Sonja Drimmer & Christopher J. Nygren:
"There is a fundamental epistemological disjuncture between what PhD-holding historians [or, I would add, other professionals in the humanities] do and what ChatGPT and its ilk do: the former meticulously, purposefully, and rigorously comb through a mountain of human-curated documents looking for revealing details that diverge from the baseline, offer indications of cultural shift, or elements humanity embedded in seemingly mundane activities; the latter processes terabytes of machine-harvested data in order to predict what will be the most likely next token in a string[.]"
"Computers are good at pattern recognition; but pattern recognition and token prediction are not learning. To continue calling them machine “learning” or artificial “intelligence” is to agree with a fallacious metaphor that risks irreparable harm to students, the citizenry, and, by extension, humanity in the form of death-by-a-thousand-cuts. "