(Extended Play) Using AI to Write
I admittedly don’t write enough on this blog, and part of the problem is time. Back when I was working full-time I came home with barely enough energy to feed myself and get myself ready for the next day, and trying to write on the weekends or days off meant losing time that I had truly to myself, so writing unfortunately took a backseat. Another part of the problem is inspiration. As much as I want to write, sometimes I would sit at my desk, turn on my computer and nothing would come out. Sure, I could brain dump onto the page, but there were times where even that yielded nothing. So with nothing to write about, what could I do? If I had the energy, I could have just pushed through the writer’s block in order to put something to page, but if we circle back to the first problem, you could see how that never really worked out.
It would be nice if there was something that would make things easier for me, like having more time or more energy. However it never occurred to me that the thing that I needed to do in order to write more was to… not write. While it is difficult, even under the best conditions, I like the process of writing as much as I like the finished product, because the process is what makes the finished product worth it. It’s not just about the destination, it’s about the road you take to get there.
On this week's episode of Smut Book Club, I learned about a woman who, under 21 pen names, is using AI to write romance novels. The woman (who in the interview goes by the name Coral Hart, but does not disclose her other pen names, and possibly her real name, for fear it will hurt book sales), states that she is able to prompt the AI program she uses with an outline and it will “write” the novel for her. Just last year she managed to “write” over 200 books that all together sold 50 thousand copies. While this isn’t the first time AI has been used as a cheat code for writing, and romance isn’t the only genre that has fallen victim to AI slop, the increase in use and the hiding of said use, is even more concerning.
I said before that writing is a process, not just the end result. Much like the plot of a story itself, all of the twists and turns, edits and rewrites are part of what makes the process satisfying. I’ve spent countless hours on stories, from outline to finished product, figuring out what is relevant to the story and what isn’t. I’ve read countless books on the process of writing, and what makes a good plot versus a bad one, and what makes a well-rounded character versus one that is flat and unrelatable. Hell, even with these blog posts, I’ve written long screeds about my thoughts on certain topics only to cut them down or scrap them completely because they just don’t work. Do I like killing my darlings, so to speak, when I have to get rid of large chunks of the things I’m trying to write? No, but I do it anyway, because that’s part of the process. The work, the tweaking, the rewriting, all of it is not only what makes the story or blog post work in the end, but also what gives the story my unique voice. It’s my signature on the page, and I wouldn’t have been able to put that signature on it if I hadn’t gone through the hours and hours of writing and instead put a prompt into AI.
But with AI, I could forge someone else's signature.
With all the early issues surrounding all the harm that AI does (and believe me, there is a LOT), one of the big issues is how much plagiarism is involved in training them. Pages and pages that equate to hours and hours of work by authors across genres are being fed into these machines without asking permission from the writers. The artistic voices of each of these authors has been uploaded into these Large Language Models (the AI) and are being used to mimic their styles without regard to any of the work and development that these writers put into finding their own style. The most egregious part is, many AI users seem to know that this is wrong and are doing it anyway. Last year a lawsuit was brought up against Anthropic for using hundreds of novels to train it’s AI, without the permission of the authors (Anthropic’s Claude is also conveniently the program that was being used by the author with 21 pen names to “write” her books). One law professor, Ed Lee, commented that if ruled in favor of the authors, the amount of money needed to fairly payout the authors could end Anthropic entirely, which some think is a good enough reason not to compensate the authors for their work (it is not good enough reason, in my opinion). 21 Pen Names herself refuses to disclose the other pen names that she is writing under. She states that it’s an experiment, but the lack of disclosure makes her use of AI seem more nefarious than scientific.
While all of this seems very daunting, I don’t want to be pessimistic and assume that the coming AI wave is inevitable. There have been a lot of “unavoidable” tech innovations that were pushed to the general public that have lost momentum and money (crypto, NFTs), and it seems as though there’s starting to be a public backlash to AI, with many people calling for AI regulation, and opting out of anything marketed as AI due to the negative stigma. Personally, I’ve been doing my best to keep up with how to spot AI on my own, and not engaging with it. And that’s not just with written AI, but also with videos, photos and audio, despite the fact that every day companies make it harder and harder to distinguish what’s real and what’s not. It’s time consuming, it’s arduous, and it’s most of all annoying but it’s a process, and if this process helps me find art that people have put time and effort into, that people had to write and rewrite, film and re-shoot, draw, erase and draw again then I’d rather that then anything else.
