AI Bugs and Failures: How and Why to Render AI-Algorithms More Human?

Abstract

AI systems such as self-driving cars, or autonomous lethal weapons are expected to work in a framework called ‘explainable AI’, under meaningful human control, in a fail-proof way. In this chapter, the author discusses case studies where the opposite framework will prove more beneficial: i.e. in certain contexts, such as cultural and artistic production or social robotics, AI systems might be considered humanlike if they deliberately take on human traits: to bluff, to joke, to hesitate, to be whimsical, unreliable, unpredictable, and above all to be creative. In order to uncover why we need ‘humanlike’ traits -especially bugs & failures, the chapter considers representations of intelligent machines in the imagination of popular culture, and the deeply ingrained fear of the machine as the ‘other’.