Genre Grapevine on Two Secrets about LLM Usage Every Writer Must Remember

Like many writers in the SF/F community, I was horrified when on December 19th the Science Fiction & Fantasy Writers Association (SFWA) briefly changed the rules for the Nebula Awards to allow partial LLM usage. In response to member outrage, SFWA undid this change a few hours later and banned all large language models (LLM) usage in any works under consideration for the awards.

In my follow-up post the next day, I examined what was likely involved in this overstep by SFWA. I truly believe the organization was trying to deal in good faith with the ever-creeping integration of generative AI systems in many aspects of all our lives, such as in search engines used for research and many popular writing software systems. They didn’t want authors declared ineligible for the awards through unintentional LLM usage. But they also wanted to say that SFWA was opposed to LLMs trained on the stolen works of many writers, including myself and much of the SFWA membership, and that these works wouldn’t be eligible for one of the genre’s premier writing awards.

Since SFWA’s quick about-face has resulted in the awards currently banning works created with any LLM usage, some people are worried that if authors used spellcheck or researched with ChatGPT their stories are now ineligible for the Nebulas. I don’t believe that is the intent of the new “no LLM” rule but it does mean there is still some confusion on all this.

To SFWA’s credit, they apologized and are now asking both members and the larger writing community to share their thoughts on LLM usage. And that’s exactly what we’re witnessing.

Erin Underwood published an essay on File770 about all this, noting that while “the fundamentally human act of creation must remain in human hands,” at this point there’s “no putting the genie back in the bottle.” She then suggests ways in which LLMs may be mainstreamed into not only helping authors write but also assist in book publishing, marketing, publicity, and even analyzing the contracts we sign.

The responses to Underwood’s essay from many SF/F writers was not positive. Chuck Wendig wrote an excellent response, opening with consideration of the many so-called reasonable questions or exceptions that people want to make for LLM use.

As Wendig said, these “reasonable questions”

“... are a Trojan horse to allow a lot of other exceptions in through the city gates. To continue to mix metaphors, if you give a mouse an AI cookie, well, he’s gonna want the AI milk, the AI straw, until eventually you’ve given him an AI nuclear bomb where he kills all the human beings and can feast on our smoldering corpses at his rodenty leisure.”

Wendig also disputes the idea that the AI genie is out of the bottle and that as writers we have no choice but to accept it. As he stated several times in his essay, “AI is not inevitable.”

For example, with regards to the publishing industry using AI, he said

“Publishers can and must avoid using generative AI and LLM AI. Publishers remain competitive by hiring and training real people to do real people jobs that support real people authors and real people readers. AI remains a broken foot. Bad for the environment, bad for writers, and also, generally doesn’t work well — it certainly doesn’t work as well, or as creatively, as actual humans! Remember, the AI is fed with the work of actual humans. Why do you think that is, exactly? If you use it, it means you’re replacing people. People who could’ve done the job better. People who actually did the job, and now their work is pilfered and duped.”

I recommend people read Wendig’s entire essay.

Foz Meadows also wrote an excellent response to Underwood’s essay, exploring more with regards to the ethical reasons why people shouldn’t use LLMs to write.

As Meadows said,

“AI is foundationally unethical in ways that go far beyond its creative theft, and which do not only concern the communities from whom that work was stolen. AI is unethical the way billionaires are unethical, and exists for much the same reason and with many of the same risks, because the one is inextricable from the other: a culmination of wildly unregulated capitalist excess which, if left unchecked, legitimately threatens both global democracy and the long-term habitability of our one and only planet.”

I recommend people also read Meadows’ entire essay.

Shortly after SFWA’s Nebula Awards rule change and immediate reversion, I shared my thoughts on where to draw the line with LLM usage:

“If a writer uses tools like spelling and grammar checks, speech to text or other accessibility tools, translation programs to read sources that aren’t in their own language, or search engines for research, we shouldn’t hold any of that against them even if an LLM powers those tools. The same if they use a writing tool like Google Docs that has an integrated LLM. As long as the writer isn’t using the ‘Help me write’ prompt in Google Docs then Google Docs can be seen as just another word processing program.”

I still stand by those standards.

Earlier this year I wrote about why I don’t use generative AI in my writing. All my points still stand. However, I’ve since realized two additional and very powerful reasons why writers should avoid using LLMs to assist with their writing.

The first is because of a comment on File770 in response to Underwood’s essay. As Madame Hardy said,

“First drafts are where the thinking begins. In a first draft, you discover that Subsidiary Character A is actually the protagonist; that the second scene you’d planned is the real story, and that the rest is irrelevant; that you need to spend \an hour\ days researching Edwardian shoes; that actually there’s a separate story to be written springing from something you found while researching Edwardian shoes. If you shortcut the first draft, you’re boarding an express train before you have ever explored the territory.”

I’m 100% totally and forever in agreement with this! If you’re looking to spit out a predictable story as fast as possible, then an LLM may be the tool for you. But if you’re looking to create a story that only you can tell and that’s something truly worth exploring, you need to write and draft and rewrite and keep exploring until you get to the heart of that story.

The other thing writers must remember about generative AI usage is that LLMs are a tool. If you use an LLM instead of fully developing your own skills as a writer, then you become dependent upon that tool.

And while a skill is a part of you, a tool can easily be taken away!

Here’s a simple truth: Those pushing LLMs as a way to replace as many aspects of our work and lives as possible are corporations and billionaires that have no more connection to us than a tiger to a tadpole. When have these corporations and billionaires ever had the best interests of humanity or even your own individual life at heart?

Writers may remember how social media was once a great way to promote your books and engage with readers. At least, it was a wonderful tool until certain corporations and billionaires changed how those sites worked. Now much of social media is only useful for promotional purposes for bots and rich influencers.

Or remember how great Amazon used to be for self-publishing your own books and reaching eager new readers? Then Amazon changed how their site worked. Now you have to pay Amazon tons of money to promote your self-published book to reach those same readers.

Essentially, those useful tools for writers have changed so much that they are far less useful today than they were 10 or 15 years ago. Many writers who depended on those tools lost out.

So say you’re a writer who grows dependent on using LLMs to help you draft and write your stories. And say that readers ignore the shortcomings of your LLM writing and you become a big success. Unlikely that’ll happen anytime soon, but let’s pretend it could.

Then a few years down the road, the corporations and billionaires controlling LLMs change how these systems work, or block you from using them, or go belly up. Now you’re missing the tool that helped you write. If you’d learned the skills to write on your own, no one could have ever taken that from you. But you didn’t do that, so you’re screwed.

Writers across our community are struggling with how LLMs will affect both their own writing and the stories they love creating. Erin Underwood was correct that there needs to be a broader conversation on all this.

But during these conversations, writers should also never forget these two secrets of using LLMs to help your writing:

• First drafts are where the thinking begins.

• If you become dependent upon a tool, that tool can be taken away.