Dr. Sven Weinzierl is a postdoctoral researcher at the Friedrich-Alexander-Universität (FAU) Erlangen-Nürnberg. He studied business informatics at the Otto-Friedrich-University Bamberg and received his doctoral degree from FAU Erlangen-Nürnberg in 2022. Since the beginning of his doctorate, he has been working on the development of machine learning-based solutions for companies. His research results have been published in internationally recognized journals such as European Journal of Operational Research, Decision Support Systems, and Business & Information Systems Engineering. His research has been funded by programs such as FAUnext, ETI, and Software Campus.
Explaining Internal Mechanisms of Large Language Models: Methodical Design, Empirical Evaluation and Implications for Society
Large Language Models (LLMs), such as GPT, have shown impressive capabilities in algorithmic language processing and are increasingly becoming an integral part of our society. Despite the hype surrounding LLMs, their internal mechanisms are largely opaque. This lack of transparency poses undesirable risks for downstream applications such as ChatGPT. For example, it is not possible to understand in which situations applications generate fictitious content, i.e., hallucinate. The research project therefore aims to design and evaluate a method for explaining the internal mechanisms of LLMs and to derive implications for society.