Tomato Novel Tests AI-Assisted Writing Feature: How AIGC Will Impact the Online Literature Business
-
According to online literature authors, Tomato Novel's Tomato Writer Assistant has begun internal testing of AI writing tools. Some social media users revealed that as early as November last year, Tomato Novel had already opened AI-related features to high-level authors.
The author mentioned that the related features are currently fully functional. However, Tomato Novel has issued a notice through the author backend, requiring authors not to disclose the feature interface or the content generated by the tools to the public.
Around January 19 this year, Tomato Writer Assistant conducted a questionnaire survey through the backend, asking authors whether they had used or were willing to use AI to write novels. According to responses from a Tomato Novel author contacted by New Perspective: "I have indeed used Bing and ChatGPT to assist in writing novels, primarily through conversational exchanges to validate certain inspirations or viewpoints. I'm very willing to try more AI-assisted methods, so I answered 'yes' to all survey questions. On January 23rd, I obtained beta testing access and immediately tried it out."
Below is the functional list compiled by New Perspective based on the author's verbal description:
Using AI for content creation falls under the AIGC domain, but Tomato Novel isn't the only online literature platform currently offering AIGC tools. The "Seven Cats" platform provided authors with "AI Assistant" and other auxiliary writing features in 2023, developed in collaboration with Baidu's "ERNIE Bot". Additionally, China Literature Group's large model 'Yuewen Miaobi' and its derivative 'Writer Assistant Miaobi Edition' had already opened for internal testing before September last year. Authors can use this tool for character or scene development, though it hasn't been made available to all of China Literature's authors.
However, since different online literature platforms have distinct reader expectations and business focuses, the impact of their AIGC applications based on large models naturally varies across platforms.
"It's somewhat helpful in the short term, and might become better with further iterations," author Xiao Liu revealed in an interview with New Position. "The biggest issue is that the generated content often lacks coherence with the original text's intrinsic elements. The most prominent problem is the inconsistency in the protagonist's emotions or personality. For instance, while the protagonist might be a complex, contradictory entity in my original content, the AI-generated portion suddenly reduces them to an NPC-like character. Both authors and readers can easily spot which parts of a novel are AI-generated. Moreover, during expansion or rewriting, it only generates based on selected content without considering the context." Regarding the sidebar functionality, Xiao Liu commented: "The AI naming feature is somewhat redundant. Since character names are a fundamental part of the setting, I find it faster and more personal to come up with them myself. Though it might be useful for generating names of minor characters like passersby A, B, or C."
"Currently, the inspiration generation feature is the most useful. I can input my novel's main settings, and some of the generated plots align well with Tomato Novel readers' preferences, which is an advantage over Bing or ChatGPT... But it's only suitable for generating less important side plots," said Xiao Meng, another writer specializing in fantasy novels, to New Perspective.
However, AIGC writing isn't something that can be used right out of the box. According to Xiao Meng, AIGC "has a certain learning curve and requires understanding how to input instructions more effectively." According to observations by 'New Position' and interviews with authors, Tomato Novel makes it "relatively easy to get signed," but "it's hard to gain exposure early on—authors need to write over 80,000 words before entering the recommendation system." The Tomato Novel author backend shows that "recommendation" means: first targeting a subset of users to display the work, then continuously expanding the audience based on reader engagement metrics. The algorithm adjusts recommendations dynamically according to reader behavior.
This indicates that on the Tomato Novel platform, readers hold significant influence over content direction. If AIGC (AI-generated content) features are successfully integrated with creators in the future, web novel production could more rapidly adapt to reader preferences, leading to faster shifts in taste and creating a more agile content iteration cycle.
The aforementioned Qimao platform, which incorporates Baidu's "Wenxin Yiyan" large language model, is known as the "guaranteed income leader" in the web novel industry. Here, "guaranteed income" means authors can earn fixed payments by meeting daily word count requirements under the platform's incentive program, regardless of their work's actual readership numbers. For web novel platforms with higher guaranteed income, AIGC tools can help authors better secure their 'guaranteed wages'. However, according to feedback from some authors on social media, "it's becoming increasingly difficult to obtain guaranteed wages on certain novel websites." New Perspective speculates that as AIGC makes content production easier, platforms may raise the thresholds for their guaranteed incentive programs.
