The Pen Is Mightier Than The Large Language Model

Creators of all kinds have been forced to consider how artificial intelligence will affect their work and livelihoods. The advent of large language models (LLMs) like ChatGPT crystallizes the particular threats writers face; while LLMs can’t replace the humanity necessary for good reporting, fiction and poetry, they can flood the internet with mediocre content that could make higher quality work even less marketable than it already is—if we, the reading public, let them.

The Authors Guild released an open letter, signed by over 8,000 authors, addressed to the CEOs of OpenAI, Meta, Microsoft, and other companies demanding that authors get the opportunity to give consent and get credit and compensation when their work is used to train AI models. The letter is at least partially retroactive in nature as AI companies have already used writers’ copyrighted content without permission and for profit.

The Authors Guild letter is (perhaps unsurprisingly) well-written. “Millions of copyrighted books, articles, essays, and poetry provide the ‘food’ for AI systems, endless meals for which there has been no bill… It is only fair that you compensate us for using our writings, without which AI would be banal and extremely limited,” it says, directly addressing CEOs.

Credit and compensation would not amount to a full solution, especially because one-time payments can’t make up for the loss of entire markets and future earning potential. This point is especially salient given the low wages the current market already forces writers to tolerate in pursuit of their craft: according to the open letter, the median salary for full-time writers in 2022 was $23,000.

The letter also addresses the issue of what some writers consider the original sin: content theft.

“Some creators object to the fact that their stuff was taken in the first place and is being used to generate material that is similar to theirs,” Mary Rasenberger, Executive Director of the Authors Guild, said in an interview. The Authors Guild is, according to Rasenberger, “trying to start the nice way by getting legislation that will help us get paid.” But she also notes that the Guild is “fully prepared to litigate” if it comes to that.

Rasenberger said that the U.S. government is so worried about the ongoing AI competition with China that there is no chance it will mandate that companies destroy and remake their LLMs fairly. Credit and compensation are the relatively low-hanging fruits in this context.

“Even if all Americans don’t necessarily value the arts, they certainly understand that culture represents a massive part of our economy,” she said, pointing out that creative fields make up 9% of the United States’ GDP and are major national employers—not to mention the fact that the U.S. is a top exporter of cultural work.

It is perhaps this last point—that the U.S. sustains its soft power largely through cultural exports written by American humans—that may sway the lawmakers hesitant to regulate artificial intelligence based on the misconception that doing so will weaken America’s power relative to China’s. American cultural output is made possible in part due to the nation’s unique strength: that it houses a diverse, open society where ideas can clash, usually peacefully.

AI regulation that protects America’s creatives would also preserve the quality of its journalism, poetry, television shows, movies, stories and books that the U.S. is known and admired for around the world. Without our written culture, the ground we stand on as an exemplar of speech that is not only free, but also high quality, will substantially weaken.

In fact, safeguarding the quality of American writing is a central argument in the letter: “Generative AI threatens to damage our profession by flooding the market with mediocre, machine-written books, stories, and journalism based on our work,” it states. Even models that are trained on good or great writing are still likely to produce lower quality, literally lifeless renditions of the materials they were “fed.”

That’s because, according to Rasenberger, the regurgitation AI models perform couldn’t be more distanced from the work of a writer: “There’s no thinking going into it; it’s just math,” she said. “There’s no connection to the real world.”

Writers who have trained and practiced their whole lives to be good at what they do can write in ways that machines can’t. The real question is directed at cultural consumers: will we accept a barrage of half-decent content that succeeds in distracting us but fails to make us think or feel? If so, we will be complicit with a market shift that threatens the existence of artists and writers who are, in every sense of the word, real.

Many writers have already pushed themselves to their financial limits, eking out salaries through side hustles and pursuing their creative or more interesting work in addition to whatever they do for still-little money. What may be underappreciated is the fact that they do that while contributing to the cultural lifeblood and resultant soft power of the United States. Further threats to their craft, AI-powered or otherwise, should not be tolerated.

As Rasenberger put it, “What we’re seeing right now is existential.”