Shalizi's frame of artificial intelligence as mechanized tradition (www.argmin.net)

At the recent Cultural AI conference held at NYU, Cosma Shalizi presented a thought-provoking framework that characterizes artificial intelligence, particularly large language models, as a form of "mechanized tradition." Shalizi argues that these models operate by retrieving and synthesizing information from vast amounts of internet text, ultimately reproducing the patterns and regularities inherent in that training data—essentially mimicking human cultural formulas. He emphasizes that following traditions allows for externalized thinking, thereby enhancing cultural communication and understanding. This perspective reframes AI not just as a technological advancement but as a critical extension of collective human intellect. The significance of Shalizi's discussion lies in his critique of the notion of "artificial intelligence" as a standalone concept, suggesting instead that it is a reflection of the mechanization of our communal cultural practices. By emphasizing the implications of this mechanized tradition, he encourages a deeper exploration of identity, creativity, and the role of AI in societal discourse. The conference also featured experiments comparing human and AI-generated literary styles, revealing nuanced preferences that highlight human biases toward traditional authorship and raise questions about originality in AI outputs. Ultimately, these discussions promote a more nuanced understanding of AI as an evolving manifestation of human cultural practices and shared intellect.