Soenke Ziesche Published New Paper on AI

Nov 7, 2018

The objective of this article is to identify synergies for two relevant challenges, which are currently faced by the world community, yet addressed separately: The artificial intelligence value-loading problem and the United Nations Sustainable Development Goals. First, the challenges and their significance are outlined. Subsequently, opportunities and risks are discussed to utilize the Sustainable Development Goals to set the values of an artificial intelligence. History has shown that it is complex to agree on universal and sufficiently specific human goals. Nevertheless, this is a prerequisite to approach the artificial intelligence value-loading problem, which is closely linked to artificial intelligence safety. So far, it has not been considered harnessing the Sustainable Development Goals in this context. Yet, the Sustainable Development Goals can be seen as the closest and most comprehensive existing approximation towards common human goals since it is what the United Nations, i.e. the world community, currently agrees upon. Such an attempt entails various risks, which are discussed and which are anticipated given that the artificial intelligence value-loading problem is considered very hard. However, due to the urgency it is argued here that the Sustainable Development Goals constitute an innovative as well as promising interim heuristic towards artificial intelligence safety.