Reader Bot: What Happens When AI Reads and Why It Matters

In this thoughtful yet unfocused inquiry, linguist Baron (Who Wrote This?) examines how large language models’ increasingly proficient reading capabilities may transform the public’s relationship with the written word. The author raises particular concern over what enriching benefits may be lost in favor of efficiency (“Reading an AI summary of Beloved can feel... like working through the earnings report for a Fortune 500 company”) and how outsourcing obligatory reading to AI may diminish the already paltry frequency of reading for pleasure (“only 54 percent of Americans read at least one book in 2023”). The book’s best bits spotlight ways LLMs are already being employed—such as evaluating college admissions essays, Hollywood scripts, and submissions for peer-reviewed journals—and AI’s limitations and negative impacts, including its tendency to “hallucinate,” or fabricate results, and its “vast” energy requirements. While skeptical of AI, Baron resists catastrophizing, instead highlighting tools that may offer a useful path forward, like “a fine-tuned version of GPT-4” created by educator Sal Khan using pedagogical principles that seek to “improve student reading comprehension” rather than read for them. However, with a lot of space given over to fun but distracting historical trivia about reading (Hitler and Eisenhower were both avid readers), the book can feel a bit haphazard. It’s nonetheless a unique discussion of AI’s impacts that will be of particular interest to educators. (Jan.)