What Do We Talk About, When We Talk About Rejecting Generative AI?An Essay by Chisom Umeh
- Chisom Umeh
- Jun 4
- 4 min read
Five months ago, the Science Fiction and Fantasy Writers Association (SFWA) updated its Nebula Award eligibility information, stating that, “Works that used LLMs at any point during the writing process must disclose this upon acceptance of the nomination, and the nature of the technology’s use will be made clear to voters on the final ballot.”
This update immediately drew the attention of writers and editors including Premee Mohammed, Meg Elison, Neil Clarke[1], and more. They unanimously criticized this new addition to the Nebula Award eligibility information for its subtle permission of AI-assisted or generated works. The SFWA’s statement that “…the nature of the technology’s use will be made clear to voters on the final ballot” meant that works would be allowed to stay on the ballot even though their authors had confirmed AI usage in their creation.
After writers correctly pointed this out, the SFWA rewrote the rules again and informed its members and the general public of its latest position on the issue within a matter of hours. They apologized[2] for the implications of the earlier wording and assured every one of their strict stance against AI usage. The updated rule read:
“Works that are written, either wholly or partially, by generative large language model (LLM) tools are not eligible.”
Unfortunately, even though the SFWA has now taken a hard, rather than passive, stance on the matter, this inevitably creates the problem of people not
being transparent about their AI usage.
Very recently, Jamir Nazir, a Trinidadian writer whose work was published on Granta after winning the Commonwealth Writers Prize for the Caribbean region, was suspected by readers of using AI to either generate the story in full, or in part. They provided several pieces of evidence[1] within the story to back up their claims, many of which seemed valid. They also called out the Prize judges for letting such a work slip through. Many others, though, maintained that the writing didn’t look like anything a human couldn’t create and labelled the accusations a witch hunt. Granta and the Commonwealth Foundation released statements backing both the publication of the work and the integrity of the writer, respectively. The story is still up and can be read here[2].
I think we genuinely have a problem on our hands.
The only definitive way to confirm that a piece was the product of an algorithm is, unfortunately, a direct confession from the creator. Not AI checkers nor our intuition. But, in the absence of the writer admitting their usage of AI themselves as the SFWA requires, an action which would be as common as a thief returning their “hard-earned” steal, what do the rest of us do?
When we don’t have the luxury of a guilty plea, do we continue to sit around and publish or vote everything that comes our way just because we can’t prove anything for certain?
Of course, there’s a real danger of throwing the baby with the bathwater. We can, in the process of expressing our hate for a machine owned by capitalist billionaires, misjudge a human-written story and greatly harm an innocent writer who was only trying to be experimental with their work.
But, consider what this may mean if dubious AI ‘writers’ learn of the inability of the writing community to definitively dismiss anything. They’d know that, if the work somehow gets published, after all is said and done, we’d still pay them and leave it on our sites, no matter how many writers raise suspicions of the origins of the work.
If, in a case like Jamir Nazir’s, whose work was replete with telltale generative AI patterns (from meaningless and over-the-top metaphors, negative parallelisms, his author photo being generated by AI, to his LinkedIn chock-full of AI advocacy), we are still divided on the matter, then on what grounds could we ever prosecute any future “writer” whose work has similar problems?
I’m not in any way suggesting that this is an easy problem to solve or that I have some one-size-fits-all solution. I admit that it is a tricky issue that requires a lot of thinking through and caution. In fact, one may weigh the harm that could potentially be done to a writer if indeed they have been accused falsely and their reputation marred, against the harm that would happen to the writing/reading community if one AI work is actually published, and conclude that the former is more severe since there’s an actual person involved, and the latter is almost nearly a victimless crime (in the case of Jamir Nazir, neither the Commonwealth Foundation nor Granta accept being wronged. So, who, really, is the victim?) Therefore, they may wonder what anyone stands to gain by constantly policing submitted works for AI usage.
But then, we’d have to ask ourselves, what is the point of doing literature? What is the purpose of art? Why do we bother translating our unique human experiences into stories if anyone at all can type prompts into a computer and an algorithm fakes it for them?
What do we mean when we talk about rejecting generative AI in literature if we lack the willpower to enforce that rejection when the preponderance of the evidence tilts that way? My sincere hope is that we find a meaningful and lasting solution to this.
A tiny hole can sink a ship. I pray we recognize those holes when we see them, and when we do recognize them, we have it in us to plug them. It may be one story today that we let slide. But that singular piece can open the door for a deluge that’ll drown all our endeavours.
Reference List of Cited Links
Comments