I'm writing this largely to try and figure out what I think about the controversies over generative AI and its use in the creation of entertainment, arts, literature, essays, and job-related writing. I should start by making two things clear: 1) I have never used AI to create anything except once when, for a lark, I let Gmail's AI write a two-sentence email reply to a friend, and it wrote exactly what I would have; 2) I'm currently something of an agnostic about AI; I understand the concerns about its use but I also understand that it could be useful in some areas.
1. AI is wrong, a lot! I encounter AI quite a bit on social media. It seems to be being used as a prime engine for creation of content on some Facebook groups. For example, some groups about movie and theater history are sustained mostly by what I suspect are AI articles. Though on the surface, these short articles are readable enough, they have no real authorial style to them and sometimes leave out important information or get things wrong—one mentioned a musical stage show with Lucille Ball that never existed; an AI hallucination, I guess. I've also seen it get people's birth dates and death dates wrong, and it claimed that reporter and author Mark Harris, who is plenty alive and kicking, was dead, perhaps mixing him up with the Mark Harris who wrote Bang the Drum Slowly. Many cases of AI providing fictional source citations in academic documents add to the argument that AI is bad at generating nonfiction.
2. As a former teacher, I have great sympathy with teachers who are currently struggling with students using AI to write papers. Not only is AI not going away, but colleges are largely welcoming it. When I was teaching college English many years ago, I struggled with use of the Internet; first, with coaxing students to use it for sources, and later to put on some brakes and to be mindful of the authoritativeness of such sources. Teaching students how to use AI effectively and honestly (which can be done) is probably a current goal of many teachers. Because AI is often wrong, it can be problematic to use it heavily in research; in terms of writing a paper, it has been shown that students who do this are not only not learning to write (i.e. communicate clearly with an audience) but they are also falling behind in learning how to think: about topics, about argumentation, about research, about the ability to pick and choose the elements of an individual writing style.
3. The use of AI in creating art and writing literature is quite controversial. Obviously creators who make their livings on their own without AI are worried about losing readership and jobs to those who create using AI, like Coral Hart, the romance writer interviewed by the New York Times a while back who published 200 romance novels in the space of one year using AI. As she put it, none of the books has become a bestseller, but the sheer amount of content produced keeps dollars rolling in. Lots of readers assume they would know and not like fiction produced by AI, but what happens when AI gets better at it and people actually start liking it, not just the easy-to-please romance readers? I don't mean to dis romance fans; I read mysteries and given the general sameness of a lot of mystery types (cozy, gay, hard-boiled), I'm not sure I could necessarily smoke out an AI mystery if I read one.
The recent skirmishes over AI fiction published by large mainstream houses sharpens the focus on an important point: if people like the fiction, does it matter that it was generated (completely or partially) by AI? The recent book "It Was Either That or Die" by Thelyson Orelien was in line for a major literary prize in France until AI detection programs determined that the book was written at least in part using generative AI. Honestly, if critics loved the book two months ago, does it matter that much if the author used AI? Isn't it still (apparently) a very good novel that Orelien had some hand in authoring? This problem goes back to the issue of "authenticity." In literature and music in particular, we are supposed to value that work came from an artist's heart or soul and not from whole cloth. In fact, life experiences probably do make an artist's output richer, but that hardly negates work that is not based on the artist's life. Joni Mitchell didn't make it to Woodstock, but she wrote a song about it that has come to be considered the music festival's theme. Many singers record songs they did not write; Tony Bennett and Frank Sinatra weren’t songwriters, but interpreters of other people’s songs, and no one docked them any talent points for that. Edgar Rice Burroughs never went to Africa but that didn’t stop his Tarzan novels from becoming wildly successful. We haven't had a major fuss over AI-generated songs but just wait—it will happen.
One of the problems here is the use of AI detectors. No one claims that they are incapable of error; they can and have misidentified human writing as AI. In the case of Mr. Orelien, the book does appear to spring from his life experiences and interests. I assume that virtually all AI creation still involves a human being at some point: providing prompts and guidance, editing, rewriting, etc. My musings here are not the thoughts of a person whose living is threatened by AI, so some may find that they don't have much power. I still don't quite know where I stand on these issues. The use of AI in areas with real-life consequences, like science or the law or nonfiction writing, seems much less ambiguous to me. In these cases, AI should only be used as a tool and should always be checked and double-checked. In the arts, I’m not sure where the boundaries lie.
By the way, the illustration at the top of my post comes from a post on Medium by Robert Henahan who writes about using AI as "a reading partner, not a replacement," using it to discover and to think about literature. Most of what he says makes sense, though he's not really talking about using it to create art.
