The Best Scientific full Paper prize was awarded to ZHAW researchers for their paper “Towards the Certification of AI-based Systems”. Two of the authors, Philipp Denzel and Stefan Brunner presented in a talk their ongoing effort to develop a validated AI certification scheme. The certification scheme is a framework for assessing the trustworthiness of AI systems. It combines legal requirements and validated techniques for assessing the compliance of AI systems together. The result is the implementation of a workflow to support AI certification.

“My highlights at the conference were definitely the surprise of the best paper award for me and my co-authors,” says Philipp Denzel after the conference. But there were also other talks, that left an impression on Denzel, such as Thomas Brunschwiler’s talk on AI for mitigating the risks of climate change, or Lukas Tuggener’s talk on efficient GPTs for private Large Language Models (LLMs). “One talk emphasises that AI can really be a research tool to do good - which is becoming more and more urgent now, especially when it comes to the climate; the other talk emphasizes how much research has been done to enable AI for private use. Both are arguments that we are not necessarily in competition with the big companies like Google, Meta, or Microsoft. But above all, the most important thing was the exciting discussions during the breaks with like-minded people I met for the first time and with old friends I met again,” says Denzel.

Next to Denzel, Brunner and Tuggener, many other ZHAW researchers participated in the 11th IEEE SDS with presentations and workshops on various current data science topics. For instance, the talk by Mark Cieliebak from the Center of Artificial Intelligence focused on the reliable evaluation of Large Language Models (LLMs). In the workshops, ZHAW expert Jürg Meierhofer showed how business can be elevated through social values. Ricardo Chavarriaga and Christoph Heitz on the other hand demonstrated the implementation of standards towards a responsible AI. Other talks by ZHAW experts focused on group fairness (Christoph Heitz) or anomaly detection (Lilach Goren Huber). A summary of some of the talks can be found in this blog.

As the official scientific partner, the ZHAW Datalab was also present with a booth, making its commitment to the conference visible to the outside world: “For years, the ZHAW Datalab has been heavily involved behind the scenes in expanding the scientific appeal of the conference, for example through cooperation with the IEEE or involvement in the scientific board that puts together the scientific programme,” says Thilo Stadelmann, founding member of the Datalab.

More information:

• Blog about the SDS2024
• Paper “Towards the Certification of AI-based Systems” by Philipp Denzel, Stefan Brunner, Yann Billeter, Oliver Forster, Carmen Mei-Ling Frischknecht-Gruber, Monika Reif, Frank-Peter Schilling, Amin Amini, Marco Repetto, Arman Iranfar, Joanna Weng, and Ricardo Chavarriaga.