The Knowledge Centre Data & Society has kicked off their DataDate event series in the VUB AI Experience Centre this week.
The event series ‘Data-Date’ invites external speakers to spark discussions on AI. Their very first event already presented a hot topic; the White Paper of the European Commission ’On Artificial Intelligence – A European approach to excellence and trust’. The panelists debating were Nathalie Dumarey (VVSG), Astrid Eichstädt (Unia), Simon November (Test Purchase), Jelle Hoedemaekers (Agoria) and Dieter Somers (Voka). They responded to the presentation and shared their critical reservations on the White Paper.
- The panel members welcome the intention to create a European discussion regarding the definition of AI. Still, they do so whilst raising some concerns. Some stakeholders may struggle to provide knowledgeable input, partly due to a lack of knowledge and skills. The question is then, how such a broad consultation is to proceed and if it can really offer greater stakeholder involvement.
- Should there be an authority or a regulator to take care of a framework and appropriate legislation on AI? The panel members didn’t articulate a united answer. It is clear and also crucial that – when such an authority arises – it must have the knowledge and the skills to evaluate AI applications and intervene if necessary. The Knowledge Centre Data & Society can contribute to this by identifying problems in current regulations so that our governments can adjust them.
- Data ownership and who should have access to data were also discussed. These are important elements which should also be noted in the White Paper. These elements are already cited in the European Data Strategy.
- A key element in the White Paper is the distinction between high-risk sectors and applications, and others. According to the panel members, it is challenging to make the distinctions. It is difficult to estimate in advance whether or not a sector or an application has a high risk. There may be some sectors and applications where a high risk can be estimated from the start (e.g.: matters related to safety or health care), but there may also be a high risk in other sectors and applications. The panel wondered what will be the precise parameters that determine whether a sector or an application is of high risk?
- Can certifications offer a possible solution here? At the moment, there are still insufficient standards established for AI which means that certification is not yet possible. A possible solution is to start from a current demand from a group of similar customers or providers. A good example of this is the standard clauses for local authorities with regard to the procurement of AI. This approach is currently being developed and tested in the Netherlands.
The panel agreed that the White Paper on AI sets a number of clear targets for Europe. The document was positively received and it is considered as a conversation starter to further elaborate the policy. However, the White Paper does not yet explain how we can achieve these targets, and therefore, a number of difficult political choices have not yet been made. It is time to consider which steps we want to take further from Flanders.