[author]Wang Qian

[content]

The Qualitative Analysis of Content Generated by Artificial Intelligence in Copyright Law

Wang Qian

Professor, School of Law, East China University of Political Science and Law

Abstract: The purpose of copyright law is to encourage creation, and only humans can understand

and utilize the incentive mechanism of copyright law. Therefore, only human’s creative achievements can be protected by copyright law as works, and the “Objective Originality Theory” that argues works do not need to come from humans is untenable. The copyright law may treat a person or legal entity other than the real author as the author, but such a legal fiction is based on the fact that there exists a work created by a natural person. So it is illogical view that the content generated by artificial intelligence constitutes work just because the developer or the user may be recognized as the author by the copyright law. Both developers and users of artificial intelligence cannot directly determine the content generated by artificial intelligence based on their free will, so this content is not created by humans using artificial intelligence as a tool. Due to the different business models of utilizing the content generated by artificial intelligence from utilizing the work of authorship, not recognizing the content as a work will not affect investment in artificial intelligence technology, nor will it violate the concept of the unity of rights and obligations.

Keywords: Artificial Intelligence; ChatGPT; Originality; Creation

Introduction

Marked by the victory of the artificial intelligence Go program Alpha GO over Go world champion Lee Sedol in 2016, the rapid development of artificial intelligence technology in recent years has triggered academic attention. Some AI writing programs, drawing programs, and composing programs have been able to generate content that is formally close to human works, and whether such content can be protected by copyright law as a work has also provoked heated academic discussions. The author has used the process of generating content by AI as an entry point, suggesting that, at least at that time, AI was merely the application of certain algorithms, rules and templates, which were far removed from the author's creations based on his or her unique personality and emotions, and did not satisfy the requirement of originality.

However, the emerging new generation of generative AI seems to compound the above problem. ChatGPT, a general-purpose dialog system that typifies it, "exhibits strong capabilities for dialog comprehension of arbitrary tasks, complex logical reasoning, generation of multi-style long texts, and automatic generation of program code, initially realizing general-purpose cognitive intelligence". ChatGPT's algorithms are more advanced than those of previous AIs. "The neural convolution model it relies on is more complex than traditional algorithmic models, and the analysis of various data elements is more in-depth", and it responds to questions or requests from users based on data from almost all publicly available content on the Internet, and it can generate flexible and diverse new content based on further instructions from users, and "personalize" new content. New content can be generated based on further instructions from the user, and "personalization" is greatly increased. This results in the generation of content that is closer in form to the results of human creativity.

In this regard, I believe that the content generated by ChatGPT is still the result of the direct application of certain algorithms and rules. It is just that it applies more probability theory, through the training of a large amount of data, with a huge amount of calculation to calculate the probability of the distribution of the collocation between words and words, and a reasonable continuation of the user input text. This is still fundamentally different from human creation. However, based on the complexity of the issue, but also in view of the emergence of many different academic views in recent years, it is necessary to explore from other perspectives to ChatGPT as a representative of a new generation of artificial intelligence-generated content can be protected as a work.

1.the "originality objective" defense

The author's old article mainly focuses on the process of AI-generated content as the object of discussion, but this is not because it is not important to determine the inevitable connection between the creator and the work, but rather to avoid starting from the obvious logical starting point of "copyright law only protects the work of man", and to directly conclude that AI-generated content can not be protected as a work of art by the copyright law, so as to avoid the conclusion that AI-generated content can not be protected as a work of art. Instead, we hope to avoid the obvious logical starting point of "copyright law only protects human works" and directly conclude that contents generated by artificial intelligence cannot be protected as works by copyright law, so as to discuss them from other perspectives. However, in recent years, quite a number of theses have rejected the concept of "copyright law only protects human works", arguing that the new philosophical theory of subjectivity categorically abandons the natural human author-centered view of creation, and it is no longer important to know who creates the work and why the author creates it. The work is no longer considered to be the result of the intellectual effort of the natural person author and should be treated "without any question as to whether or not it embodies the intellectual nature of mankind"; a doctrinal analysis of the law from the legislative history and the legal texts is not sufficient to argue that "creation by human beings" is the sine qua non condition for the protection of creations by the copyright law; and human beings also have sufficient reasons to There are also good reasons for human beings to abandon the traditional objective requirement that the object of copyright belongs to the creations of the human intellect. Therefore, it is proposed that copyright law should shift its attention from intellectual labor and creativity to the value of the object itself. In line with this, the judgment of originality should also follow the objective standard, as long as the specific content itself has the minimum creativity and can satisfy the needs of the public, i.e., it can provide the public with the same benefits as the works of human beings, then it should be protected as a work of authorship, and the subjective criterion of considering whether there is a human being behind the specific content should not be adopted. The subjective criterion of "whether there is a human being behind" the particular content cannot be applied. In other words, copyright should protect creativity itself, not only human creativity; since AI "can independently select, process, refine and optimize the required materials, and use different techniques and methods to create new and random content, it is not difficult to see that AI creations can reflect its individual choice and judgment, and can meet the requirement of 'creation'". the element of 'creation'." In other words, since the content generated by artificial intelligence has the external indistinguishability as well as the content similarity with human intelligent creations, it should be given equal opportunity with human intelligent creations, that is, to recognize the originality of the results of artificial intelligence creations. The author collectively refers to this type of view as the "objective view of originality".

There is another way to express the "objective theory of originality", which is to propose that "the content generated by artificial intelligence is not a product that exists naturally in the natural domain, it must be a product generated by artificial intelligence software through certain processing and operation, and it is essentially generated through 'labor' processing", so it is consistent with the basic theory of "labor creates property" to recognize it as a work and provide it with copyright protection. It is essentially a content generated through 'labor' processing", so it is consistent with the basic theory of "labor creates property" to recognize it as a work and provide copyright protection. This view, in effect, argues that the content generated by AI is the product of independent labor to transform the state of nature, and therefore must meet the requirement of originality in form. There is also the view that in order to argue for the "objective theory of originality", "our law does not explicitly stipulate that originality must come directly from a natural person", and the Copyright Law does not mention that a work must be the fruit of human intellect when it defines a work. Therefore, as long as the content generated by AI meets the requirements of originality and reproducibility, it should be recognized as a work under the copyright law, hoping to prove that the "objective theory of originality" has a legal basis.

It can be found that the "objective theory of originality" does not mean that objective criteria should be used to evaluate whether a certain expression meets the requirement of originality, so as to exclude as much as possible the interference of the subjective factors of the judges, but emphasizes that originality can be not originated from human beings, as long as a certain achievement in form meets the criteria for the recognition of the achievement originated from human beings as a work. This is why some opinions also call it "result theory".

