"The emergence of AI agents, the practice of the 'AI Training Supplementary Agreement', and the concerns of writers contracted by online platforms remind us to rethink and define the nature of creation. We must give full play to the value of artificial intelligence while affirming and protecting it. The dominant position of human creators." Recently, Xing Zheng, director of the management committee of Shanghai Jinshen Law Firm and a doctor of law at the Shanghai Academy of Social Sciences, said in an exclusive interview with this newspaper.
Recently, the "Lingka Lingka" creative writing agent launched by the School of Communication of East China Normal University was released in Shanghai Bookstore. In recent years, platform AI-assisted writing tools, represented by Yuewen Group’s “Yuewen Miaobi”, “Writer’s Assistant Miaobi Edition” and Qimao Chinese website’s “AI Assistant”, have shown their prowess. With the upgrading and optimization of technology and the rise of AI writing software on major platforms, human-machine collaboration may become a new trend in creative writing. Is this the beginning of a new era of writing?
Intelligent tools, with their powerful algorithms and data processing capabilities, have brought unprecedented convenience and efficiency to literary creation. However, the development of AI writing also faces new problems. In July this year, the Tomato novel platform attempted to incorporate authors’ original content into corpus training and required the signing of an “AI training supplementary agreement,” which triggered resistance and complaints from some contracted writers. Such incidents not only reveal the problems of how to protect copyright and the rights of original authors in AI writing, but also prompt us to re-examine and think about the nature and ethical norms of literary creation in the era of artificial intelligence.
When AI agent creation meets online literature
Through deep learning and natural language processing technology, AI writing software can simulate human writing styles and ways of thinking, providing authors with a large amount of creative materials and inspiration. The team of Professor Wang Feng, Dean of the School of Communication at East China Normal University, used the method of "domestic artificial intelligence large language model + prompt word engineering + manual post-polishing" to successfully complete the novel "Apostle of Destiny" with more than 1.1 million words in May this year. In October 2023, a team of Shenyang professors from the School of Journalism and Communication of Tsinghua University used generative artificial intelligence to create the novel "The Land of Machine Memories", anonymously participated in the Jiangsu Youth Popular Science Science Fiction Competition, and won the second prize.
What impact has AI writing agents had on online literature? Will AI writing software bring a new model of writing?
At present, many online writers are using AI to assist creation, such as data query, story conception, scene imagination and character characterization. For online literature, which requires physical strength, intelligence and endurance, the intervention of AI can help speed up writing. Jiacheng, a member of the Internet Literature Committee of the Chinese Writers Association and an Internet literature critic, analyzed the special expression methods and production models of Internet literature. He believed: "The elements of 'database writing' such as the types, routines, and memes of Internet literature have become necessary components of the story. Innately capable of being dismantled and reorganized At the same time, the meaning of online literature is relatively simple and straightforward, and it does not pay much attention to the poetic "meaning". These all provide convenience for AI writing. In essence, AI writing still works according to the instructions given by people. Requires resources to be mobilized, texts to be recombined or new texts to be generated.”
AI-generated works remind us to rethink and define the nature of creation. Xing Zheng said: “In this process of thinking and defining, we need to make it clear that AI-generated works are also the fruits of intellectual labor. We must leave enough room for development of technological progress and new creative methods, while also encouraging and protecting Innovative human original works that contribute to the development of cultural diversity.”
Literature is an important way to demonstrate the originality of the human spirit. Will the large-scale application of AI weaken humans' ability and opportunities to express their creativity?
Jiaocheng bluntly stated the problems brought about by AI writing entering online literature: "AI writing leads to the generalization of the subject of literary creation. AI does not have human subject emotions, and all the data resources it calls are human creations. For those who really want to express Literary creation of subjective emotions and in-depth thinking Generally speaking, AI-assisted writing has little value, which is why it has less impact on the traditional literary field. The development process of AI writing reflects the struggle between humanistic values and commercial interests in the literary field, and it is necessary for us to use it. New technologies while safeguarding human values.”
Corpus training and copyright ownership of AI agents
AI agent training usually requires a large amount of data, including public texts, images, audios and videos. Has the corpus used by the intelligent platform obtained the corresponding use rights through legal channels? In recent years, copyright cases surrounding AI-generated content have involved novels, images, sounds and other fields. How to determine the ownership of copyright generated by AI? These issues are not only related to the protection of creators’ rights and interests, but also directly affect the positive development of AI technology.
After the "AI Training Supplementary Agreement" of the Tomato novel platform caused a boycott by contracted writers, Wang Feng was keenly aware of the copyright issues brought about by the new model of AI agent creation. He and his team members had an in-depth discussion about the issues facing the field of large model research and writing. challenge. Regarding the legality of the corpus used by the platform, Wang Feng did not mince words: "Every AI agent faces huge challenges in this regard. The 'Lingka Lingka' creative writing agent does not involve corpus pre-training for the time being. Of course, it will be done later We also have to face this difficulty. We will obtain the training rights of the corresponding corpus through various legal methods to protect the legal rights of the original creator.”
Whether the corpus used to train AI agents has obtained legal usage rights is not only related to the legal operation of the platform itself, but also involves the protection of the rights and interests of original authors and the balance of the entire content creation ecosystem.
In February 2024, the Guangzhou Internet Court ruled that the AI-generated Ultraman infringed copyright. The court held that the plaintiff owned the copyright of the Ultraman series image, and that an AI company generated pictures that were substantially similar to the plaintiff's Ultraman image without permission, which constituted copying. , the act of adapting the image of Ultraman constitutes infringement.
Artificial intelligence is a strategic technology that will lead the future, and my country's artificial intelligence technology is also in a stage of rapid development. Therefore, it is equally important to balance the development of the AI industry and the protection of creators' rights.
Xing Zheng believes: “The legal source of training data is a key issue that needs to be solved in the development of AI writing. The training stage of AI is equally important as the output stage. The infringement of AI works not only refers to infringement in the output stage, but also includes infringement in the training stage. Infringement. "Existing artificial intelligence training is highly dependent on data and requires huge training samples to learn. From an intellectual property perspective, the original author has not explicitly authorized his work to be used for artificial intelligence training, and unauthorized use may infringe upon the rights of the original author.
Do human users who collaborate with intelligent platforms to produce artwork own copyrights? Since the overall artistic creation process is not entirely completed by the user, they can only have a partial share in the work. Faced with this "thorny problem", Wang Feng came up with a constructive solution: dividing copyright into two steps, one is the product, and the other is the declaration and empowerment of the copyright. Wang Feng believes that the relationship between smart writing platforms and users in the future is that of co-creators. “In the final product, the platform and the user share the copyright and use rights of the work. This is a brand-new cooperation model. Of course, in The use of the final product can be negotiated according to the agreement.”
How to solve the problem of copyright ownership of AI-generated works? This requires a forward-looking intellectual property legal system that affirms the copyright protection of generative artificial intelligence "works" and establishes a hierarchical protection mechanism for human original creations and artificial intelligence-involved creations. Xing Zheng analyzed from a legal perspective that our current intellectual property legal protection system must take into account the particularity of artificial intelligence creation. On the one hand, it is necessary to strengthen legal research, improve the relevant laws and regulations system, and clarify the copyright protection standards and rights ownership of AI works. in principle. On the other hand, it is also necessary to strengthen intellectual property protection through technological innovation, and to research, protect and manage AI agent creation models through technical means to prevent infringements. At the same time, it is also necessary to strengthen industry self-discipline and supervision, promote the establishment of industry standards and norms, and provide a strong guarantee for the healthy development of AI technology.