【GLA EVENTS】Colloquium meeting of Spring 2026 on Tuesday, June 2, 16:40-18:15 in AN524

GLA is inviting you to the second colloquium meeting of Spring 2026 on Tuesday, June 2, 16:40-18:15 in AN524. We will be hosting Prof. Anna Kruspe (Munich University of Applied Sciences) for her talk, “Peeking Behind the Curtain of Large Language Models.”

Please see the information below for the abstract of the talk and Prof. Kruspe’s bio. Large language models are currently one of the most consequential technologies, and this is a rare opportunity to learn about this important topic from an established expert.

We hope to see as many of you as possible at the event!

In today’s digital landscape, Large Language Models (LLMs) like ChatGPT often appear as "black boxes"—tools that seem to possess human-like intelligence and an answer for everything. But what is actually happening beneath the surface? Is there a mind at work, or is it just math? This lecture "peeks behind the curtain" to demystify the technology for participants without a technical or computer science background. We will break down the complexity into three accessible parts:

- From Words to Numbers: An introduction to how machines "read" by turning human language into mathematical data.

- The Next-Word Engine: Exploring the core mechanism of LLMs—predicting the most likely next word based on vast patterns rather than true "understanding."

- Capabilities vs. Hallucinations: Understanding why these models can be brilliant at creative tasks but can also confidently provide false information.

The goal of this session is to replace the "magic" with an intuitive understanding. By the end, participants will be better equipped to use these tools critically and effectively in both their academic and professional lives.

Prof. Anna Kruspe received her diploma and Ph.D. degrees in media technology from Technische Universität Ilmenau, Germany, in 2011 and 2017, respectively. She is a machine-learning researcher at the Munich University of Applied Sciences and the German Aerospace Center. Previously, she was a member of the Fraunhofer Institute for Digital Media Technology, Ilmenau, Germany, where her work focused on the application of speech recognition technologies to singing (e.g., for language identification, keyword spotting, or lyrics-based search), as well as the analysis of world music. She conducted research at Johns Hopkins University in Baltimore, Maryland, and at the National Institute of Advanced Industrial Science and Technology in Tsukuba, Japan. Her current work deals with the development of machine-learning technologies for the analysis of social media data in the context of disaster management, biases in large language models, and AI models for text and audio.