A country aims for more artificial intelligence
What potential does data & AI hold for a state government? Quite a lot – Switzerland wants to play a key role in the global digital transformation. A big task for Bertrand Loison, Vice Director and Head of the Data Science and Statistical Methods Division (DSSM) for the Swiss Confederation. In an interview, he talks about current developments, major goals and ethical concerns in the use of artificial intelligence.
Oliver Bosse spoke with Bertrand Loison
The Federal Council has identified artificial intelligence (AI) as "a central theme of the 'Digital Switzerland' strategy". What does this mean exactly?
Artificial intelligence holds immense potential for society as a whole. It is imperative that this potential is also properly and fully exploited in Switzerland. The goal is for Switzerland (both as a country and as a business location) to play a key role in the global digital transformation. For this to succeed, all actors in society are called upon to play their part.
As head of the Data Science and Statistical Methods Division at the Federal Statistical Office (FSO), you deal with digital transformation on a daily basis. Where in particular do you see potential for optimisation through AI in the federal administration?
The potential at the federal level is huge. For this reason, three years ago the Federal Council commissioned the Interdepartmental Working Group Artificial Intelligence (IDWG AI) to examine the establishment of a competence network for artificial intelligence (CNAI). In this context, we conducted a survey within the Federal Administration at the beginning of 2020. It turned out that many employees have fears and doubts about the use of AI in the federal administration. This was also one of the reasons why the rapid establishment of a network such as the CNAI met with great approval.
You are currently in charge of setting up the CNAI, which is scheduled to become operational in the summer of 2022. What tasks will the competence network perform in the future?
The task of this network should be, on the one hand, to quickly and sustainably promote confidence in artificial intelligence as well as the use of this technology within the administration and beyond. On the other hand, the CNAI is to systematise, bundle and pass on the national and international knowledge that is necessary for the successful realisation of AI projects. On top of that, it is to be the point of contact for internal and external networking at the federal level. The CNAI serves as an enabler (person, unit or subject that makes something possible) and a facilitator (person, unit or subject that makes an action or process easier or simpler).
The goal is for Switzerland (both as a country and as a business location) to play a key role in the global digital transformation. For this to succeed, all actors in society are called upon to play their part.
As the CNAI project database shows, several pilot projects are already underway. Where exactly is artificial intelligence already being used by the federal government today?
Currently, the majority of AI applications in the federal administration are still in the early stages (studies and prototypes) and only very few AI applications are already in productive operation. Examples of promising pilot projects can be found in several departments. The following topics are particularly relevant for the federal administration:
- Text Recognition
- Speech Recognition
- Image Recognition
- Fraud Detection
- Plausibility Checks and Data Validation
- Predictive Maintenance
Here at the Federal Statistical Office, the use of AI offers potential for automating and streamlining tasks of a repetitive nature. This CNAI project database provides an overview of AI-relevant projects in the Federal Administration and creates transparency about the AI projects available in the Federal Administration.
What is your conclusion on these projects so far? Is the use of AI proving successful?
As I said, the use of AI is still in its infancy in our country and there is still a lot to learn. I personally see four lessons that we can learn from so far:
- Training: Staff training is critical in the areas of data science and AI. That's why, in collaboration with the EPFL Extension School, we have developed a training programme (including learning paths) that enables all roles in a federal office to train in these topics as needed.
- Governance: AI is characterised by a professional, a technical and an algorithmic component, which is a challenge in terms of governance that should not be underestimated. This governance still needs to be established in the various federal offices, especially among the heads of the legal services, the IT officers, the data protection officers and the directorates of the federal offices.
- Data quality: The old adage "Garbage In, Garbage Out" is still relevant and must not be forgotten under any circumstances. Data science and AI require data management and upstream data quality control that are particularly powerful on the one hand and cross-cutting within the federal administration on the other.
- Ethics: The development and use of algorithms that can make decisions inevitably leads to key questions of ethics, legal foundations and data protection. These questions must be addressed in a professional and pragmatic manner in the context of public policy. The functioning of the rule of law is always at the centre – also in the use of AI. If adjustments to the legal basis are necessary, they can only be made in this context.
Keyword "functioning of the rule of law": You also thought in advance about the possible challenges and risks of using AI. What findings did you come to and what experiences have you had in this regard?
