The Bullshit Problem

Since the launch of ChatGPT in 2022, generative artificial intelligence (GenAI) has proliferated across the tech industry, accompanied by a persistent critique: it produces bullshit. Not lies, but outputs shaped by statistical correlations with no regard for truth—what Frankfurt called the sociolinguistic condition of bullshit. This framing, however, misses the point. In this chapter, VandenBroek argues that the real issue is not bullshit as a technical flaw but as an epistemology—one that obscures the origins, logics, and commitments behind its knowledge. Using Dumit’s implosion project, she compares the epistemics of different AI systems and anthropology to draw out the distinctive differences that she argues draw us toward more productive frameworks for thinking about and living with AI. Symbolic AI and anthropology offer situated, partial ways of knowing, while GenAI generates a new monstrous, situated and partial position constructed from the tokenized scraps of other’s lives, thoughts, writing, and art with each response—what she calls a bullshit epistemology. This chapter challenges the assumption that the solution lies in fixing GenAI’s confabulations. Instead, she asks what GenAI reveals about our expectations for AI and how we might live with them more deliberately.