AIGC will also reshape the business model of the entire cultural and creative IP industry chain. In July last year, when China Literature unveiled the industry's first large model "Yuewen Miaobi" and "Writer Assistant Miaobi Edition" at its creation conference, it stated that this model would establish a "new infrastructure" for web novel creation, including author services, data operations, and technical tools. Moreover, it aims to build a new integrated ecosystem for IP upstream and downstream, with the core being using AIGC to bridge "content + platform" to enhance IP incubation and ecosystem efficiency. If Tomato Novel's strength lies in using AIGC tools to swiftly bridge the gap between readers' ever-changing preferences and online literature creation, then Yuewen's advantage is in utilizing AIGC to better facilitate the complete incubation of IP from novels to films and derivatives.
The requirements for creative ability in online literature are vastly different from those in other literary fields, which leads to entirely different attitudes towards AIGC across various writing domains. The first field in China to strictly reject AI-generated content (AIGC) in creative works is science fiction. Known as the 'Whampoa Military Academy' of sci-fi literature, Sci-Fi World magazine announced in its April 2023 issue: "This magazine does not accept submissions generated by artificial intelligence (AIGC). If undisclosed AIGC submissions are discovered, the author's works will be permanently banned. We firmly believe that the fruits of human imagination possess unique and irreplaceable value, far beyond what AIGC can achieve. We also encourage sci-fi creators to expand the boundaries of human imagination through scientific, forward-thinking, and original creations."
As the first magazine to publish original works of popular sci-fi IPs like The Wandering Earth and The Three-Body Problem, Sci-Fi World holds significant influence in the industry. Its stance will likely shape the sci-fi literary community's attitude toward AIGC.
This trend is not limited to China. Since February of last year, American sci-fi magazines such as Clarkesworld and Asimov's Science Fiction have even temporarily halted submissions due to the proliferation of AIGC. This phenomenon is quite interesting. As the literary field most acquainted with AI, science fiction literature actually exhibits the strongest resistance to AIGC. However, this has little to do with whether the science fiction genre is conservative or not. Online literature belongs to the category of daily consumables, requiring authors to produce works that are "fast, voluminous, and cater to readers' tastes." Science fiction, on the other hand, is deeply intertwined with technology and philosophy, demanding that authors create logically rigorous settings and plots with foresight and imagination.
Therefore, it can be anticipated that, compared to other literary fields, AIGC's application capabilities might have greater complementarity with the core competencies required for online literature creation. Regarding the most contentious copyright issues in the web novel industry, these problems did not emerge with the advent of AIGC. Controversies such as "plagiarism" and "plot borrowing" have persisted for a long time, involving conflicts between different creators as well as between creators and platforms.
However, this is not an issue unique to the web novel industry. On December 27, 2023, The New York Times sued OpenAI, the owner of ChatGPT, and Microsoft, alleging that OpenAI used its copyrighted content to train the system without permission. The companies were accused of being responsible for "billions of dollars" in damages. The lawsuit highlighted instances where ChatGPT generated "verbatim excerpts" from The New York Times articles when asked about current events, which are otherwise inaccessible without a subscription.
On January 8 of this year, OpenAI responded on its official website, stating that The New York Times' lawsuit lacked a complete basis. It further argued that "ChatGPT displays real-time content with attribution, offering news publishers new ways to connect with readers," implying that the platform drives traffic to news publishers. Both parties have their own arguments, and there is still no verdict in this lawsuit.
In fact, lawsuits against AIGC tools including OpenAI, such as Stability AI and Midjourney in the text-to-image field, have never stopped. The conflicts between AIGC tools/platforms and creators are entirely new issues of this era, which cannot be resolved overnight. However, the outcomes of these lawsuits do provide some reference for how online literature platforms can better regulate their use of original online works.