‘How to build trustworthy Artificial Intelligence (AI)’ was an overarching question for the first European AI Alliance Assembly that took place in Brussels on 26th June 2019. The highlights of the event were the launch of the piloting process of Ethics Guidelines for Trustworthy AI and presentation of the Policy and Investment Recommendations for Trustworthy AI. Both documents have been prepared by the High-Level Expert Group on AI. Topics discussed at the Assembly include international governance, legal framework and social and environmental impact of AI.
Piloting AI Ethics Guidelines
An important step towards developing trustworthy AI is a launch of the piloting process for ethics guidelines. The initial draft of these guidelines was published in December 2018 and during the stakeholder consultation more than 500 contributions were received. These were considered in the final version of the guidelines published in April 2019. According to the guidelines, Trustworthy AI should be lawful, ethical and robust and four main ethical principles based on the EU fundamental rights are: respect for human autonomy, prevention of harm, fairness and explicability. The European Commission has expressed support to these guidelines in its communication Building Trust in Human-Centric AI which understands a human-centric approach to AI as one where ‘AI is not an end in itself, but a tool that has to serve people with the ultimate aim of increasing human well-being’.
The piloting process aims to collect feedback in order to improve the guidelines and publish a revised version in 2020. The piloting process includes a questionnaire for all interested stakeholders and in-depth interviews with selected organizations in different sectors. The Human Brain Project is one of over 300 stakeholders that already have expressed interest to participate in the piloting phase. The online survey will be open until 1 December 2019.
Policy & investment recommendations
As the next step towards building trustworthy AI, the High-Level Expert Group presented its 33 policy and investment recommendations, addressed to the EU institutions and member states. These recommendations aim to guide AI towards sustainability, growth and competitiveness and inclusion. They focus on four major areas of impact: humans and society, the private sector, the public sector, and academia and research. To achieve positive impact in these areas, the recommendations highlight the importance of four enablers: data and infrastructure, education and skills, governance and regulatory framework, and funding and investment. Takeaways from this report include securing a Single European Market for trustworthy AI and embracing a holistic way of working, combining a 10-year vision with a rolling action plan.
These recommendations contribute to the EU approach to AI which was initially set out in the EU AI strategy published in April 2018. This strategy aims to boost the EU’s technological and industrial capacity, prepare for socio-economic changes and ensure an appropriate ethical and legal framework. Additionally, EU member states have agreed to cooperate on AI. Joint EU level and member states’ approach is outlined in a coordinated action plan launched in December 2018 and to be updated in 2019. Future EU funding for AI is planned within the Horizon Europe and Digital Europe funding programmes to be launched in 2021.
The Assembly marked the one-year anniversary of the creation of the EU AI Alliance. The Alliance was set up in parallel to the High-Level Expert Group as broad multi-stakeholder forum to support EU policy-making on AI. More than 3000 stakeholders have joined the Alliance.
Beyond Europe?
The High-Level Expert Group and the European Commission have argued for a need to build an international consensus and work towards a global framework for Trustworthy AI. One of the main panels at the assembly was dedicated to international governance and partners in AI. It brought together representatives from OECD, the Council of Europe, UNESCO, IEEE and the European Commission. They highlighted common themes of the EU AI ethics guidelines, OECD AI principles, UNESCO approach to AI, IEEE work on AI, AI activities undertaken by the Council of Europe and others. While emphasising the benefits of cooperation, the panel members also recognized that current times are not the best for multilateralism. The European Commission representative explained that the EU approach is to share its ethics guidelines widely but without compromises on fundamental rights and data protection.
During the event, trustworthy AI was presented as an EU brand. However, it was also recognized that the EU does not have a monopoly on it and also has to turn a mirror on itself and deal with practices that are not compatible with trustworthy AI.
You can watch the first AI Alliance Assembly here.