This review examines the Critical AI, "Humanities in the Loop," which argues for the constitutive participation of the interpretive humanities in artificial intelligence research, design, and governance. The review evaluates the essay's framing of three core dilemmas: reductive and contested meanings of "intelligence," problematic benchmarks for "AGI," and the intertwined issues of bias, error, and concentrated technological power. The argument is situated within converging scholarly traditions, including the "stochastic parrots" critique advanced with the post-Saussurean theory from Derrida, Bakhtin, and Jameson, and the historical critiques of Weizenbaum and Arendt regarding the separation of judgment from calculation. The review finds that the essay's central methodological contribution is its shift from langue to parole, from language as system to language as enacted use, which reframes the human-machine relation around the irreducibility of vouloir-dire, the "desire to say," to statistical prediction. However, the review identifies three limitations: the essay's scope as an editor's introduction, its relative inattention to the political economy of AI development, and its predominantly English-language focus, which limits its applicability to low-resource language contexts where linguistic erasure, rather than anthropomorphic projection, is the primary risk. The review concludes that the most valuable contribution lies not in definitive answers but in carving out space for patient, theoretically informed critique that refuses both technodeterminist despair and techno-utopianism, positioning the humanities as essential interlocutors in the governance of accountable technology in the public interest.
Copyrights © 2026