In the author's view, the "objective theory of originality" is logically untenable, contrary to the legislative purpose of the copyright law, and lack of legal basis. The so-called "law has provisos, logic has no exceptions", the law can be based on some special factors and make special provisions that superficially violate the logic of the law. Assuming that China's copyright law specifically provides that as long as the content generated by artificial intelligence is "original" in form, it can be protected as a work, then the special provisions of the "legal proviso", can be recognized as a work generated by artificial intelligence as a legal basis. Legal Basis. However, the above argument that the content generated by AI is a work is not based on the "proviso" of the special provision of the law, but on the legal logic of the contingency state. Since formal logic is not subject to exceptions, and a single counter-example is all that is needed to cause the collapse of the formal logic system, if one believes that the source of originality of a work can be recognized outside of human beings, i.e., as long as it has an "objective originality" it can be recognized as a work, and therefore a work can be derived from a non-human being, this conclusion must logically be universal and not only apply to artificial intelligence. This conclusion must logically be universal, applying not only to AI-generated content, but also to other content of non-human origin.

Take the famous macaque selfie case as an example, a macaque monkey snatched a photographer's camera and mimicked the photographer's action by holding up the camera and pressing the shutter, and one of the photos it took was a selfie. By chance, the timing and focus of the photo was just right, and the macaque bared its teeth as if posing. The photographer published the photo on his website and published it in a book, claiming himself and the publisher as the copyright holder. An animal protection organization in the United States sued the photographer, the website and the publisher, asking the court to find that the macaque monkey had a copyright on the selfie, and that the act of displaying the selfie and publishing it in the book infringed on the macaque monkey's copyright, and that the profits from the infringement should be returned. In this case, the court found that the selfie was taken by a rhesus monkey, should the court "categorically abandon the view of creation centered on the author of a natural person", not ask "whether there is a human being behind the selfie", and deny that it was "created by a human being"? whether the selfie is "created by human beings" is "a necessary condition for obtaining protection under the copyright law", but rather follows an objective standard in judging originality, emphasizing that "copyright protects creativity itself, not just human creativity", and "protecting creativity" with "the principle of human creativity". (b) To emphasize that "copyright protects creativity itself, not just human creativity", and to focus attention "more on the value of the object itself" from a "consequentialist" standpoint? If the answer is yes, then the macaque's selfie is certainly "original" in form, since it is not the result of a rehash of any existing photograph, but must be a "product of the macaque's manipulation" and "essentially a 'labor'-produced content". It must have been a "product of manipulation" by the macaque, and "essentially 'labor' to produce the content". At the same time, macaques have intelligence, and the fact that they were able to imitate the photographer's manipulation of the camera and take the photograph indicates that they are intelligent enough, so "it is not difficult to see" that the photograph reflects the macaque's "individual choice and judgment" and itself "has a minimum of creativity. minimally creative". At the same time, the selfie was posted on the Internet and became so popular that it was reprinted on various websites, making the macaque an "Internet sensation". This fully demonstrates that the selfie and the photograph taken by a human being are "outwardly indistinguishable and similar in content", and of course "can satisfy the public's needs" and "provide the public with the same benefits as a human work". ". For this reason, should the Court have recognized the macaque's selfie as a photographic work, so that it would be "afforded the same opportunity as an intelligent human creation", and thus support the plaintiff's claim in the present case?

Certainly, the macaque's selfie cannot be protected as a photographic work. But the reason for this is not that the photo lacks "originality" in a purely formal sense, or that it is not the product of the macaque's independent labor or that the macaque has no intelligence, but rather that identifying it as a work and providing it with copyright protection would be fundamentally contrary to the purpose and spirit of the legislation on copyright law, and would also be inconsistent with the provisions of the law. Even putting aside the ultimate philosophical reflection on human creativity, it should be recognized that the reason why countries enact copyright laws is to encourage authors to create more and better quality works by guaranteeing that authors can receive appropriate remuneration from the exploitation of their works by others.

Any law has a legislative purpose and spirit. Discussing the design of a specific system in isolation from that purpose and spirit is tantamount to building a building in the air. 1710, the first copyright law in human history, the Queen Anne's Laws, stated in Article 1 that: "Whereas of late printers, booksellers, and others have often taken the liberty of printing, reprinting, and publishing books and other writings without the consent of the authors or possessors of the books and writings, which have done them great injury, and have often led to the bankruptcy of them and of their families. bankruptcy. In order to prevent such practices in the future and to encourage learned people to create and write useful books, ...... (therefore it is stipulated that) ...... The authors and their assigns of any book which has been written but not printed and published, and of any book to be written in the future, shall have the exclusive right to print and reprint the same for a period of fourteen years. Article 1 of China's Copyright Law also begins by stating the purpose of its legislation - "to encourage the creation of works beneficial to the construction of socialist spiritual and material civilization ......" The copyright law on the one hand creates a series of On the one hand, the Copyright Law creates a series of exclusive property rights, such as the right of reproduction, distribution, performance and information network transmission, etc. On the other hand, it defines the unauthorized use of a work in a manner regulated by the exclusive rights as infringement, and stipulates the corresponding liability for infringement (unless there are limitations on the right or exceptions to the law), and even imposes criminal liability for certain serious infringements. This ensures that others seek permission and payment from the author prior to copying, distributing, performing or interacting with the work in a manner regulated by the exclusive right, so that the author receives a reasonable income and has an incentive to continue to create.

Of particular importance, the establishment of protection periods and exclusive property rights also takes into account the objective laws governing the creation and exploitation of particular types of works. For example, the creation of musical works often embodies a strong cyclical nature, that is, a large number of authors are in the creative peak period which lasts not too long through the inspiration and passion of the great burst contributed to the excellent works, and after that will often fall into the creative trough, even if the author himself is very hard work, but also very difficult to create the quality of creation with the peak period of comparisons, and before the creation of the excellent works have been repeatedly utilized over a long period of time of a lot of The demand is high for repeated use of previously created works over a long period of time. In response to this, the Copyright Law provides for the right of dissemination, such as the right of performance and the right of information network communication, and a longer period of protection, so that authors and their heirs can be fully rewarded for the repeated use by others of a work that may be small in number, but is widely popular. As a result, authors feel a real sense of economic benefit from the creation of their works. The higher the quality of the work and the greater the market potential, the stronger the need for others to obtain a license from themselves and the more revenue they can make. At the same time, the personal right of authorship under copyright law ensures that others may not remove and tamper with the author's signature on the work, list as authors those who did not participate in the creation of the work, or even, after obtaining permission to utilize the work, distort and tamper with the work in such a way as to change the thoughts and feelings that the author wishes to express through the work to the point of damaging the author's reputation that he wishes to gain. This allows the author to always be closely associated with his work and be recognized as the source of the work that correctly expresses his thoughts and feelings. It is thus clear that the personal and property rights of authors under copyright law enable the author of a good work to achieve "fame and fortune". This not only encourages the authors themselves to actively devote themselves to the creation of their works in order to obtain greater fame and more economic benefits, but also has a good demonstration effect on other people with creative potential, especially young people and their parents, so that they believe that there is a future in creative work. Through this incentive effect, the legislative purpose of copyright law to encourage the creation of works is realized.

