Can A.I. Produce Writing That We Actually Want to Read?
In the previous installment of this series on the future of higher education, I talked with professors about the ways that A.I. has changed their classrooms. Most felt despair over the breakdown of a contract between student and teacher, one predicated on the faith that, even if students weren’t always perfect, they would at least challenge themselves to think every once in a while. If students rely on A.I. summaries to do their “reading” for them, if they don’t attempt to put their ideas into prose, are they really learning anything?
TL;DR
- AI writing can be identified by specific stylistic tells, such as overused em dashes, "not X but Y" constructions, and certain verbs.
- A key limitation of current AI writing is its tendency towards describing stasis and absence, with characters often doing very little.
- An experiment involving AI-generated passages in the style of classic authors revealed that most participants could distinguish AI text from human writing, though accuracy varied.
- Even when mimicking specific authors like Bram Stoker, AI-generated text often lacked the dynamic action and character engagement found in human literature.
- The author suggests that while AI may become more proficient, the human process of writing and the reader's desire for genuine human connection will ensure the continued value of human-authored literature.
- The discomfort associated with reading AI-generated text stems from a feeling akin to discovering an opponent using a bot in chess, highlighting the value placed on the human effort and process.