Hardly a day goes by without a new application based on artificial intelligence methods (AI) appearing in our news. ChatGPT interacts with us in the form of a dialogue, and Midjourney generates images based on our description of pictures, to name just two examples. These technical achievements are often accompanied by questions about the risks involved. For example, do we want an AI system that carries out biometric facial recognition in the public sphere on a real-time basis? Or one that creates a social assessment of us based on personality-related characteristics? The EU is proposing a new AI law which will prohibit such practices and aims to regulate AI according to risk. It assumes four levels of risk: unacceptable risk, high risk, limited risk and minimal risk. In cooperation with the Swiss Drone and Robotics Centre (SDRC) of armasuisse Science and Technology (S+T), the University of St.Gallen organised a unique competition in which the law was subjected to a “stress test” to investigate its applicability.

 “The First University of St. Gallen Grand Challenge” and the SDRC DDPS

The competition was launched by Professor Dr. Thomas Burri from the University of St. Gallen. He was inspired by the robotics competitions of the American defence research institute DARPA. “I was fascinated by this energy and wanted to adapt the concept for my research in the legal field”, says Professor Burri. He was able to win the SDRC DDPS as a strategic partner for this idea. Pascal Vörös, who manages the social science research projects in the SDRC DDPS: “In the Grand Challenge, we want to put the implementation of the AI law to the acid test using practical application cases. We have various partners from industry, in particular start-ups, who are trying to grow in the European market. As they use AI in their robotic systems, they will be confronted with the new EU regulations. By participating in the Grand Challenge, they haven’t just supported legal research. They will also benefit from a broad spectrum of expert feedback which will help them to be competitive in the European market. In addition, we are building competencies to compare the EU’s AI law with other methods for regulating AI.

Strict selection process and a test run in the boot camp

Of the more than 30 teams who applied in March this year, twelve teams from ten different countries were selected according to the criteria of competence, credibility and diversity. At the same time, in Swiss-Mile, Gravis Robotics and Ascento the SDRC DDPS was able to win three of its research partners, who answered the teams’ questions and provided deep insights into their use of AI. As a test run for the grand final, a “boot camp” was integrated into the field robotics event ARCHE on 14 July, where the teams were able to gain valuable practical experience with autonomous systems.

Assessments within a very short time and two winners

On 18 July, the teams arrived at the University of St. Gallen to compete in the “Grand Challenge Final”. Four AI systems had to be assessed. An international jury selected the two best teams for the second round. Here, they had to work on another AI application case. The overall performance of both teams was assessed as equal by the jury, so that the prize money of CHF 100,000 was split up equally with CHF 50,000 going to each team, one from Spain and one from South Africa, Canada and European countries.