“A natural language interface for everyday life”: The social and political functions of AI capabilities discourses

We interpret and historically contextualize AI capabilities discourses: spectacular narratives that ostensibly lay out the capabilities and use cases of the latest generative AI models in marketing and scientific communications. We complement existing work scrutinizing these discourses as components of a wider phenomenon of "AI hype"- which tends to focus on "debunking"them - by exploring how they rationalize wider social projects. Via an analysis of marketing materials produced by companies like OpenAI, we trace how AI capabilities discourses extend longer-running capitalist discourses, which construct aspects of self and social life as segmentable "skills"with economic value, and which responsibilize people for their survival in precarious economic conditions. We then show how AI capabilities discourses build on top of this ideological scaffolding. We suggest that these discourses have power because they reiterate hegemonic ideas about society and personhood that are entwined with material conditions. As such, contesting the power of the AI industry necessitates more capacious forms of political action, beyond interventions aiming to merely correct narratives.