Viewpoints | Issues and Challenges for Intellectual Property Protection in ChatGPT
Published:
2023-12-18
From the existing technology can achieve the application, ChatGPT can help content creation. ChatGPT powerful function, no doubt to the existing law, especially the protection of copyright and other intellectual property rights, has brought far-reaching influence.
ChatGPT is a language processing model launched by OpenAI, which can be used to generate articles, code, machine translation, question and answer, and other types of content. The first generation of products was released in 2018, and the ChatGPT released in November 2022 belongs to the 3.5 generation of GPT. Each iteration upgrade means an explosive growth in the number of parameters, for example, from 1.5 billion to 175 billion from version 2.0 to version 3.0. It is precisely because of its exponential iterative upgrade speed that the future growth is full of imagination.
Advances in technology often pose new questions and challenges to the use of the law, and ChatGPT is no exception. As a representative of the current artificial intelligence generated content ChatGPT, its principle is not complicated, can be simply understood as a pre-training generation model, that is, first deep learning of a large amount of existing work information, that is, the pre-training of a large amount of information "feeding", and then ordinary users issue instructions, ChatGPT according to the instruction feedback generated content. But unlike some previous models, ChatGPT can continuously answer questions and even correct responses, admit errors, question incorrect premises and reject inappropriate requests. ChatGPT not only follow the instructions, at the same time, the generated content is not simply copied or spliced, but has a certain literary and logical integrity, it can be said that at least reached the level of ordinary middle school students.
Judging from the applications that the existing technology can achieve, ChatGPT can help with content creation, such as writing reports, writing papers, writing poems, writing songs, and can also help users write code, debug programs, organize meeting minutes, make PPT, etc. ChatGPT powerful function, no doubt to the existing law, especially the protection of copyright and other intellectual property rights, has brought far-reaching influence.
"Creation" in the sense of copyright law belongs to intellectual activities, while works refer to "intellectual achievements that are original and can be expressed in a certain form in the fields of literature, art and science". In short, creation belongs to intellectual activities, and works are intellectual achievements. Therefore, the understanding of "creative" activities needs to be based on "originality" to consider.
For the "originality" of the work, the first standard is the objective form standard test, that is, "forehead sweating" and "minimum creativity". As long as it can prove that the work has been done in the process of creation and there is a little creativity in the objective form of expression, it will be recognized as a work. The second is the overall feeling standard. In order to avoid plagiarism, plagiarism and other opportunistic acts as "creation", it is usually based on the dichotomy of "thought and expression", emphasizing the abstract feelings of "ordinary observers" on the whole of the work. Through the "abstraction" of the work, the concept, subject and emotional part of the work that are not protected by copyright law are "filtered" out, and finally whether the content of the expression constitutes substantial similarity is judged through overall observation.
Therefore, the evaluation of "originality" should be carried out from both subjective and objective aspects. The objective form of judgment excludes the work that produces a high degree of "repetition" and the work that lacks intellectual input. For example, although the copy of the original work or the restoration of the old film has invested a lot of manpower, material resources and even "intellectual labor", this process is generally not regarded as "creation". In particular, the author's creative intention should be included as one of the necessary conditions for judging the work. For example, in the famous "Monkey Selfie Case", the court held that the monkey had no subjective intention to create the work and did not meet the requirements of the author of the work. Although the photographer put on a wide-angle lens, set various parameters, and fixed the camera and remote control on a tripod, the monkey's behavior cannot be regarded as "creation".
The next question is whether the generated content of generative artificial intelligence represented by ChatGPT can be recognized as a work.
From the perspective of ordinary users, the instructions issued by them are generally simple sentences or words. Simple words themselves do not constitute works. Although there is a small amount of inspiration behind the selection and combination of words, it is relatively random, and it is difficult to be evaluated as "originality". In the domestic "Artificial Intelligence Copyright First Case" (2018) Beijing 0491 Minchu No. 239 case, the court also held that the user of the software only submitted keywords in the operation interface for search. This behavior did not convey the original expression of the user's thoughts and feelings, and should not be regarded as the completion of the user's creation. The user should not be the author of the intelligent content generated by the computer software, nor can the content constitute a work.
