There are some interesting bits in this article (22 page PDF) even if, in my view, the research basis doesn't allow us to generalize meaningfully. The first is the proposition that news reporting by humans is fundamentally different from that produced by machines. "Journalists engage in selective representation, deciding which events in the world are noteworthy or relevant to their audience, thus shaping public discourse. They accordingly choose words based on what they deem best captures what they wish to report or analyze... While human text represents ideas and can typically provide reasoning behind the choice of words and constructions, algorithmically generated texts merely render outputs without such explanations." Second, and as a result, "the instrumental, efficiency-oriented purposes served by LLMs exist in tension with the values expressed by the individuals interviewed in this study, particularly around accuracy, transparency, editorial autonomy, and accountability." My scepticism exists along two fronts: first, whether the reporter's art is based as much on reason as averred in the article, and second, whether machines are not in fact capable of exercising the same mechanisms themselves.
Today: Total: [Direct link] [Share]
