Non-artificial non-intelligence: Amazon’s Alexa and the frictions of AI

Abstract

This paper examines a case where Amazon’s cloud-based AI assistant Alexa accidentally ordered a dollhouse for a 6-year-old girl. In the press, the case was defined as a technical recognition problem. Building on this idea, we argue that the dollhouse case helps us to analyze the limits of current AI applications. By drawing on the writings of Gilles Deleuze and Francois Laruelle, we argue that these limits are not merely technical but more deeply embedded in the structures where the thinking of AI can potentially happen. We point out that AI research has been compromised by the concepts of what constitutes both ‘artificial’ and by what constitutes ‘intelligence’. First, we use the notion of artificial non-intelligence to explain how different modes of digital capitalism such as voice commerce establish limits for AI. Second, we use the notion of non-artificial intelligence to illustrate the limits of associating AI’s modes of thinking with human thought.