The State of Salzburg is funding four innovative AI projects with a total of €2.5 million

One of these is the project RAPID (Reliable AI for Public Law and Intelligent Decision-making), a cross-faculty collaboration between legal and digital sciences and, not least, another outcome of the EXDIGIT initiative funded by the State of Salzburg.

As part of the Science and Innovation Strategy (WISS) 2030, the State of Salzburg launched the call “AI Research Salzburg 2025,” setting targeted priorities in the field of digital transformation. The call was aimed at researchers from Salzburg universities and non-university research institutions who wish to carry out interdisciplinary projects at the interface between basic and applied research.

The submitted projects were evaluated by an international jury, and three projects with researchers from the DAS Faculty were successful:

• CANVAS: Craquelure Analysis via Neural Vision and Surface Scanning

  Project lead: Christoph von Hagke, Paris Lodron University of Salzburg (PLUS)

  Project partner: Andreas Uhl, Paris Lodron University of Salzburg (PLUS)

• INSPIRE: Intelligent Novel Support for Personalized Instruction and Robust Evaluation in STEM Lessons at Primary School

  Project lead: Christina Egger, University of Education Salzburg (PH Salzburg)

  Project partner: Simon Hirländer, Paris Lodron University of Salzburg (PLUS)

• RAPID: Reliable AI for Public Law and Intelligent Decision-making

  Project lead: Sebastian Krempelmeier, Paris Lodron University of Salzburg (PLUS)

  Project partner: Frank Pallas, Paris Lodron University of Salzburg (PLUS)

We are particularly pleased that, with Prof. Pallas as the project partner of RAPID, a member of the Department of Computer Science is also part of this success.

RAPID investigates the potential of artificial intelligence (AI) in public administration.

Using concrete application examples, the project analyzes the legal framework and technical possibilities for deploying large language models (LLMs) in administration. The focus is on automating repetitive tasks, e.g., checking the completeness of applications for heating cost subsidies or housing allowances. Various privacy-preserving LLMs—with their respective strengths and weaknesses—are examined and evaluated in an interdisciplinary manner.

On the DAS Faculty side, under the leadership of Prof. Frank Pallas, extensive experimental and prototype-based studies are being conducted on administration-specific LLM adaptation that also satisfies the particular legal requirements prevalent in this field.

Image credits: Innovation Salzburg/Benedikt Schemmer

Translated with ChatGPT.