Artificial Intelligence at the University of Potsdam

Current Status

The massive impact of AI technologies on educational institutions became clear, if not before, with the global release of ChatGPT in November 2022. The UP began exploring the various facets of the topic “AI in Higher Education” relatively early on. Under the coordination of the VPL division and the ZfQ, an initial position paper on “AI in Higher Education Teaching” was drafted as early as 2023, and a second version has been available since October 2024. In conjunction with this, the ZfQ and PoGS have developed training programs for faculty and researchers, which have been continuously refined ever since.

Since ChatGPT and similar commercial AI technologies—which are supposedly offered “for free” on the internet—do not meet the data protection and examination regulations of German universities, a task force “Campus AI License” was temporarily tasked with evaluating various commercial products and open-source solutions for university-wide use from educational, legal, and technical perspectives. Based on the evaluation results, the first AI chat assistant, GPT.UP, was launched by ZIM in October 2024.

Technically, GPT.UP is based on the open-source AI technology HAWKI, which allows for the integration of both commercial and open-source language models. For testing purposes, HAWKI is being used at the UP to integrate both the “open-source” language models hosted by the GWDG and the commercial language models from OpenAI (e.g., GPT-4o) provided by Microsoft in Europe in compliance with data protection regulations, which are currently funded by MWFK grants. In addition, a user guide and basic informational resources have been created and made available on the GPT.UP website. Despite a lack of AI personnel capacity, ZIM is also ensuring the technical monitoring of the ongoing pilot operation for the purposes of system stability and IT security, so that the pilot operation can be evaluated within the framework of the MWFK-funded project. These activities have, in particular, exhausted the staffing capacity of the technical members of the AI Task Force. A project-based approach is therefore necessary for the provision and further development of additional AI tools for the “Study & Teaching” area and to address new needs, e.g., from the “University Administration” sector.

Challenges and Needs

Based on feedback from GPT.UP users, the ZUV’s AI use cases aggregated by D6, the legal requirements of the AI Regulation, and general trends, the following fundamental needs can be identified:

• Functional enhancements to GPT.UP (document uploads, image and video generation, creation of presentation slides, voice input, etc.)
• AI chatbot with access to internal information—e.g., from the Z-Wiki—as well as to local files and emails
• AI-supported meeting minutes, data analysis, and process optimization
• Integration of AI functions into existing line-of-business applications
• Provision of API access for the integration of AI language models into faculty applications
• Systematic teaching of AI skills to students, faculty, and administrative staff (AI literacy)
• Ensuring compliance with the legal requirements of the EU AI Regulation
• Design and establishment of a central AI cluster for research and for operating local language models, e.g., for processing UP-internal information
• Monitoring global AI developments and maintaining regular exchange with other universities

Project Planning

As part of ongoing development efforts, several departments at UP have been addressing the cross-cutting topic of AI since 2023. Notable among these are the units already involved (the Legal Affairs Office, the VPL Division, ZfQ, and ZIM-BA), as well as D6-L, ZIM-L, and the CIO/VPD Division. To future-proof the UP in light of dynamic AI developments and to address the diverse needs arising from the areas of “Academic Studies & Teaching” and “Administration” in the coming years, the topic of “AI for Academic Studies, Teaching, and Administration” is therefore being addressed as part of the “AI@UP” project.