It is self-evident that the key to achieving the effect of encouraging creativity through the exclusive rights and protection mechanisms provided for in the copyright law is that the targets of the incentives set out in the legislation are able to understand and utilize the said incentives. Rhesus monkeys and other animals can certainly be incentivized. For example, circus animals are able to perform hilarious or awe-inspiring acts because their trainers make use of the animals' physiological conditioned reflex mechanisms, and provide food and other rewards after guiding them to perform specific actions, thus incentivizing the animals to complete various tasks in accordance with the training requirements. However, animals are unlikely to understand the rights and protections afforded by the law, let alone utilize copyright law to assert their "rights". Therefore, animals certainly cannot be encouraged by copyright law. As for artificial intelligence, the core of which are computer programs, algorithms, models and databases, etc., it is even more unlikely to be "incentivized" by copyright law. Therefore, it is not in line with the legislative purpose of the Copyright Law to recognize the contents generated by artificial intelligence as works and provide copyright protection.

Based on this, it is irrelevant whether or not the copyright law explicitly stipulates that only the intellectual achievements of "human beings" can constitute a work when stating its legislative purpose or defining a work, and that the copyright law encourages the creation of "human beings". What is more, article 2, paragraph 1, of China's Copyright Law provides that "Chinese citizens, legal persons or unincorporated organizations shall enjoy copyright in their works, whether or not they are published, in accordance with this Law." Paragraphs 2 to 4 of the same article stipulate under what circumstances the works of foreigners and stateless persons can be protected by the Copyright Law. The term "work of a legal person or an unincorporated organization" does not refer to a work "created" by a non-human being, but rather to a legal person or an unincorporated organization as the author of a work by a human being (as will be explained below). This clearly excludes animal and AI-generated content from the scope of protection. From this perspective, even if Locke's property theory of labor is used to explain the legitimacy of copyright law's granting of rights, the "labor" in question must be human labor (including labor performed by humans using a variety of tools), and cannot be the "labor" of animals or machines or software. In the case of the macaque selfie mentioned above, the "labor" must also be human labor (including labor performed by humans using various tools).

In the macaque selfie case, the U.S. Copyright Act, which was the basis for the court's decision, did not define "Works of Authorship" and "Author" as used in the Act. The animal rights organizations that brought the lawsuit argued that since the U.S. Copyright Act does not limit what constitutes an "author," an animal that creates an original work of authorship should be entitled to sue as an "author. The U.S. District Court for the Northern District of California held that the U.S. Copyright Act does not expressly extend the concept of authorship to animals, and nowhere does the Act refer to animals. The U.S. Supreme Court and the U.S. Court of Appeals for the Ninth Circuit (the district court's appellate court) have consistently pointed to "human beings" when analyzing authors in copyright law. At the same time, the U.S. Copyright Office has made it clear in its Manual of Practice that "to constitute a work, it must be created by a human being. Results that do not meet this requirement lack copyrightability". Therefore, it only registers works created by human beings, not content produced by "nature, animals, or plants," as in the case of photographs taken by monkeys. The court agreed with the U.S. Copyright Office that the macaque in this case was not an "author" under copyright law and had no right to sue. In the second instance, the United States Court of Appeals for the Ninth Circuit further noted that it could also be inferred from the language of the relevant provisions of United States copyright law that the author must be a "person". For example, the copyright law provides that the author's children, whether legitimate or not, may inherit certain rights under the copyright law. Another example is that upon the death of an author, the author's terminated (licensed) interest is enjoyed by his or her spouse; if the author's children or grandchildren are still alive at that time, the author's spouse is entitled to 50% of the said interest. The terms "child," "grandchild," "legitimate," and "spouse" all imply that the author can only be human. Animals are of course excluded. Animals cannot form marriages, nor do they have heirs who can legally inherit property. Of course, it is impossible for artificial intelligence to have "children", "grandchildren", "legitimate" and "spouse", and naturally, it cannot be called "author". Naturally, it cannot create "works" as an "author".

For the same reason, when Taylor, an American, applied to register his AI-generated painting "A Shortcut to Heaven" as a work, the US Copyright Office refused. The U.S. Copyright Office's Copyright Review Board held that copyright law protects only the fruits of intellectual labor based on the creative faculties of the human mind. The U.S. Copyright Office will not register content generated by a machine or a purely mechanical process without the creative input of a human author, because a work must be created by a human being. The U.S. Copyright Office's recently released Guidelines for Registration of Works Containing Content Generated by Artificial Intelligence reaffirms that "it is well established that copyright protects only the fruits of human creativity" and that "if the traditionally creative elements of a work (for which registration is sought) are generated by a machine, it lacks human authorship, and will not be registered by the Office". not register it".

Thus, the "objective theory of originality" holds that it is not necessary to consider whether the work protected by the copyright law comes from a human being or not, but only whether the specific content is "like" a work, that is, formally "original" and beneficial to the public. "This is in essence a deflation of the legislative purpose and spirit of the copyright law, and an attempt to utilize the "form" of the copyright law without its "spirit", which is not desirable. In the case of Beijing Filene Law Firm v. Baidu, which is known as the first case of AI-generated content copyright infringement in China, the Beijing Internet Court held that "the completion of the creation by a natural person should still be a necessary condition for a work under the copyright law." ....... The analysis report is formed by WK Priority Library using input keywords in combination with algorithms, rules and templates ....... Since the analysis report is not created by a natural person ...... is not a work within the meaning of the Copyright Law". This finding is in line with the legislative purpose and spirit of the Copyright Law and is commendable.

2. the defense of "proposed author"

For the characterization of AI-generated content in the copyright law, there is a class of viewpoints, although recognizing that only the results of human beings can be protected as a work, but believes that "the provisions of the existing law on the civil subject, and does not limit the civil subject to natural persons, with the development of society, inanimate subjects to become a civil subject has become possible, and legal persons to become civil subjects is the best example of the civil subject "; "It is not possible to deny that artificial intelligence has the qualification and possibility to become a copyright subject because it is not clearly stipulated in the law that it can exist as a copyright subject." This type of viewpoint suggests establishing a legal personality in a virtual sense for the intellectual property protection of AI-generated objects, that is, through legal mimesis, mimesis the AI as a legal person so that it has a legal personality, that is, "following the historical logic of legal persons being regarded as authors", mimesis the AI as an author, or mimesis the investors, developers, managers, etc. of AI as legal authors of AI works. In this way, the protection of content generated by AI is the protection of the achievements of "human beings" in the legal sense, thus formally avoiding the legal obstacle that the copyright law could not recognize and protect the achievements originated from non-human beings as works. In the copyright infringement litigation involving the content generated by Dreamwriter, an intelligent writing program developed by Tencent, Tencent, as the plaintiff, claimed that it "published the article in question on its official website in such a way as to indicate at the end that 'this article is automatically written by Tencent's robot Dreamwriter' to express that the article belongs to Tencent. Tencent, as the plaintiff, claimed that it "published the article in its official website with a note at the end stating that 'This article was automatically written by Tencent robot Dreamwriter. The article in question is hosted by the plaintiff, on behalf of the plaintiff's will to create, and by the plaintiff to assume responsibility for the work, the plaintiff should be regarded as the author of the article in question according to the law, the copyright of the article in question belongs to the plaintiff." The author collectively referred to this type of view as "proposed author".

