Learning Outcomes:
By the end of the course, students will be able to:
1. Understand the fundamentals of large language models and their applications in the social sciences.
2. Implement LLMs for various tasks relevant to political and social science research using Python
3. Critically evaluate the ethical and societal implications of AI technologies, particularly LLMs, in the context of social science
4. Apply best practices in research documentation, reproducibility, and validation when using AI tools.
5. Engage in informed discussions about the impact of AI on society, including issues of bias, sustainability, and concentration of power.
Indicative Module Content:
- Using Python and APIs for programming tasks with the help of AI tools (GitHub Copilot)
- Effective use of development environments (VSCode, Jupyter Notebook) and version control systems (Git)
- Fundamentals of LLMs, pretraining, fine-tuning, and transformer-based models
- LLM applications, including classification, transcription, and topic modelling
- Research best practices: reproducibility, validation, and documentation
- Ethical implications: bias, misinformation, environmental impact, training data ownership, and copyright in AI-generated content
- An introduction to advanced techniques, which may include quantisation, parameter-efficient fine-tuning (PEFT), retrieval-augmented generation (RAG) and multimodal models
- Future challenges and regulation