Understanding LLMs: Between Fancy Autocomplete and AGI
14:00-15:30
This is an introduction to LLMs, examining different perspectives on them and the critical problems they pose.
Work in Progress: This page and all linked
resources are still under development. Content may change, and there
might be errors or broken links. If you spot something that needs
fixing or have any feedback, please write me – I'd really appreciate
it! :)
christopher.pollin@dhcraft.org
1
Large Language Models Fundamentals
Day 1
Large Language Models Fundamentals
16:00-17:30
A "non-machine-learning" introduction to the core concepts, architecture and training principles of LLMs.
2
Prompt Engineering & AI Engineering
Day 2
Prompt Engineering Basics
09:00-10:30
From "Raindancing" Prompts to Context Engineering
AI Engineering & Applied Generative AI
11:00-12:30
From Demos to Workflows (Tools, Techniques and Agents)
Advanced Prompting Techniques
14:00-15:30
PRISM, Kaskade and other advanced prompting techniques
Hands-on Session
16:00-17:30
Practical prompt engineering for DH research workflows
3
LLM-Supported Digital Editions
Workshop track
Day 3 & 4
LLM-Supported Modeling, Operationalization, and Exploration for Digital Editions
Day 3 09:00-15:30
Comprehensive workshop covering LLM applications for digital editions: TEI modeling, extraction and annotation workflows, and enrichment processes for research data
4
Promptotyping
Note: This section was not part of the original summer school curriculum but has been added as supplementary material
Additional Resources
Introduction to Promptotyping
Self-paced learning
Extremely rapid, researcher-centric, research-data-driven prototype creation of research tools, workflows, and models using frontier LLMs.
Resources
- Summer School: Access Course Materials
- Google Drive Folder (Materials): Access Course Materials
- Zotero: www.zotero.org/groups/5319178/agki-dh
- DHd-AG Angewandte Generative KI in den Digitalen Geisteswissenschaften (AGKI-DH): agki-dh.github.io
- YouTube Playlist: Applied Generative AI: View Playlist
Additional Information
Prerequisites
-
Access to a Frontier LLM (Required)
- Recommended: Claude 4 Sonnet (Claude Pro)
- Alternatives: GPT-5, Gemini 2.5 Pro, Claude Sonnet / Opus 4.1
- Note: No free/open-source models match these capabilities
- Cost: ~$20-25/month subscription or $5-10 API credits
- Python experience (optional but helpful)
What to Bring
- Laptop with modern web browser
- Your research questions and use cases
- Sample texts or datasets (optional)
- Active LLM subscription or API key or local LLM
Workshop Notes
- On proprietary models: Apologies for the Big Tech dependency, but experiencing frontier capabilities is essential for understanding the field
- Materials: Currently hosted on Google Docs/Slides for easy collaboration. After the summer school, polished versions will be archived on Zenodo as PDFs