The "proposed author theory" can be further divided into two different directions, one is to propose the artificial intelligence itself as having a legal personality, and the other is to propose the investors, developers and managers of the artificial intelligence as the authors of the content generated by the artificial intelligence in the legal sense. In the author's view, neither direction is feasible.

The first direction tries to make artificial intelligence itself a "person". However, if the law wants to anthropomorphize non-human beings into "human beings" and give them virtual personalities, it must solve a series of problems such as the subject of rights, behavioral capacity and legal responsibility. For example, article 11, paragraph 3, of China's Copyright Law treats a legal person or an unincorporated organization "as an author" in order to allow the legal person or unincorporated organization to enjoy and exercise, in the capacity of an author, all the rights conferred on authors by the Copyright Law, including the use of the work through the issuance of a license. In cases where the work is substantially similar to a prior work of another person and the exploitation of the work constitutes an infringement of copyright, the legal person or unincorporated organization "deemed to be the author" is also liable for the infringement of copyright. If the law grants a virtual personality to a non-human being for the sole purpose of creating a "person" in the legal sense, and does not allow the fictitious "person" to exercise rights or assume responsibilities, such fictitiousness lacks significance and will lead to absurd results. Still taking the case of the macaque monkey selfie as an example, is it possible to formally circumvent the legal obstacle that the copyright law does not protect the achievements of non-human beings by making the macaque monkey taking the photo a "human being", and thus recognizing that the selfie is a photographic work of a "human being" and is protected by the copyright law? If the answer is yes, can the personified macaque monkey be licensed? If the macaque takes a picture of a painting that is still under protection, can the macaque's act be recognized as an act of reproduction of the work under the copyright law and bear the responsibility of infringement that may arise from it? The answer is of course no. Some scholars pointed out that: artificial intelligence science determines that artificial intelligence is not a legal subject, the value of the law determines that artificial intelligence does not have the nature of the legal subject, artificial intelligence does not have the legal relationship subject qualification, that is, there is no right, behavioral ability and responsibility ability. The author fully agrees with this.

The investors, developers and managers of AI are proposed to be the authors of AI-generated content in the legal sense, which reverses the logical relationship in the law. Article 11, paragraph 3 of the Copyright Law stipulates: "If a legal person or an unincorporated organization hosts a work that is created on behalf of the will of the legal person or the unincorporated organization and for which the legal person or the unincorporated organization bears responsibility, the legal person or the unincorporated organization shall be regarded as the author." It is clear from this paragraph that there is a "work" first, and then the legal person or unincorporated organization is "regarded as the author", rather than establishing an "author" for a certain content through legal drafting first. Instead of first establishing an "author" for a certain content through legal drafting, and then recognizing the content as a "work". In other words, "legal persons or unincorporated organizations are considered as authors" is premised on the existence of works protected by copyright law. This is the only way to "follow the historical logic of legal persons as authors". Just as a child is a "human being" only if it is the result of the genetic union of human parents. Whether or not the child is recognized as a "human being" is determined by that physical characteristic, and has nothing to do with whether or not the child can have adoptive parents (i.e., a genetically unrelated couple is considered to be the child's legal parents) in the event of the unfortunate and accidental death of its parents. If an intelligent alien being that looks like a human being were to come to Earth, it would probably lead to a discussion as to whether or not the alien being could be recognized as a "person" under civil law and whether or not it would be recognized as having the capacity to exercise civil rights and to act. However, it is of course logically absurd to argue that since a childless human couple can be appointed to adopt the alien creature and make it a guardian, the alien creature is a "person" under civil law.

Obviously, Article 11(3) of the Copyright Law does not solve the problem of recognizing works. The identification of a work should be based on the legislative purpose of copyright in Article 1 of the Copyright Law, the definition of work in Article 3, and the definition of "creation" in Article 3 of the Regulations for the Implementation of the Copyright Law. As mentioned above, the "objective theory of originality" does not hold, and only human achievements can be recognized as works under the above articles. Returning to the macaque selfie case, of course, it cannot be argued that the macaque selfie is a work just because the photographer whose camera was snatched can be technically fictionalized as the author of the legislation. Since the content generated by artificial intelligence is not created by a human being and by its nature is not a work protected by copyright law, it does not matter who is recognized as the author and who enjoys copyright.

In the aforementioned case of the registration of the painting "A Shortcut to Heaven", the applicant for the registration of the work, Taylor, submitted that the painting, which was "automatically generated by a computer algorithm on a machine", was authored by a "creativity machine" of his own invention but should be registered as a work for hire in the name of the owner of the machine (i.e., Taylor). But it should be registered as a work for hire in the name of the owner of the machine (i.e., Taylor). One reason for this is that the U.S. Copyright Act provides that "in the case of a work for hire, the employer or other person for whom the work was created is deemed to be the author and is entitled to all the rights in the copyright unless the parties have expressly agreed in writing to the contrary". Since this section allows "nonhuman, anthropomorphic persons, such as corporations," to be authors, so can artificial intelligence. This is exactly the same as the first direction of the "anthropomorphic authorship argument" quoted above, i.e., anthropomorphizing an AI as a "person". The Copyright Review Board of the U.S. Copyright Office has held that this view cannot be sustained, as the U.S. Copyright Act provides that there are two types of works for hire, the first of which is when "the work is created by an employee within the scope of his (He) or her (She) employment", and the second is when, for eight types of works, such as test questions, the parties have specified in a signed written agreement that the work is to be treated as a work for hire. be treated as works for hire. In either case, a work for hire needs to be recognized as such under a contract of employment or a contract in which the parties agree that a particular type of work is a work for hire, and it is impossible for a machine to enter into a binding contract with a human being. At the same time, the provisions on works for hire relate only to the identification of the copyright holder and not to the existence of a copyrighted work. The painting for which registration is sought is not a work for hire because it is not the product of the author's creation, nor is it a work created "for hire". This confirms the rule on proposed authorship and is irrelevant to the determination of a work.

3. the "creative tool" defense

Regarding the characterization of AI-generated content, there is a view that AI is a tool for human creation, and the relevant content should be regarded as created by human beings using AI as a means, so human beings are the authors of the content. For example, it is proposed that human creation occupies a major position in the process of AI generation, and the 'labor' of that creation process can be generalized to be regarded as human 'mental labor' and thus human creation; "generative AI is a tool for humans to create works, which is essentially no different from tools such as pen and paper and twigs, and the works of generative AI are also the works formed by humans utilizing technological tools for creation. ...... Generative AI works represented by the works generated by ChatGPT are the intellectual achievements of human beings". The author refers to this point of view as the "creation tool theory".