Only in very few special cases, for example, the user provides an instruction specific to the expression during the instruction submission process, and the instruction ultimately affects the ChatGPT's expression content and manipulates it to form content that conforms to the user's creative ideas, it is possible to make the ChatGPT-generated content be evaluated as a work. However, even if the foregoing situation is satisfied, it is still necessary to consider whether the expression is unique and limited. If the data used exist objectively, the quantity is limited, and the manifestation of the results is also limited, this unique or limited way of expression may also be excluded from originality.
From the developer's point of view, since the intelligence of the ChatGPT comes from its large database support, the text of the database content itself may constitute the work. However, the process of generating the content is ChatGPT, or some answers set in advance are directly presented. These processes are automated by the use of "rules" and there is no human intelligence involved. Therefore, even though the data text content within the database may constitute a work, the related ChatGPT generated content should not be evaluated as a new work.
In addition, the rules set by the developer are implemented by training the reward model and reinforcement learning. According to the theory of copyright law, the rule itself belongs to the category of thought, unless it can be embodied in the form of concrete expression. The specific expression (e. g., code) corresponding to the rules set by the developer is obviously not the same or substantially similar to the ChatGPT-generated content at the expression level, and the behavior of the developer setting the rules makes the ChatGPT-generated content not to be evaluated as a work.
At present, there are different opinions on whether the content generated by artificial intelligence constitutes a work in practice. In 2019, "Beijing Filin Law Firm v. Beijing Baidu Netcom Technology Co., Ltd. infringed on the right of authorship, the right to protect the integrity of works, and the right of information network dissemination", the court held that the completion of the creation of natural persons should still be a necessary condition for works under the copyright law, and neither the software developer (owner) nor the user should be the author of the analysis report, but the software user should be encouraged to use and dissemination behavior. In the case of "Tencent v. Yingxun Technology Infringement of Copyright Dispute" decided by Shenzhen Nanshan District People's Court in 2020, the court held that the article in question belongs to the written work protected by China's Copyright Law and is the legal person work created by the plaintiff.
Based on the above analysis, there is no unified judgment rule on whether the content generated by artificial intelligence is protected by copyright law. According to the existing copyright theory, whether it is based on the subject consideration, or based on the generation process of artificial intelligence products, artificial intelligence products themselves are difficult to obtain the protection of copyright law. However, if it is not given any protection, it will undoubtedly affect people's enthusiasm for subsequent development and use of artificial intelligence. Therefore, it is necessary to discuss the idea of protecting the content generated by artificial intelligence.
Since the copyright system has formed a long-term stable theory and system, the revision of legislation should be cautious, at least it is not appropriate to easily adjust the qualitative and interpretation of legal concepts such as "work" and "creation. According to the content generated by artificial intelligence, we can consider interpreting the existing legislation in the way of judicial interpretation. We can make different regulations based on the judgment and qualitative characteristics of different content generated by artificial intelligence. We can even classify and regulate the text content, art content and music content, and explore whether we can consider the proportion of human originality, wisdom contribution and selection in the process of generating content, when a certain proportion is reached, whether the content can be determined to be original and constitute a work in the sense of copyright.
At the same time, we can also consider the characteristics of the neighboring right easy to accompany with technology, the artificial intelligence generation content into the object of the neighboring right, the neighboring right is not copyright, the object of the neighboring right is not a work. Therefore, for the current stage of artificial intelligence, if neighboring rights are used to protect the content generated by artificial intelligence, it is not only unnecessary to identify the generated content as a work, but also unnecessary to give artificial intelligence the identity of the author; and The neighboring rights system belongs to copyright in a broad sense, and its status in copyright law is relatively independent. Compared with other protection paths, it is relatively appropriate and reasonable to protect it by neighboring rights.
Key words:
Related News
Zhongcheng Qingtai Jinan Region
Address: Floor 55-57, Jinan China Resources Center, 11111 Jingshi Road, Lixia District, Jinan City, Shandong Province
