2024-08-09
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
Why has the content creation field frequently become the hardest hit area for conflicts between humans and AI? Is AI an "assistant" or an "output"? The answers in Tomato Novel are all negative.
@New entropy original
Author: Amigo Editor: Jiuli
ByteDance wants to replicate its past success in the online literature field.
Twelve years ago, Toutiao, an information distribution platform under ByteDance, rose rapidly by relying on AI algorithm recommendations and crawling massive amounts of Internet information that was almost free.UGCUsers generated content and then continuously adjusted the recommendation algorithm to enable content to actively find users, which defeated several major Internet news portal platforms that still followed the traditional content editing and distribution model at the time.
It is also in the content field, but the difference is that ByteDance now not only hopes that AI will replace the role of editors, but also expects it to replace online writers.
Recently, Tomato Novel, owned by ByteDance, was revealed to have added an AI clause to the author contract, intending to authorize the works of its contracted authors and all relevant content data (including work titles, introductions, outlines, chapters, characters, author personal information, cover images, etc.) to Tomato Novel's AI model for training, and the authors are not allowed to sue the platform for infringement.
In the same way, the news content creators who once suffered from Toutiao are now in an even worse situation. Tomato Novel and ByteDance behind it have once again become the "dark horse" in the industry, becoming the first content platform in the country to add AI training clauses to the author's contract. It has to be said that the current practice of online writing platforms is still to divide up the Internet traffic dividend as in the past, but now the Internet has transformed into AI, and high-quality data is even more important than traffic. But the difference is that the seemingly more vulnerable online writers are more daring to fire the first shot than the news media platforms in the past, and even forced Tomato Novel to give in and delete the original agreement.
01
Not long ago, there was also a conflict between humans and AI. The battle between the Carrot Express driverless car and Wuhan taxi drivers and online car-hailing drivers attracted countless people's attention. One of the representative views was: "Cars replaced horse-drawn carriages, and carriage drivers could drive cars. Now driverless cars are going to replace taxis, so what can taxi drivers do?" Behind this is an old question that is often mentioned - will AI replace humans?
Especially in the overcrowded and increasingly involuted online car-hailing industry, the contradiction between humans and AI is even more prominent. One netizen joked: "In the great new era, which one is more suitable for young people with dreams, couriers, food delivery boys, or Didi drivers?" It can be seen that couriers, food delivery boys, and online car-hailing drivers have become a reservoir for various types of idle people in society.
According to the "Platform Workers" 2023 Mid-Year Consumption Report released by Tsingshan Capital, the total number of registered online ride-hailing drivers in my country has exceeded 100 million, and there are still more than 20,000 new drivers registered every day. If based on the statistics of the Ministry of Transport on the issuance of online ride-hailing driver licenses, the number of online ride-hailing driver licenses in 2022 increased by 32.6% year-on-year, and in 2023 it increased nearly 5 times compared to 2022.
Judging from the volume, although the battle between online writers and Tomato Novel AI is far less popular than Wuhan taxi/Didi drivers and Carrot Run, if we look at the number of people engaged in the industry, online writers can actually be ranked alongside couriers, food delivery boys, and online car-hailing drivers. According to the 2023 China Online Literature Development Report by the China Audio-Visual and Digital Publishing Association, by the end of 2023, the total number of resident authors on my country's online literature platforms has reached 29.2943 million, and most of them are born in the 1980s and 1990s, who are in their prime and are the main force in creating social value.
Considering the dominant market position of Tomato Novel in the field of online literature, the conflict between online writers and AI is bound to break out. According to data from research organization Quest Mobile, Tomato Novel is currently the platform with the most online literature readers on the Internet. As of December 2023, its monthly active users have reached an astonishing 192 million, far exceeding the second and third place Qimao Novel with 90 million and Zhangyue with 63 million.