In the aforementioned case of the copyright dispute over Dreamwriter-generated content, the court should have been influenced by the "creative tool theory". The court held that "the generation process of the article in question mainly goes through four stages: data service, triggering and writing, intelligent verification and intelligent distribution. In the above links, the input of data types and processing of data formats, the setting of triggering conditions, the selection of article frame templates and the setting of corpus, and the training of the intelligent checking algorithm model are all selected and arranged by the relevant personnel of the creative team. ...... The above selection and arrangement by the relevant personnel of the Plaintiff's creative team is in line with the requirements of the Copyright Law on creation, and should be included in the creative process of the article in question. ...... From the analysis of the external expression and generation process of the article in question, the specific expression of the article and its creation process which originated from the personalized selection and arrangement of the creator and was technically 'generated' by the Dreamwriter software satisfy the conditions for the protection of written works under the Copyright Law, and the Court concludes that the article in question belongs to the category of written works protected by the Copyright Law of China. written works protected by copyright law." Obviously, the Court held that Dreamwriter was only a tool created by Tencent's creative team, and that the content generated from it was a work created by the creative team.

According to the "creation tool theory", as long as there is a connection between the intellectual input of the developer or user of the AI and the content generated by the AI, the content should be considered to have originated from the developer or user of the AI, and the AI is only a tool for generating the content, regardless of how remote this "connection" is. regardless of how remote the "connection" is. This argument, which ignores the requirement of "creation" under copyright law, is untenable.

3.1 "Creation" is the direct determination of expressive elements based on free will

Article 3, paragraph 1, of the Regulations for the Implementation of the Copyright Law stipulates that: "Creation, as referred to in the Copyright Law, refers to the intellectual activity that directly produces literary, artistic and scientific works." A work is an expression relative to an idea, and each type of work has corresponding expressive elements. For example, the most basic expressive elements of a written work are the combination of words, the formulation of sentences, and, for novels and dramas, plot design that is detailed to a certain extent; the expressive elements of a work of fine art are the colors and lines that make up the artistic modeling; and the expressive elements of a work of dance are the successive movements, postures, and facial expressions that are used to express thoughts and emotions. "Directly produced ...... The "work" emphasizes the free will of the civil subject to decide on the expressive elements required to constitute the work. The link with the work is so strong that it can only be described as "direct" rather than "indirect". In contrast, "indirect production ...... of a work", which includes the production of the work, is the same as "indirect production". In contrast, the "indirect production of a work", which consists of giving instructions, guidance, advice or suggestions on the concept, style, objectives and methods of creation of a work, does not directly determine the expression that constitutes the work, but requires the author's own choices and judgments, which he or she realizes in a personalized way.

It is important to note that acts that do not "directly result in a ...... It should be noted that acts that do not fall into the category of "directly producing a work" are not incapable of affecting the expressive elements that constitute a work. Article 3, paragraph 2, of the Regulations for the Implementation of the Copyright Law stipulates that "organizing, advising, providing material conditions for, or carrying out other auxiliary work for the creation of a work by another person shall not be regarded as creation." Of course, "organizational work" and "material conditions" have nothing to do with the expressive elements of a work, but "advice" is often related to the expressive elements of a work. For example, after the editorial department of academic journals receives a manuscript, it will often be sent to external review experts for evaluation, and the opinions put forward by the external review experts may be limited to relatively general modification directions, such as suggesting that the title of the article should be appropriately enlarged or limited, and that the overly complicated expressions should be streamlined, and they may also include more specific suggestions, such as pointing out that the argumentation on the core viewpoints of the article has missed a few important arguments (e.g., the important writings of a scholar), and should be supplemented with more specific suggestions. important works, etc.) that should be completed and how the structure of the article should be adjusted, or even suggesting changes to the wording of particular expressions. If the author of the article accepts these comments and suggestions and makes corresponding changes to the content, it can certainly be said that the expression of the article has been influenced by the external reviewer, or even greatly influenced. However, the work of the external reviewer is still considered to be "providing advice" and cannot be considered to have participated in the "creation" of the article that is later revised and published.

The reason for this is that there is a considerable distance between such "advice" and the work. Their impact on the expressive elements of the work depends on whether or not the author adopts the advice and how he or she decides to implement it. In the case of a manuscript submitted to a journal, for example, it is up to the author, after receiving the review comments of the external reviewer, to decide whether they are reasonable or not, and whether or not to comply with them by revising the article accordingly. Even if the author has to adopt others' opinions due to some other factors, for example, the author who is hired to create a work must create or modify the first draft of the work according to the opinion of the hired person, the way to realize the opinion is usually various, and the author needs to choose and judge by himself, and personalized way to reflect in the expressive elements constituting the work. Obviously, for the same manuscript, even if the author adopts the external reviewer's opinion for revision, the result is often quite different from the result of direct revision by the external reviewer himself. This shows that the author's free will still plays a decisive role in constituting the final expression of the work.

Emphasizing that "creation" is the free will that defines the expressive elements of a work does not mean that "creation" can only take place alone. "Creation can be assisted by an assistant, and the creator can listen to the advice of the assistant and reflect it in the final work. For example, if a photographer takes his apprentice to the countryside to take landscape photographs, he can do nothing himself, but instruct the apprentice to place his photographic equipment in a specific place, point the lens in a specific direction, and tell him to set the values for focal length, aperture, and shutter speed, as well as the timing of the shutter press. The photographs taken by the apprentice in compliance with these instructions are the work of the photographer, not the apprentice, because it is the photographer's choices and judgments that determine expressive elements such as composition, lighting, and chiaroscuro. The apprentice may make suggestions in the process, which may also be adopted by the photographer, and thus the photograph may be considered to contain the apprentice's contribution. However, as long as the photographic activity is still dominated by the photographer, and the substance of the photographic work originates directly from the photographer's will, it can only be considered that there exists an indirect connection between the suggestions made by the apprentice and the photographic work, and that the work cannot be "directly produced" by ....... The work".

In the famous "Gele Mountain Martyrs" copyright case, Ye Mou of the Academy of Fine Arts was hired to create the sculpture "Gele Mountain Martyrs", and after completing the first and second drafts, he and Liu Mou, who was responsible for contacting the unit, "worked together to guide the carpenters to produce an enlarged draft (also known as the finalized draft) of the skeleton according to the requirements of the basic form of the first and second drafts. ...... In addition to participating in the pile of the initial form, Liu also participated in the group carving clay sculpture enlargement production under the guidance of Ye Mou and some other work. ...... Liu Mou in the enlargement production through verbal or actual carving made some suggestions. Ye Mou think in line with their own creative intent and expression, give adopt recognition." Later Liu considered himself as the co-creator of the sculpture and filed a lawsuit. The court held that: "the enlarged draft of the group sculpture of the martyrs is completed under the personal participation and guidance of yeh mou, liu mou participated in the enlargement of the production, did some work, through the oral or actual carving and drawing of the production of the proposal, but ultimately whether or not to adopt the recognition of the author of yeh mou. Group sculpture enlargement draft and the first draft compared, in the theme idea, overall structure, basic form, expression method is the same, there is no substantial change. Some of the changes are also in the guidance, participation and recognition of the completion of the Ye Mou, is in the first draft on the basis of the modification and improvement", so Liu Mou on the sculpture does not enjoy the copyright. The judgment is in fact found that Liu only "provide advice", and the sculpture only indirect relationship.

