Participation in the UniBW AI Retreat 2024

Public sector researchers and junior researchers together with PhD students from the UniBw gathered at the UniBw AI Retreat 2024 in Garmisch-Partenkirchen. The event was held at the George C. Marshall European Center for Security Studies and organized by the Institute of Distributed Intelligent Systems and the dtec.bw Innovation Center in close cooperation with the Marshall Center. 
The retreat aimed to build an AI network and to offer a platform for technical and applied discussions in artificial intelligence and security studies.

During the retreat, participants had the chance to attend a range of activities, including technical lectures, poster sessions, and interactive workshops. Senior experts like Dr. Sae Schatz and Professor Dr. Scott Handler presented their insights on the challenges and opportunities of AI in global security. Their talks helped to spark new ideas among all the participants.

A key part of the event was the poster sessions. The six junior researchers from the CCEW, Tobias Bohne, Daniel Mittermaier, Stephanie Riedl, Sana Slouma, Clara Clipea, and Laura Braun, presented their research on using data science and machine learning for crisis early warning.
Their projects covered the following topics:

• Forests of UncertainT(r)ees: Using tree-based ensembles to estimate probability distributions of future conflict 
• Bridging Bayesian Models and GNNs for Conflict Forecasting
• Extracting Data with the help of large lanugage models: An Attempt to enhance data on kidnappings
• NLP-enhanced ethnic conflict dataset: A proof of the concept for nigeria
• Introducing a new dataset for conflict damage prediction

These presentations allowed for valuable feedback from the other researchers and participants. The discussions helped refine the research ideas and methods, showing the practical benefits of exchanging knowledge in a supportive environment.

In addition to the poster sessions, the retreat featured practical learning opportunities. An advanced PyTorch tutorial led by Dr. Thomas Viehmann provided hands-on experience with model optimization, training strategies, and exporting techniques.
Another highlight was the Open Science workshop, where participants learned about best practices for research organization and reproducibility. These sessions were especially beneficial for young researchers, giving a practical skills that will help improve their future projects.

Networking with each other was another important aspect of the retreat. Researchers from various departments and research projects, including those from the Bundeswehr University Munich and from the Hamburg Bundeswehr University, had ample opportunities to meet and discuss their work. The event encouraged the building of new contacts and the strengthening of existing collaborations. In a relaxed atmosphere after the daily schedule, participants also enjoyed a walk around the beautiful Eibsee near the Zugspitze. This informal activity not only offered a break from the sessions but also helped deepen team spirit among the CCEW team and other colleagues.

Overall, the UniBw AI Retreat 2024 was a well-balanced mix of academic exchange, practical workshops, and social interaction. The event provided participants with new impulses for their research and a better understanding of how artificial intelligence can be applied to security and crisis early warning. The insights gained from technical sessions and discussions will support the development of future projects at the Center for Crisis Early Warning.

Picture: © Mihaela Lechner, UniBw M