The Tomato Novel Agreement requires authors to provide their hard-coded content to AI model training almost free of charge, causing many online writers to worry about "raising a tiger to cause harm". If the AI model of the online writing platform can automatically generate online articles in the future, the authors will be "abandoned after their work has been done" and will have no "way to survive". The scary thing is that such terrible speculation is becoming a reality.
According to netizens, a suspected AI author named "Jiang Yuan Storyteller" has appeared on the Tomato Novel website. This account has put more than 200 works on the shelves in just three months. The speed and frequency of updates are abnormal, which is obviously beyond the scope of normal authors. Ordinary online writers who can update 10,000 words a day are already outstanding. Although Tomato Novel declares that "it has not published any works written purely by AI, and will not use AI writing capabilities against the author's personal wishes", the AI account that can generate tens of thousands of words in minutes has obviously made the authors panic.
Does AI assist authors in their creation or replace them?
02
Although AI cannot create original works, it is full of "routines". In the online literature world, the most popular ones are often those with explosive and exciting points, and the plot is intensively advanced. These "exciting points" require the author to achieve them through "settings" such as upgrades, cheat codes, time travel, and harems. The plot is even more full of routines, which has led to the formation of different schools of online literature according to certain fixed routine settings, such as: system flow, infinite flow, farming flow, spiritual recovery flow, mortal flow, etc. Fixed "routines" are exactly what AI is good at. As long as the AI model learns a large number of routine template online articles, it can generate batches of plots that conform to the "routines" based on the prompt words and rules of the above-mentioned various styles of online articles.
In addition, on free online literature reading platforms such as Tomato Novel that make profits by inserting advertisements, most free users do not have such high requirements for the quality of online literature plots, and only need "large quantities to satisfy their hunger". Therefore, the fixed model and online literature plots provide Tomato Novel with a breeding ground for AI online literature generation for users who do not have high requirements for quality.
The Tomato Novel agreement, on the one hand, did not compensate the authors for the relevant AI training, and on the other hand, if the authors later discovered that AI used similar plots and character names in their works, they would no longer be able to protect their rights due to the agreement. This is equivalent to a portion of the profits that the author should have owned being transferred to the platform. However, under the joint struggle of all parties, Tomato Novel began to make concessions, stating that "for authors who have signed supplementary agreements or signed contracts that include AI clauses, if they have any concerns, they can provide feedback through the backend, and channels will be opened to assist in the cancellation of relevant AI clause agreements as soon as possible."
But it is obvious that the original intention of online writing platforms has begun to change, from "developing auxiliary tools that can help authors significantly improve writing efficiency and reading experience" to using AI "supply side" to revolutionize the lives of online writers. After all, online writing platforms are not only the judges of traffic and online writing recommendations, but also the financial backers behind the authors. It is really difficult for online writers whose lives hang on the platform to feel at ease when the platforms develop so-called AI writing assistants.
Online literature platforms want to use AI to reduce costs and increase efficiency, but this will inevitably affect the income of creators. According to the 2023 financial report of China Literature Group, one of the three major online literature platforms, its total annual revenue is 7.012 billion yuan, of which content costs are 1.646 billion yuan, including copyright and author remuneration, accounting for 19.8% of the group's total revenue. This also means that online literature platforms need to give one-fifth of their revenue to authors every year.
Let's take a look at the composition of various costs of China Literature Group. In addition to the content cost, which accounts for the bulk of the expenditure, the platform distribution cost is 808 million yuan, accounting for 11.1% of revenue; the production cost of TV series, web series, animation and movies is 631 million yuan, accounting for 8.7% of revenue; the amortization cost of intangible assets is 148 million yuan, accounting for 2.9% of revenue, etc. It seems that the company can use AI to significantly reduce costs and increase efficiency. The only two options seem to be content costs and the production costs of TV series, web series, animation and movies, and these two options happen to belong to the field of content creation.