In a lawsuit in Spain concerning the attribution of rights in a theatrical work, the Spanish theater artist Albert applied a model he called "collective creation" in the creation of a theatrical work called "La Torna". Based on a script written by Albert, the writer and director, he guided the actors to improvise some of their lines during each rehearsal, which he then selected and finalized. A dispute arose between some of the actors and Albert over the copyright of the play. The Provincial Court of Barcelona (Spain), referring to the usual practices and views of the theater world, held that Albert, in applying this mode of "collective creation," "retained the power to determine all the expressive elements of the work. He directs the actors to make proposals based on his own ideas, reshaping them, transforming them and determining the parts that correspond to the main idea of the work." At the same time "a small contribution to parts of the work, even if it is indeed 'creative,' is not sufficient to bring about co-authorship." The court found that "the actors were the means of this creative process, and that the final work stemmed from Albert's fundamental and decisive creation," and accordingly ruled against the actors. In this case, the actors' contribution to the theater work should be greater than that of Liu's contribution to the sculpture in the case of "The Martyrs of Gele Mountain", which is also the reason why the formation process of the theater work is called "collective creation". However, the lines improvised by the actors were not only guided by Albert according to his personal and unique conception of the work, but it was also entirely up to Albert to decide how the scattered lines of the different actors were to be selected, combined, modified, and integrated into the pre-created script. In other words, the actors' contributions, while perhaps indispensable, still belonged to the material Albert used to create his work. The act of "directly producing ...... The act of "directly producing the work" comes from Albert, not from the actors.

Although the "direct production of a ...... work" requires the civil subject to determine the expressive elements that make up the work based on his or her free will, the "determination" is not limited to identifying every detail of the work 100 percent, but leaves room for contributions from other sources. It is only the contributions from other sources that do not substantially change or develop the expression of the work and thus cannot be considered as constituting a "creation". For example, when a magazine editor is working on a manuscript and finds that it is one or two lines longer than the intended space, and if he or she is pressed for time and unable to contact the author, he or she will usually select one or two sentences from the original manuscript and reduce them, while preserving the original meaning, in order to keep the length of the manuscript within the space reserved. In such cases, some of the text combinations and sentence formulations of the manuscript are modified by the editor, and there is an intellectual contribution from the editor, but it is clear that these textual changes, which do not change the original meaning, cannot be considered as "creative work".

For example, the designer of a computer game not only creates the elements of the game's visual effects, the static and dynamic images and the art modeling of the characters, animals, scenery, and props, but also sets the logic and rules of the game's operation. Although the specific game continuous screen is generated in real time during the player's actual operation, its essence is the combination of various preset game elements according to the predetermined logic and rules. Even if there is no player's operation, the game will generate a series of consecutive screens in real time according to the same logic and rules when the game is running automatically. Although there is a small difference in the surface of the screen produced by different players' operations and different times of the game's automatic operation, it is impossible to go beyond the game designer's preset, but only a continuous arrangement and combination of various elements. This is like the Beijing 2022 Winter Olympics mascot "Ice Dun Dun" (caricature in the art work) of the overall shape and the characteristics of the various components of the body and the proportionality of the relationship are well known, even if the Olympic Committee released to make a variety of poses of the "ice Dun Dun" There is no squatting image, strictly in accordance with the "ice pier" the characteristics of the various parts and proportionality of a squatting "ice pier" is only the use of drawing skills, any basic level of caricature of the artist without the need to play the artistic imagination and creativity will be able to Any cartoonist with a basic level of artistic imagination and creativity can do so without using artistic imagination and creativity, and cannot form an interpretation of the work, but belongs to the reproduction of works of art. The computer program through the preset game logic and rules according to the player's operation of the various elements of the arrangement and combination, although more complex than drawing a squatting "Ice Dun Dun", but has the same nature, but should still be recognized as a game designer to create a continuous screen embodied in the audio-visual works, that is, the design of the game elements and the game logic and rules of the set The design of the elements of the game and the setting of the game logic and rules directly generated audiovisual works.

3.2 The developer or user of the AI has only indirect influence on the content generated by the AI

The above principle also applies when discussing whether AI-generated content, including ChatGPT, constitutes a work. In order to determine whether the developer or user of artificial intelligence has used artificial intelligence as a tool to "create" the relevant content, it is necessary to determine whether the developer or user has directly determined the expressive elements constituting the relevant content on the basis of his or her free will. In the case of the copyright dispute over the content generated by Dreamwriter, the court held that "as to whether there is any direct correlation between the work of the developer of the Dreamwriter software and the originality of the article in question, ...... is not in question". The court held that "as to whether there is a direct correlation between the work of the developers of Dreamwriter software and the originality of the article in question, ", which is hardly in line with the law. The court also held that the process of Dreamwriter generating the article in question had gone through four stages, all of which were selected and arranged by the relevant personnel of Tencent's creative team (see previous quote), and that "the above selection and arrangement is in line with the requirements of the copyright law on creation, and should be included in the process of the creation of the article in question". This is clearly a conflation of the concepts of "the creative team directly deciding on the expressive elements of the article (combination of words and phrasing)" and "the creative team directly deciding on the operation mechanism of the Dreamwriter program used to generate the article", the latter of which is not the same as the latter. The latter is not a "creation" of the article in question within the meaning of copyright law.

In this regard, the Revised Issues Paper on Intellectual Property Policies and Artificial Intelligence issued by the World Intellectual Property Organization (WIPO) clearly states in the "Definitions" section that "'AI-generated' and 'generated autonomously by AI' are terms used interchangeably to refer to outputs generated by artificial intelligence without human intervention. 'AI-generated' and 'generated autonomously by AI' are terms used interchangeably to refer to outputs generated by artificial intelligence without human intervention. In this case, the AI can change its behavior during operation in response to unexpected information or events. It is important to distinguish this from 'AI-assisted completion' of outputs, which requires significant human intervention and/or guidance." Since the topic of discussion is whether "AI-generated content" constitutes a work protected by copyright law, the content is of course generated by the AI on its own, without human intervention. The above definition itself excludes cases where artificial intelligence is used as a tool for creation by human beings, i.e., where human beings directly determine the expressive elements according to their free will. For example, text input programs such as Xunfei Input Method, Sogou Input Method and Baidu Input Method all have voice recognition systems, behind which is artificial intelligence. They not only recognize the user's voice, but also "intelligently" determine the user's intent based on the context in the event of accents, so that the correct text can be entered. There is no doubt that intelligent speech recognition systems are only a tool for the user (the human being), as the combination of words and phrases that appear on the screen are directly determined by the user according to his or her free will, and do not originate from artificial intelligence. However, this has nothing to do with the current academic debate on whether AI-generated content can be protected by copyright law as a work.

