On today's show, Alex and Calvin continue their series on “AI” and public discourse, focusing this time on the increasing proliferation of AI applications in government writing, policy, and social media. We characterize the second Trump administration as the "first totally post-AI presidency," which has adopted the "dumbest, most unreflective, most uncritical approach" to AI's use in communication, research, and analysis. Throughout the show, we emphasize how AI technologies are themselves rhetorical artifacts at the same time as they so often produce “bad” rhetoric, reflecting the intentions, values, and presuppositions of their creators, as well as the inherent biases of their training data and text generation models. This often results in an entry-level, overly dense writing style - often referred to as "slop" - which is almost written not to be read, but rather to fill space.
We explore several concerning examples of AI's uncritical adoption by the secondTrump administration and their acolytes in the tech world. Early executive orders exhibited AI-generated formatting errors and formulaic, generic language, demonstrating a context-blind style that could lead to legal problems and erode public trust. Furthermore, the "MAHA Report" from the Office of Health and Human Services was found to fabricate studies and misrepresent findings, reflecting how large language models are "sycophantic," and can reinforce existing (often false) beliefs. Our discussion also covers Palantir's "Foundry" product, which aims to combine diverse government datasets, raising significant privacy and political concerns, especially given the political leanings of Palantir’s founders. Finally, we examine xAI’s Grok chatbot (run by Elon Musk), which illustrates how tech elites can exert incredible political power through direct interventions in AI tools’ system prompts - which in recent months has led Grok to parrot conspiracy theories and make explicit antisemitic remarks on the public feeds of X/Twitter.
Ultimately, our analyses emphasizes - once again - that these so-called “AI” technologies are not neutral; they are, in the words of Matteo Pasquinelli, "crystallization[s] of a productive social process" that "reinforce the power structure that underlies [them]," perpetuating existing inequalities. Understanding these mechanisms and engaging in what Pasquinelli terms "de-connectionism" - undoing the social and economic fabric constituting these systems - is essential for critiquing the structural factors and power dynamics that AI reproduces in public discourse.
Have any questions or concerns about this episode? Reach out to our new custom-tuned chatbot, @Bakh_reverb on X/Twitter!
Examples Analyzed in this Episode:
Trump Admin Accused of Using AI to Draft Executive Orders
https://www.yahoo.com/news/trump-admin-accused-using-ai-191117579.html
Eryk Salvaggio - “Musk, AI, and the Weaponization of ‘Administrative Error’” https://www.techpolicy.press/musk-ai-and-the-weaponization-of-administrative-error/
Emily Kennard & Margaret Manto (NOTUS) - “The MAHA Report Cites Studies That Don’t Exist” - https://archive.ph/WVIrT
Sheera Frenkel & Aaron Krolik (NYT) - “Trump Taps Palantir to Compile Data on Americans” https://www.nytimes.com/2025/05/30/technology/trump-palantir-data-americans.html
David Klepper - “Gabbard says AI is speeding up intel work, including the release of the JFK assassination files” https://apnews.com/article/gabbard-trump-ai-amazon-intelligence-beca4c4e25581e52de5343244e995e78
Miles Klee - “Elon Musk’s Grok Chatbot Goes Full Nazi, Calls Itself ‘MechaHitler’” - https://archive.ph/SdoJn
Works & Concepts Cited in this Episode:
Bakhtin, M. M. (2010). The dialogic imagination: Four essays. University of Texas Press.
Benjamin, R. (2019). Race after technology: Abolitionist tools for the new Jim code (1st ed.). Polity.
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021, March). On the dangers of stochastic parrots: Can language models be too big?. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency (pp. 610-623).
Our previous episode with Dr. Bender about her work
Burke, K. (1984). Permanence and change: An anatomy of purpose. Univ of California Press.
Burke, K. (1965). Terministic screens. In Proceedings of the American Catholic philosophical association (Vol. 39, pp. 87-102).
DeLuca, L. S., Reinhart, A., Weinberg, G., Laudenbach, M., Miller, S., & Brown, D. W. (2025). Developing Students’ Statistical Expertise Through Writing in the Age of AI. Journal of Statistics and Data Science Education, 1-13.
Haggerty, K. D., & Ericson, R. V. (2017). The surveillant assemblage. Surveillance, crime and social control, 61-78.
Hill, K. (2025, 13 June). “They Asked an A.I. Chatbot Questions. The Answers Sent Them Spiraling.” The New York Times.
Markey, B., Brown, D. W., Laudenbach, M., & Kohler, A. (2024). Dense and disconnected: Analyzing the sedimented style of ChatGPT-generated text at scale. Written Communication, 41(4), 571-600.
Miller, C. R. (1984). Genre as social action. Quarterly journal of speech, 70(2), 151-167.
Murakami, H. (1994). Dance dance dance : a novel (1st ed.). Kodansha International.
Pasquinelli, M. (2023). The eye of the master: A social history of artificial intelligence. Verso Books.
Reinhart, A., Markey, B., Laudenbach, M., Pantusen, K., Yurko, R., Weinberg, G., & Brown, D. W. (2025). Do LLMs write like humans? Variation in grammatical and rhetorical styles. Proceedings of the National Academy of Sciences, 122(8), e2422455122.
An accessible transcript for this episode can be found here (via Descript)
