Pour Demain and the Franxini project of Reatch present a proposal for monitoring artificial intelligence (AI) in Switzerland. The whitepaper was developed during the 2022 summer session in a workshop that facilitated open exchange between members of the National Council and representatives from science, administration and economy.
Applications of artificial intelligence (AI) are being used in more and more areas of life and will have a massive impact on society in the coming decades. On the one hand, they have the potential to significantly enhance the common good, for instance by improving medical diagnoses. On the other hand, AI systems can cause material and/or immaterial damage to property or people.
The Federal Government can better understand the opportunities and risks of AI by directly following developments in the AI sector. To enable this, we recommend rapidly launching AI monitoring pilot projects to systematically capture the current capabilities and impacts of AI applications, thereby ensuring a timely and relevant evidence base for policy decisions.
The AI monitoring should achieve four fundamental goals:
Identify opportunities: Areas which can benefit from future AI applications become apparent.
Classify risks: Security risks of AI applications become detectable.
Create transparency: It is revealed which everyday decisions are made by AI systems and not by people.
Create a basis for decision-making: Society and politics receive a better basis for deciding on the use of AI.
The specific recommendations for action are:
Step-by-step approach via pilot projects to assess and optimise the concrete benefits of AI for politics and society on the basis of initial results.
Pilot project 1 "Critical infrastructures": An initial pilot project should be implemented in a safety-critical area such as healthcare, since the common good is particularly affected here.
Pilot project 2 "Meta-monitoring": This project clarifies in which areas future monitoring is particularly necessary and which methods are suitable for this.
For the successful implementation of AI monitoring, a clear distribution of roles among the actors is crucial:
Administration: It should receive sovereignty over monitoring projects to ensure that monitoring results are directly relevant to the political decision-making process.
Science: Due to their expertise in AI, Swiss universities can implement monitoring projects effectively and interdisciplinarily on behalf of the administration.
Economy: It benefits from monitoring through increased demand for AI applications in areas that have expanded application potential. At the same time, it is made more aware of risks, in a constellation similar to how the National Center for Cybersecurity raises awareness for cybersecurity risks.