For AI-generated content, AI developers need to design algorithms and models, and use a large amount of data to "train" the AI, and repeatedly correct errors, so that the AI can independently generate content in the face of different needs. However, it is not possible for AI developers to directly determine the content generated by the AI of their own free will just by doing the above. Taking ChatGPT as an example, its developer cannot even predict what kind of questions will be asked to ChatGPT by users around the world every moment, so how can it give answers to uncountable and strange questions in real time by virtue of its free will? If one thinks that the developers of ChatGPT used it as a tool to "create" content (answers to questions), then there is only one theoretical possibility that the developers had already categorized and pre-written answers to hundreds of millions of questions, which is obviously not possible. The answers given by ChatGPT are autonomously generated, i.e., "generated by the AI without human intervention" as stated in the WIPO Issue Paper mentioned above, and cannot be considered to be the result of the developer's creative work using the AI as a tool.

Similarly, a user who gives instructions to an AI to perform a specific task cannot, of his or her own free will, determine the content generated by the AI, and therefore cannot be considered to have "created" the work. Although the user's instructions delimit the direction and field of the content generated by the AI, and the user can select a satisfactory portion of the content generated by the AI, he or she cannot directly determine the expressive elements that constitute the content. This is just as a teacher cannot consider a student's independently completed work to be "created" by the teacher simply because the teacher has given the student homework (i.e., instructions for completing a particular task), given hints or notes, or singled out the best work. The Guidelines for Registration of Works Containing Content Generated by Artificial Intelligence, issued by the U.S. Copyright Office, gives an example of this, stating: "If a user instructs a text-generation technology to 'write a poem about copyright law in the style of Shakespeare,' he can expect the system to generate text that is considered to be a poem, which mentions copyright and mimics Shakespeare's style. However, it is the technology that determines the rhyme scheme, the words in each line, and the structure of the text." The registration guide suggests in this regard that "when an AI technology receives a single instruction from a single person and generates a complex textual, audiovisual, or musical result in response, the 'traditional elements of creativity' are determined and performed by the technology rather than by a human user." ...... According to the Council's understanding of existing generative AI, the user does not apply ultimate creative control over how the system interprets instructions and generates content. ...... It is the machine that determines how these instructions are executed in the output result. ...... When AI technology determines the expressive elements of its output results, the resulting content is not the result of human creation. Therefore the material is not protected by copyright and must be declared excluded (out of the scope of registration) in the application for registration."

The registration case of Zarya of the Dawn, handled by the U.S. Copyright Office in February of this year, provides an excellent illustration of the above point. In that case, Kristina Kashtanova registered with the U.S. Copyright Office the comic book Zarya of the Dawn, which she claimed to have created. But then the U.S. Copyright Office learned that Kashtanova had claimed on social media that the comics in the book were generated by her use of the artificial intelligence drawing program Midjourney. The U.S. Copyright Office examined Midjourney's process for generating the cartoons and found that Midjourney automatically generated four images for the user to choose from based on the user's input of commands (e.g., "Cute Baby Dinosaurs Write Drama Like Shakespeare"), and that it could generate new images based on further commands from the user. The U.S. Copyright Office determined that these images were not the result of human creativity and should not have been included in the registration of the work, which was not known at the time of registration of the comic book, and decided to revoke the original certificate of registration and reissue a new certificate of registration limiting the scope to the results of Kashtanova's creativity, i.e., the content of the text that she wrote and the (original) selection, coordination and arrangement of the text and graphics (i.e., the compilation work). the result (i.e. a compilation work).

Kashtanova disputed this, arguing that she had created all the content of the work herself and that the AI drawing program Midjourney was "only an auxiliary tool". She asserts that the commands she inputs into Midjourney are the "core creative input" to the resulting images, and that in order to obtain images that ultimately satisfy her, she has to continually send hundreds of new commands to Midjourney for the initial images, causing Midjourney to Midjourney to generate modified images, i.e., to produce the perfect result through "trial and error. The U.S. Copyright Office argues that the person sending the textual commands to Midjourney did not "actually create" the final image. The information in the instructions can "influence" what Midjourney generates, but the text of the instructions does not determine a particular outcome. The distance between the user's instructions asking Midjourney to generate a certain content and the actual image generated by Midjourney is too great, and there is insufficient control over the resulting image, nor can the user predict what specific content Midjourney will generate. This distinguishes Midjourney from creative tools such as Photoshop software used by artists, because when an artist uses such image editing or other aids, it is the artist who chooses how to edit or modify a particular image and determines the final graphic content. Midjourney users, on the other hand, lack similar control over the images Midjourney initially and ultimately generates. As a result, the U.S. Copyright Office found that Kashtanova did not use Midjourney as a tool to create and could not be the author of the content generated by Midjourney. The traditional work (composition) elements in the images in question originated in Midjourney, not Kashtanova.

Kashtanova also argued that the Midjourney-generated content was a presentation of "creative, human-created instructions" in the form of images, and that she should therefore be entitled to copyright in those images. In response, the U.S. Copyright Office pointed out that the process of generating images in Midjourney does not guarantee that a particular instruction will produce any particular image. The textual instructions entered into Midjourney are functionally closer to a "suggestion" than an "order," similar to the generalized requirements of an employer who hires an artist to create a painting, as to the kind of content that the employer wishes to create. It is similar to a generalized request from an employer who hires an artist to paint as to what he or she wishes to create. If Kashtanova hires an artist to paint and asks him to paint "a holographic portrait of an elderly white woman named Zarya" and "Zarya has curly hair and stays in a spaceship," and also "a spacecraft that is similar to the spacecraft in the movie Star Trek," then she will be able to create an image that is similar to the "spacecraft in the movie Star Trek," but not the same as the "spacecraft in the movie Star Trek," as well. Star Trek", has an atmosphere or style similar to that of the "spaceship in the movie Star Trek", is a "hologram" "rendered with a graphics processor", has "surreal engines", and has "a holographic image of an elderly white woman with curly hair in a spaceship". "with "cinematic effects" and "super details", and Kashtanova is not the author of the resulting painting. It is the painter who receives these instructions and decides how to realize them who creates the paintings. If Kashtanova had entered these terms into an image search engine, she could not claim that the images given by the search engine were created by her, no matter how close they were to her artistic imagination.

Obviously, merely giving an AI drawing program instructions such as "cute baby dinosaurs write plays like Shakespeare" would only limit the AI to generating pictures that contain the shape of a "baby dinosaur" and the action of its writing. But there are millions of artistic shapes made of color and line that could satisfy this requirement! Just as there are a thousand Hamlets in a thousand eyes, there are a thousand algorithms and data models that can produce a thousand different "baby dinosaurs". These instructions are similar to a teacher giving a student a writing or drawing assignment, and the additional instructions given to the student in response to the AI's initial generation are similar to a teacher's request for corrections to an assignment. No matter how many times the student revises the work before finalizing it, it cannot be said that the teacher used the student as a "tool" to "create" the work that constitutes the written or artistic work. It is not the free will of the teacher that determines the expressive elements that make up the work. To treat the content generated by the generative AI represented by ChatGPT as the same as the result of human writing with pen and paper, i.e., to regard it as a work created by human beings using technological tools, is obviously inconsistent with the facts, and also contrary to the provisions and principles of the Copyright Law.

