Controlling the Creations

Abstract

Turner explains how in order to implement constraints into AI directly, we will need to address both moral and technical questions: Which norms should be chosen? How can these be implemented? Potential basic laws for robots include: a law of identification, requiring that AI makes its status clear; a law of explanation, requiring that at least some parts of AI’s reasoning be divulged; a laws on avoiding bias; and a law setting out any limits to areas where AI can operate. Finally, a kill switch law might make it mandatory that AI systems include a mechanism for safely interrupting their processes or operations, either temporarily or permanently.