Governance of Artificial Intelligence: An Unachievable Aim?

AI governance concerns all societal levels and refers to politics, military, social systems, economy, and business. It is impossible to present a universal definition of AI governance that embodies both broadly defined institutional aspects of governance and normative recommendations concerning technical functional solutions relating to deep learning. In all attempts to develop AI governance, rules are usually presented based on ethical assumptions and involve all stakeholders. It is visible in all documents concerning AI governance, which various national and international institutions and business institutions have prepared. The universal challenges of any kind of governance deriving from the implementation of AI are identical with the traditional patterns of exercising power at all levels of societal hierarchy deriving from the infinite regress and reflected in the famous Juvenal’s saying “Quis custodied ipsos custodies?”. In reference to the institutional reality this observation meets a specific contradiction. While this saying reflects the infinite regress in social control, in the institutional reality, there always must be the highest level of control. A sovereign state is the best example. The paper aims to show how obstacles deriving from multi-criterion and multi-level governance always lead to unsolvable paradoxes at all levels of societal hierarchy