Abstract:
According to WIRED, with the proliferation of artificial intelligence-generated content, a controversy surrounding "whether novels are written by humans" is spreading in the publishing industry. After a horror novel was found to have about 78% of its content possibly generated by AI, and the publisher subsequently withdrew its publication plan, more and more writers began to worry: Even if the work is entirely created by themselves, they may be suspected of using AI because of their writing style, and may even face boycotts and reputational losses.
In March this year, the horror novel "Shy Girl" became the focus of this controversy. There has been a lot of skepticism on social media that the book was mainly generated by AI. Subsequently, the publisher Hachette Book Group canceled the publication plan of the novel in the United States and stopped distribution in the British market.
AI detection company Pangram claims that approximately
78% of the text in the book was generated by AI
. However, novel author Mia Ballard denies using AI to write. She said an editor had made AI-assisted changes to a self-published version of the book and said she was seeking legal action."Wired" pointed out that after this incident, similar accusations began to appear in the publishing industry. An author is being questioned after winning a short story award, while another is embroiled in an AI writing controversy after his crime novel received a multi-million dollar offer from a publisher. Both men deny the charges.
What worries the writer even more is that there is currently no generally accepted method for proving that a work was entirely created by humans.
Best-selling author Andrea Bartz told "Wired" that after the "Shy Girl" incident occurred, she realized almost immediately that it might trigger a "witch hunt" against writers.
“Anyone can accuse anyone of using AI.”
Once allegations emerge, even without solid evidence, the damage may have been done, she said.Baltz himself is also one of the lead plaintiffs in the copyright lawsuit against Anthropic. The lawsuit accused Anthropic of downloading a large number of copyrighted books from pirated websites to train AI models. Anthropic agreed to pay
$1.5 billion
last year Reach a settlement.Faced with increasingly common doubts about AI writing, a group of technology companies have begun to launch software specifically designed to prove the creation process of works.
Tools such as Workings, Blindproof, Wintertext, Human Ink and ObraVera claim to be able to record the author's typing process, document modification history, and even use technologies such as encrypted hashing and timestamps to generate digital "fingerprints" for written works to help authors provide creative records when questioned.
However, these tools cannot truly prove that the work does not use AI.
Pangram co-founder and CEO Max Spero pointed out that even if the software can completely record the typing process, users can still enter the AI-generated content verbatim, or fake the writing process with the help of automatic typing programs.
Furthermore, writers often create in complex ways. They may switch between multiple documents, revise drafts repeatedly, or record voice notes through their mobile phones. These processes are difficult to be completely recorded by a single software.
The developers of ObraVera also admitted that its technology can prove when a certain text was registered and who submitted it, but it cannot independently confirm whether the text was generated by AI.
AI detection tools themselves are also controversial.
Pangram claims that its tool’s false positive rate in the field of creative writing is only
0.01%
. But Mary Rasenberger, chief executive of the Authors Guild, said the association has heard of multiple cases where people have submitted their own handwritten works for testing, only to be judged to have been generated by AI.Raxenberg believes that requiring writers to provide additional evidence that they are human authors is an unreasonable burden in itself. She emphasized that the association does not support forcing writers to use such tools, and reminded authors not to easily accuse their peers of using AI.
Publishing industry expert Jane Friedman points out that publishers and literary agents are unlikely to universally accept authorship authentication systems that require additional fees or are time-consuming. Even if unified standards are established, it is still difficult to avoid certification materials being forged or manipulated.
This distrust has affected the market evaluation of the work.
Baltz revealed that she had heard of writers receiving a lot of bad reviews on the book review website Goodreads simply because their works "read like they were written by AI." Some readers even said that they are unwilling to read works published after 2022 because they are worried that they will be mixed with AI-generated content.
Ironically, Baltz's new novel "Delete Your Account", scheduled to be published in 2027, happens to be based on generative AI. In her book, she deliberately imitated the expressions of large language models to create fictional AI chat records and marketing copy.
Now, she is beginning to worry that these words that deliberately imitate AI will make herself suspicious.
"What should I do if any of the content is judged to be AI writing?"
Baltz said."Wired" pointed out that with the continuous development of AI writing tools and detection technology, the publishing industry is facing a difficult problem to solve:
When machines are getting better and better at imitating human writing, and human words may be misjudged by machines, how can writers prove that their works are from their own hands?
Comments