Course language: English
Location: TUM (in presence)
Offered in summer semester 2024, winter semester 2024–5
Description
The seminar explores philosophical questions concerning contemporary Generative Artificial Intelligence (GenAI), and in particular Large Language Models (LLMs). Although the core mechanisms of LLMs are generally well understood from a technical point of view, it is little understood why they can generate novel text that is relevant and meaningful to humans. The secret lies in their training and their training data, and in particular the patterns of human language use in the text corpora on which LLMs are trained. We discuss
(1) what LLMs extract from text corpora and
(2) the relationship between language and in particular writing to meaning.
Text corpora reflect a specific form of human language use, namely written text, and writing itself is a transformative technology. The profound changes that writing has brought to memory, thinking, communicating, and philosophical understanding are often overlooked today, but they were already problematized by Plato in his—writing. The generative nature and interactive uses of LLMs seem to counter some of the problems of writing, but do LLMs really communicate? In how far do they revert to features of oral communication such as contextualization and interaction, and to what extent do they initiate new language games? What does AI-generated text reveal about human patterns of thought, language, and interaction? How does it change them? Are superficiality, “hallucinations,” “bullshitting,” bias, and clichés preventable malfunctions, or are they part and parcel of LLM output?
Upon successful completion of this seminar, students will be able to:
- analyze philosophical texts with respect to their assumptions and arguments
- present and critically discuss philosophical claims
- assess the possibilities and limits of the generation of meaningful text with LLMs
- understand the general relation between writing, communication, and sense-making
Teaching and learning method
- text analysis and interpretation
- collaboration
- presentations by lecturer and students
- problem-oriented discussion