want to learn¶
- LoRA, ComfyUI, ControlNet and other tools for image generation
- How does 3D asset generation work?
- How does audio generation for music, sound effects and speech work?
- Which specialized hardware exists for GenAI and how does it work?
- How can inference times be reduced? How can models be compressed or optimzed for faster inference?
- What is agentic AI and how does it work?
- Emerging AI / Safety of AI systems
- Anthropic Research - Agentic misalignment
- Frontier Models are Capable of In-context Scheming ArXiv, YouTube
Things students want to create¶
- Image generation with recent workflows and tools like ComfyUI, ControlNet, LoRA, ...
- 3D asset generation including normal maps, UV maps, etc.
- Experiments with audio generation for music, sound effects and speech
- A GenAI agent that can perform tasks autonomously (e.g. using OpenClaw)
- Personal GenAI assistant for various tasks including visualization of an avatar