The most important insight is reflected in our guidelines for dealing with AI. With every technology we use, it must be possible to answer the questions of transparency, reproducibility, traceability and ethics at any time. It must always be clear how an AI system arrived at its results. This question ultimately determines whether a project should go into production or not. After all, comprehensibility plays a major role, especially in the public sector, since it must guarantee the rule of law and always be able to justify the decisions it makes. If we do not understand how an algorithm arrived at its results, we can no longer guarantee the rule of law. This is a risk that must be avoided at all costs. For this reason, every AI project that is initiated in the federal administration is carefully examined with regard to the above-mentioned questions.
If we do not understand how an algorithm arrived at its results, we can no longer guarantee the rule of law. This is a risk that must be avoided at all costs.
The guidelines for dealing with AI also clearly define the goal that Switzerland wants to develop into a leading location for research and application as well as companies in the field of AI. How are you proceeding with this? What framework conditions and measures are needed to achieve this?
In order to reach this goal, we need to map out and plan the way to get there as precisely as possible. In other words, we need – as with a hike in unknown territory – a precise map that shows everyone involved the same path so that we can reach our destination safely. Two strategies form the basis for this "hiking map": the "Digital Switzerland" strategy and the "Digital Foreign Policy 2021-2024" strategy. The former serves as an umbrella strategy and provides the guidelines for government action in the area of digitisation. The latter defines the conceptual foundations for helping to shape international governance in the area of digitisation.
Building on this, we drew up the guidelines for the use of AI in the federal administration, which now provide precisely this "hiking map" – a signpost, a general orientation framework for the federal administration as well as the bearers of administrative tasks. In addition, the Data Science Competence Centre (DSCC) also acts, among other things, as a neutral authority for public administrations. In this role, the DSCC provides a scientific judgement on the quality of the data and the algorithms used in a project. The aim is to avoid biases that arise from the use of automated data processing algorithms.
At the global level, Switzerland wants to play an active role in shaping standards and norms for artificial intelligence. In what framework will this take place and what is Switzerland advocating?
Switzerland is actively involved in the development of an international regulatory framework for AI – together with the OECD, the Council of Europe, UNESCO and the International Telecommunication Union (ITU). The Confederation is therefore also close to the discussion on the planned AI regulation in the EU and is analysing the potential impact on Switzerland. In this context, for example, in November last year the Federal Department of the Environment, Transport, Energy and Communications (DETEC) showed the Federal Council (FR) how AI-supported communication platforms affect public communication in Switzerland and the formation of opinion among the Swiss population. As a result, the FR mandated DETEC to show by the end of 2022 whether and how communication platforms should be regulated.
In addition, the Platform Tripartite Suisse, led by the Federal Office of Communications (OFCOM), serves as an open exchange body for all stakeholders on AI issues and has a management committee that coordinates Swiss positions in international bodies and processes.
Are stakeholder groups such as the Swiss business community also involved? To what extent?
Absolutely! As already mentioned at the beginning, digitisation affects all actors in society and therefore everyone should have the opportunity to contribute to this topic. One of the main tasks of the CNAI is the networking of experts, project leaders, data scientists, etc., in order to find solutions for AI-based problems together in interdisciplinary teams. For this reason, we are also constantly looking for interested stakeholders who are willing to actively participate and share their AI experiences. Interested individuals, institutions, organisations or companies can become part of the "Community of Practice". Proven AI experts can also apply for the "Community of Expertise".
One of the main tasks of the CNAI is the networking of experts, project leaders, data scientists, etc., in order to find solutions for AI-based problems together in interdisciplinary teams. For this reason, we are also constantly looking for interested stakeholders who are willing to actively participate and share their AI experiences.
What is your personal goal that you would like to achieve within the framework of the "Competence Network for Artificial Intelligence"?
I have two personal goals: On the one hand, the CNAI should succeed in creating this "hub" that makes it possible to bundle the existing competencies within the federal administration in Switzerland in the area of AI and actively contribute to Switzerland's digital strategy. On the other hand, we naturally also have financial and personnel goals for the CNAI that we want to achieve – so that we can fulfill the task of this network and thus further advance digitalisation in Switzerland.
About Bertrand Loison
Dr Bertrand Loison is Vice Director and Head of the Data Science and Statistical Methods Division (DSSM). In addition to the Data Science Competence Centre of the Confederation (DSCC), which provides data science services within the administration – in the sense of "Data Science as a Service" (DSaaS) – this also includes the Competence Centre of the Federal Statistical Office for statistical methods.