Author Admits to Using AI to Write Award-winning Novel
Rie Kudan won Japan's highest literary award for her novel. Generative AI wrote 5% of it.
At a Glance
- Rie Kudan won Japan's highest literary award for her novel about futuristic Tokyo in which gen AI is everywhere.
- Kudan admits generative AI wrote 5% of it. Literary award groups in Japan are re-evaluating the terms of future submissions.
- Last year, a Chinese professor wrote an award-winning novel generated by AI, using 66 prompts and completed in three hours.
An author who won Japan’s most prestigious book award revealed that around 5% of the content in her novel came directly from generative AI.
Rie Kudan won the 170th Akutagawa Prize for emerging authors for her novel, Tokyo-to Dojo-to (meaning, Sympathy Tower Tokyo), according to Kyodo (Japan) News.
Her novel is set in a futuristic Tokyo in which generative AI is being used everywhere. One of the characters is ‘AI-built,’ which is similar to ChatGPT.
After Kudan’s comments garnered widespread attention on social media, she clarified that she only used AI-generated text in AI-built’s responses.
"While I borrowed (from AI) in parts, I made the appropriate modifications to the story so as not to disrupt the text's flow," she said. "When you read it back, the applicable text barely makes up a whole page, and I think my comments were overblown."
One of the judges of the prize, Shuichi Yoshida, said the subject of AI hardly came up in the selection committee’s discussions but instead it was thought to be a character in the novel.
However, Kudan's admission is making other editors think twice. "Similarly to adding citations, I think we would like to be informed in advance," a veteran editor told Kyodo News. "Realistically, it is hard to discern (if AI was used during the writing process), and some readers may feel cheated if they find out after the fact."
At least one literary award group is already implementing changes.
The Hoshi Shinichi Award for science fiction literature developed specific details for using AI-generated content in submissions: They include a ban on using verbatim text, a requirement to keep records of the process, and other rules.
However, literary critic Akira Okawada told the news agency that "many authors are already using AI to help them come up with ideas for themes or writing structures."
"While we still cannot create superior stories simply by using AI-generated text, we should discuss whether it will (eventually) supersede writers' creativity," Okawada said.
66 prompts in three hours
Last December, journalism professor Shen Yang of China’s prestigious Tsinghua University used 66 prompts in about three hours to generate a Chinese-language novel, complete with title, illustrations and pen name, according to China Daily.
Called The Land of Machine Memories, the novel went on to win second prize in a popular youth science and sci-fi competition put on by the Jiangsu Popular Science Writers Association in Nanjing, Jiangsu province. Out of 200 entries, 18 won second prizes and 14 first prizes.
The novel is about Li Xiao, a fictional neural engineer who loses all her memories about her family. To regain her memories, she goes on a journey to explore the land of machine memories with an AI friend, Neura.
Shen first created an outline, improved the content, sharpened each paragraph’s point and then put it all together. The AI generated 43,000 words, which was whittled down to 6,000 for the winning submission.
Fu Changyi, director of the association’s science fiction committee, knew about the professor’s involvement in AI but did not tell the judges that the novel was generated by AI, as a bit of “an experiment.”
To Fu’s surprise, three of six judges voted for the AI-generated novel, which was enough to win a second prize.
One judge and novelist, Wang Yanzhong, was impressed. "At first I didn't realize that it was created by AI. There is no doubt that AI surpasses many humans in terms of density of knowledge. Compared with the works of beginners, it is outstanding.”
Another judge and fellow novelist, Suo Hefu, was less enthusiastic. The literary quality was “weak” and the story was “disjointed,” he told China Daily. "Despite this, I still gave it a score of 72, whereas some works by humans got only 27 from me."
Read more about:
ChatGPT / Generative AIAbout the Author
You May Also Like