Authors feed their own literary works into AI models for the sake of creativity: NPR

The vast majority of authors do not use artificial intelligence in their creative process – or at least do not admit it.

Yet according to a recent survey from writers advocacy group The Authors Guild, 13% said they use AI for activities like brainstorming character ideas and creating outlines.

Technology is a controversial topic in the literary world. Many authors are concerned about the use of their copyrighted material in generative AI models. At the same time, some are actively using these technologies, even attempting to train AI models on their own work.

These experiments, although limited, teach their authors new things about creativity.

Best known as the author of technology and business-focused nonfiction books like The long tail, Lately, Chris Anderson has tried his hand at fiction. Anderson is working on his second novel, about drone warfare.

He says he wants to put generative AI technology to the test.

“I wanted to see if AI could do more than just help me organize my thoughts, but also start injecting new thoughts,” Anderson says.

Anderson says he fed parts of his first novel into an AI writing platform to help him write this new one. The system surprised him by moving its opening scene from a corporate meeting room to a karaoke bar.

“And I was like, you know? This could work!” Anderson said. “I ended up writing the scene myself. But the idea was AI.”

Anderson says he didn’t use a single word generated by the AI ​​platform. The sentences were grammatically correct, he said, but failed to replicate his writing style. Although he admits to being disappointed, Anderson says that ultimately he’s OK with having to do the heavy lifting himself: “Maybe it’s just the universe that tells me that writing actually involves the act of writing.”

Train an AI model to imitate style

It is very difficult for commercial AI models like GPT and Claude to imitate the styles of contemporary literary authors.

Authors NPR spoke with say that’s because these models are primarily trained on content scraped from the Internet, like news articles, Wikipedia entries and how-to manuals — standard, non-literary prose .

But some authors, like Sasha Stiles, claim to have successfully adapted these systems to their stylistic needs.

“There are times when I ask my IT person to write something, and then I use what came out verbatim,” Stiles says.

The poet and AI researcher says she wanted to make the commercially available AI models she had been experimenting with for years more responsive to her own poetic voice.

So she began personalizing them by entering her finished poems, drafts, and research notes.

“All this with the intention of somehow becoming a mentor to a tailor-made poetic alter ego,” Stiles says.

She has collaborated with this tailor-made poetic alter ego on various projects, including Technology (2021), a volume of poetry published by Black Spring Press; And “Repetitions: again, again,” a multimedia poem created last year for luxury fashion brand Gucci.

Stiles says working with her AI character made her wonder if what she’s doing East in fact poetic, and where the border is between human and machine.

“It’s been really provocative to be able to use these tools to create poetry,” she says.

Potential problems arise with these experiments

These types of experiences are also provocative in another way. Mary Rasenberger, CEO of Authors Guild, says she’s not opposed to authors training AI models on their own writing.

“If you’re using AI to create derivative works of your own work, that’s totally fine,” says Rasenberger.

But building an AI system that fluently responds to user prompts requires large amounts of training data. So the fundamental AI models that underpin most of these literary style investigations may contain copyrighted works.

Rasenberger pointed to the recent wave of lawsuits filed by authors alleging that AI companies trained their models on unauthorized copies of articles and books.

“If the result actually contains other people’s works, that creates real ethical problems,” she says. “Because you should get permission for that.”

Getting Around Ethical Issues While Being Creative

Award-winning speculative fiction writer Ken Liu says he wanted to sidestep these ethical issues, while creating new aesthetic possibilities through AI.

The former software engineer and lawyer therefore attempted to train an AI model based solely on his own results. He says he fed all his short stories and novels into the system – and nothing else.

Liu says he knew this approach was doomed to failure.

This is because a writer’s entire oeuvre simply does not contain enough words to produce a so-called large viable language model.

“I don’t care how prolific you are,” Liu says. “It just won’t work.”

Liu’s AI system, built solely on his own handwriting, produced predictable results.

“It didn’t even generate almost any sentences,” Liu says. “A lot of it was just gibberish.”

Yet for Liu, that was the problem. He put that gibberish to good use in a short story. 50 Things Every AI Working With Humans Should Know, published in Uncanny Magazine in 2020, is a meditation on what it means to be human from the perspective of a machine.

“Dinoted concentration crushes dead gods,” is an example of a line from Liu’s story generated by his custom AI model. “A man reached for the torch for something darker than seemed the message,” is another.

Liu continues to experiment with AI. According to him, the technology is promising, but it is still very limited. If anything, he says, his experiences reaffirmed the importance of human art.

“So what’s the point of experimenting with AIs?” Liu said. “The goal for me is really to push the boundaries of what art is.”

Audio and digital stories edited by Meghan Collins-Sullivan.