We’re joined by Leif Weatherby, associate professor at NYU, founding director of the Digital Theory Lab, and author of the new Language Machines: Cultural AI and the End of Remainder Humanism, to think with us about AI, structure, and what happens when computation meets language on their own shared turf. Language Machines is easily the best book about AI written this year and is just a killer antidote to so much dreary doomer consensus, it really feels like one of the first truly constructive pieces of writing we’ve seen out of academia on this subject.
 
This episode follows really well after two others — our talk with Catherine Malabou earlier this summer and the episode with M. Beatrice Fazi about a year ago (both faves). It feels like theory is opening back up again into simultaneously speculative and structural returns, powered in no small part by the challenges posed to conventional theories of language (from Derrida to Chomsky) by Large Language Models. This episode absolutely rips, literally required listening.
 
Structuralism is so back (and we’re here for it).

Some important references among many from the episode:

• Roman Jakobson, “Linguistics and Poetics.”
• N. Katherine Hayles, Unthought: The Power of the Cognitive Nonconscious .
• Beatrice Fazi, Contingent Computation: Abstraction, Experience, and Indeterminacy in Computational Aesthetics.
• Steven Pinker, The Language Instinct (1994).
• e.g. Noam Chomsky, Ian Roberts & Jeffrey Watumull, “The False Promise of ChatGPT,” NYT (link) 
• Anthropic, “Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet” (featuring the Golden Gate Bridge example - link)
• LAION-5B dataset paper and post-hoc analyses noting strong Shopify/e-commerce presence in training scrapes.
• Weatherby in the NYT