After a successful start in 2018, the sixth after-work seminar under the motto ‘Knowledge in the Evening’, which was organised by the Centre for Continuing Education and Knowledge Management (ZWW) at OTH Regensburg on 23 May 2025, was once again very well received. Around 60 participants gathered information and discussed the topic ‘Behind the scenes of ChatGPT & Co.’. Prof Dr Timo Baumann, Faculty of Computer Science and Mathematics at OTH Regensburg, and Dr Manuel Kirschner, BSH Haugeräte GmbH, were invited to give a keynote speech and report from the field. The annual after-work seminar is intended to provide a platform for brief and compact information and dialogue on a current topic.
The evening began with Prof. Dr Rösel, Vice President of OTH Regensburg and Head of the Centre for Continuing Education (ZWW), revealing in her welcome address that she had used AI to write the welcoming remarks. According to Prof Dr Rösel, the topic is something that now occupies us both in a professional and private context.
Modern chatbots demystified
Prof Dr Baumann then allowed the participants to delve deeper into the ‘AI world’ and prepared them for a ‘demystification of modern chatbots’. We are fascinated by artificial intelligence, but is it really intelligent and if so, to what extent?
Modern chatbots are based on a Large Language Model (LLM) and Prof Dr Baumann began by giving an overview of this, including the history: the N-grams, the first language models over a hundred years ago. Technical refinements - including the use of neural networks - and scaling to very large amounts of data finally led to the qualitative breakthrough of the technology from around 2023, which enabled its use in AI chatbots such as ChatGPT, Microsoft Copilot or Google Gemini. With a so-called Large Language Model (LLM) as the core and Retrieval Augmented Generation (RAG) and Function Calling as additions, it was now possible to conduct natural dialogues with the chatbots.
Prof Dr Baumann is currently carrying out a project with Master's students at OTH Regensburg: An AI chatbot should be able to answer any questions about studying or the IT infrastructure and, among other things, access the existing support wiki and module manuals using RAG.
Generative artificial intelligence with Langflow at BSH
Dr Kirschner then provided interesting insights into the use of AI at BSH, the largest manufacturer of household appliances in Europe with almost 60,000 employees worldwide and the headquarters for global electronics and motor development in Regensburg, with his practical report. He presented several case studies of how generative AI is used in the company, focussing in particular on so-called RAG flows. ‘AIM’, a company-internal chatbot, was designed specifically for BSH and serves as an internal BSH knowledge source, similar to the ongoing student project by Prof Dr Baumann. Dr Kirschner presented the ‘RAISE’ project for improving internal software development processes and the ‘BAIA’ project, a voice dialogue system for household appliances, as further case studies for the application of generative AI at BSH. He vividly demonstrated this using a video in which a washing machine could be operated using voice control and answered the user's questions about stain removal, for example.
The speaker then showed the practical structure and application of an LLM- and RAG-based chatbot based on the open source software ‘LangFlow’ to create a natural-language user interface for internal knowledge sources with minimal effort, which can also be operated completely locally and thus data-secured if desired.
The many questions from the audience that were asked during and after the presentations showed how stimulating the topic was. The discussions continued into the late evening at the subsequent get-together.
The ZWW will also include the topic of AI more comprehensively in the further education programme at OTH Regensburg in the future. For example, a part-time university certificate in AI is already planned for September 2025. Further information can be found atwww.oth-regensburg.de/weiterbilden.
