Can AI Be Governed?
A Proposal for an IPAI Framework.
The artificial intelligence revolution will potentially change the composition of all human society and might even transform humanity itself. The transnational, boundary-bridging character of AI calls for a multistakeholder platform of global governance. Based on a shared proposal of the governments of France and Canada, the Intergovernmental Panel on Artificial Intelligence (IPAI) aims to follow the example of the IPCC, its role model on climate change, to become the primary source of objective scientific advice on AI. To help policy makers balance its significant opportunities and risks, the IPAI will work towards an ethical AI for the good of human society.
The mandate of the IPAI is to assess the academic, technical and socio-economic information relevant to form the scientific base of the impact and future risks of artificial intelligence on human society. It aims to mediate between entrenched multilateral positions on the right balance between velocity and robustness, between disruptive innovation and risk mitigation, which is essential given the enormous power of AI technologies.
The thematic focus will be especially on areas where an unprecedented shift towards autonomous decisions is to be expected, including (but not limited to) autonomous weapon systems, transportation, biotechnology, health care, finance, industrial production, aerospace and energy production. Given the wide range of applications of AI, various studies have indicated many use cases to further the United Nations SDGs (Sustainable Development Goals). The IPAI aims to ensure a positive impact of AI on all seventeen SDGs while preventing any adverse effects.
Another focus will be on the ethics of autonomous and intelligent systems as well as the collection, handling and usage of data. The IPAI will build upon previous work in the field such as the EU’s Ethical Guidelines for Trustworthy AI, the IEEE Ethically Aligned Design, or the G20 AI Principles. The IPAI is aware of variations in risk appetite, privacy concerns, and cultural sensitivities among member states and will consider multilateral differences in its working methods and scientific communication. Emphasis must also be put on the importance of inclusive growth of AI in developing countries to avoid rising inequality, loss of competitiveness, and the exploitation of digital resources.
The IPAI aims to be an inclusive platform for all relevant stakeholders in order to ensure credibility, public trust and international consensus. Therefore, participation is open to all UN member countries. Since the IPAI will not have a legal enforcing mechanism for cooperation and implementation, it must convince member state of its paramount scientific quality. This multilateral approach might require more persuasiveness to convince sceptics of multilateralism, but is the only way to reach the above-mentioned principles.
The IPAI Bureau as well as the IPAI Working Groups will be modelled after the IPPC structures and reflect balanced geographic and demographic representation. Similar divisions of the working groups into physical science basis, impact & adaptation and mitigation are conceivable. Specific tasks groups on subtopics such as gender equality may be established at any time. Panel members from governments, academia, civil society, industry, IGOs and NGOs will be invited based on their individual expertise, to minimise potential interference and outside influence.
The IPAI will publish annual assessment reports. In order to ensure maximal transparency, the IPAI will give free access to all published scientific advice. Meetings will be regularly open to outside observers and the media. As an intergovernmental body, review of IPAI reports will involve both peer review by experts and review by member governments. Unlike for the IPCC, there cannot be an overall benchmark to measure success due to the ever-changing nature of AI. Advancing consensus amongst stakeholders and enabling effective global governance of AI will be the criteria by which the effectiveness of the IPAI will be measured.
3. Global Governance
While undoubtedly various new institutions will evolve in the next decades, the IPAI constitutes an important element of the global governance of AI. It’s role model institution, the IPCC, has a proven track record which culminated in the 2015 Paris Agreement. The IPAI can be a similar base for international negotiations and prospective governance mechanisms, by reducing information asymmetries and bridging the gap between policy makers and technology. The AI revolution requires both soft and hard measures and a high speed of policy making, legislators are building the aircraft while flying it. Much like the IPCC, the IPAI will give policy makers the scientific resources and provide a multilateral platform for achieving these objectives and creating an ethical AI for the good of society.