Artificial Stupidity and Jailbreaks as a Form of Algorithmic Creativity

Abstract

The article discusses the practices of creativity, where the aesthetics of conscious error is transferred to the realm of algorithms. Taking the so-called perspective of artificial stupidity as a point of reference, the article analyzes the conditions of creativity in an algorithmic culture where unpredictability has become unlikely. As one of the practices of artificial stupidity, the article discusses so-called jailbreaks, the purpose of which is to deliberately confuse large language models, forcing systems to behave outside the intended purpose defined by the manufacturer. The article puts forward the idea that jailbreaks are a new type of dual-purpose narratives, which, on the one hand, act as a technical tool that “breaks” the limitations of algorithmic media, and on the other hand, they are significant structures, facing the user and performing a certain function of self-reflection. The article formulates the narratological structure of jailbreaks and highlights its similarity to the narrative structure of conspiracy theories.