03
The field of content creation is indeedAIGCIn March this year, a16z, a well-known American technology venture capital company, counted the top 50 AI applications in terms of monthly visits. Among them, AI product applications in the fields of content editing, content production, and emotional companionship accounted for nearly 70% in total, while content editing and content production are the tracks with the most concentrated AI products, accounting for more than 50% in total. This means that more than half of the 50 most popular AI applications in the world are related to content generation.
According to data from Quantum位, AI writing applications are also the most numerous products in the domestic AIGC field. In fact, it is not only Tomato Novels, but also various online writing platforms have already launched their own AI writing applications. On a deeper level, there are shadows of Internet giants behind these online writing platforms. This is no longer a simple competition between online writing platforms, but an investment bet by Internet giants in the future AI era.
For example, the AI assistant of Qimao Novels is backed by Baidu and Baidu Wenxin Yiyan Big Model; Yuewen Miaobi and Writer Assistant Miaobi Edition of Yuewen are backed by Tencent Holdings, the major shareholder behind the group. Tencent owns Tencent Hunyuan Big Model, as well as domestic big model unicorns.Dark Side of the MoonHuge investments from Alibaba and Kingsoft’s office platform WPS have also launched AI writing applications; not to mention ByteDance and Doubao behind Tomato Novel. This is why AI and human creators have the most frequent and fierce conflicts in the field of content creation, whether at home or abroad, and in the fields of painting, text, pictures, sound, video, etc.
Although Tomato Novel, which was the first to try this, has made concessions under pressure from all sides, and online writers seem to have achieved a temporary victory, if we look at the world, human writers are not always the winners in conflicts with AI companies. First of all, regardless of whether the company that uses the content of human creators is AI or the Internet, from the perspective of copyright disputes, content always seems to be innovated into different forms for dissemination, and the copyright of creators often gives way to the innovative application of technology.
From Google Books, which was established by Google and collected a huge amount of physical books and merged e-books into Google's search function, to Toutiao, which captured "free news" from various institutional media on the Internet, the former won the lawsuit in the past decade based on the principle of "fair use", and the latter compensated through traffic revenue and purchased some news copyrights after the platform became bigger. Both of them obtained massive content at a relatively small cost.
In the era of artificial intelligence, the most representative conflict between humans and AI is the New York Times andOpenAIThe other is Hollywood’s ambiguous attitude towards AI. On December 27 last year, The New York Times suedChatGPTOpenAI, the company behind the company, violated its copyright, used its content for AI training, and generated the same "verbatim excerpts" when generating results. OpenAI responded by "cooperating with news organizations and creating new opportunities," saying that this move brought traffic to news publishers and had certain public welfare attributes of the "fair use" principle. So far, the lawsuit between the two parties has not yet been concluded.
Hollywood's attitude towards AI is even more divided, with many practitioners trying to use Sora, Wonder Studio,Midjourney, Stability and other AI tools, and even start-ups, in order to seize the initiative. Similarly, many actors are worried that their images, portraits and voices will be abused by AI companies, and they actively sue to protect their rights. It can be seen that Hollywood hopes to embrace AI to significantly reduce costs, and also hopes to firmly grasp its own copyright interests while AI is rampant.
At present, the formulation of AI-related laws is still being actively explored around the world. For example, the European Union recently implemented the world's first comprehensive regulation on AI, the "Artificial Intelligence Act", which clearly states that audio, video, text and image content generated by AI can be detected as AI-generated content; a proposal in the United States, the "AIGC Copyright Disclosure Act", requires AI companies to disclose copyrighted works used in their training; and the "Interim Measures for the Management of Generative Artificial Intelligence Services" implemented in my country last year emphasizes that algorithm data and platform advantages must not be used for unfair competition and monopoly.
rightFor technology optimists, this AI copyright issue may be just a minor problem brought about by a new round of diffusion of new technologies. But for the hundreds of millions of online writers who rely on online writing to receive "minimum living allowances", even a grain of sand in the AI era is of great weight.
Collect and organize the latest news information and make it available for free for everyone to read