4.the defense of "incentive investment"

For the viewpoint of providing copyright protection for AI-generated content as works, there are also arguments from the perspective of incentivizing investment in AI technology and promoting industrial development. For example, it is argued that "based on the incentive theory, AI generated content should be protected by the copyright law"; granting the legal attributes of works to AI created content can provide a more favorable legal environment for the development of AI, and stimulate and promote the rapid development of AI science and technology and industry; only by granting copyright protection to AI generated content can we ensure that the relevant subjects can invest their resources in AI technology and industry. Only by granting copyright protection to AI products can we ensure that the relevant subjects can obtain corresponding returns on their resource investment costs, which will in turn stimulate their enthusiasm for further development of AI and AI-generated products, and ultimately promote the sustained and healthy development of society as a whole. On the contrary, if AI-generated content is not given the legal status of a work and provided with copyright protection, and is "allowed to enter the public domain", it will "not be conducive to the development of AI" because it will "harm the economic interests of investors" and "jeopardize the economic interests of the investors". the economic interests of investors", "dampen the enthusiasm of AI research and development and investment, and also deviate from China's current science and technology policy of vigorously developing AI as well as industrial and economic policy, which is not in line with the law of market economy development and restricts the broad prospect of transforming science and technology into productivity" "And thus ultimately harm the economic benefits of the entire market." "From a commercial point of view, AI smart companies have spent a lot of money and technical capital to build highly intelligent AI programs, and it is also against fairness not to protect the 'works' derived from the programs at all"; if AI-generated content is denied a copyright law status, "it will result in the economic value attached to such products not being realized, and the investors behind AI not being able to obtain returns, which will seriously hinder the development of the AI industry ...... that is clearly going to undermine investor confidence."

In the author's view, the current state of development of the artificial intelligence industry strongly reveals the problem with the "incentive to invest" argument. As mentioned above, the U.S. Copyright Office has always insisted that the U.S. Copyright Law only protects human works, and therefore does not register content originating from artificial intelligence. In the case of the registration of "A Shortcut to Paradise", the U.S. Copyright Office has taken a detailed look at the views of the U.S. federal courts and other government agencies since the introduction of computers, and has pointed out that a consensus has long existed on this issue. In other words, the U.S. does not seem to have "provided a more favorable legal environment for the development of artificial intelligence" by "allowing the content generated by artificial intelligence to enter the public domain" and failing to "ensure that the resources invested by the relevant subjects are rewarded accordingly". It fails to "ensure that the relevant subjects receive a corresponding return on their investment of resources" and "stimulate their enthusiasm for the further development of AI and AI-generated materials", and therefore "does not conform to the laws of market economy development, and restricts the broad prospect of transforming science and technology into productive forces. ". However, ChatGPT, which is currently being hotly debated in all circles, was developed by Open AI, another U.S. company invested by Microsoft. This undoubtedly represents a huge advancement in AI technology, and of course huge investments are needed to develop and perfect its algorithms, models and "training" for it. However, for the U.S. AI industry, it does not appear that U.S. copyright law's failure to protect AI-generated content has "harmed the financial interests of investors," "undermined investor confidence," "discouraged investment in AI research and development," or "discouraged investment" in AI research and development, investment" and "seriously impede the growth of the AI industry." Otherwise, how can we explain that ChatGPT, a world-shaking generative AI product (and other generative AI products such as Midjourney) was first produced in the United States? How to explain that other U.S. high-tech companies are also competing to invest in the development of similar AI products?

At the same time, China's Internet companies have also been developing AI products similar to ChatGPT in the absence of a copyright law that explicitly protects AI-generated content as works. Baidu has launched an intelligent chat and mapping program called "Wenxin Yiyin". Tencent and Alibaba have also announced that they are developing AI products similar to ChatGPT. How does the "incentive investment theory" explain this R&D boom in China in the absence of copyright "incentives"?

The problem with the "incentive to invest" argument is that it ignores the different business ecologies of investing in AI and creating works. Taking movies, for example, which also require huge investment to produce, the failure to recognize movies as works and provide copyright protection will inevitably "jeopardize the economic interests of investors" and discourage them from investing in movies. This is due to the way in which movies are utilized to generate economic benefits. Once a film is popular, there is a demand for its reproduction, distribution and dissemination by various means, and these acts are easily carried out by others without authorization. If the right of reproduction, distribution and dissemination under the copyright law cannot be used to protect films and regulate the corresponding acts, the film industry cannot survive and develop without enabling film producers to obtain a reasonable return from the abovementioned acts of exploitation of films by others.

In stark contrast, the content that generative AI, represented by ChatGPT, can generate depends on the user's instructions, which is difficult to predict in advance even for its developers. In line with this, the mode in which the developers of AI derive economic benefits, which is vastly different from the mode in which the producers of movies derive economic benefits, is not achieved by regulating the reproduction, distribution and various dissemination practices of the AI-generated content itself. In the case of ChatGPT, for example, there are a variety of profit models. One is the distinction between free and paid users, with free users having to endure longer waiting times when the volume of questions is high, whereas paid users will be able to access ChatGPT's answers in real time, thus enabling the developers to reap the rewards of the fees paid by paid users. Second, Microsoft will embed ChatGPT into its own products, including Office and Bing search engine, so as to enhance the user experience of these products and attract more users. As Office itself is a paid software, an increase in the number of users will of course bring more profits to Microsoft, and although Bing search engine is a free product, the use of more users will of course lead to more commercial organizations to use its bidding ranking services or advertisements, which will also enhance Microsoft's revenue. Third, Microsoft will also widely license other business organizations to access ChatGPT in their own services and products, even including the application of ChatGPT in robots, in order to provide intelligent services to their users, "the economic value generated by its integration into the development of various industries is even more incalculable," which will generate considerable commercial profits. It can be seen that AI developers utilize the profit model of AI mainly to provide it as a commercial service to individual users and commercial users. As some scholars have pointed out, "The application of all network information technologies such as artificial intelligence, big data, blockchain, cloud computing and so on at the moment is to realize the profitability of the technology, which is also why the capital market is keen to chase the investment in new technologies such as blockchain." And there are a variety of ways to achieve technology profitability, and in many cases licensing the technology itself is sufficient for profitability. Given the business models for realizing the profitability of AI technology, not recognizing AI-generated content as a work and providing copyright protection will not affect investment in AI technology.

At the same time, not providing copyright protection for AI-generated content does not mean that AI is irrelevant to copyright and other intellectual property laws. For example, the computer programs behind AI are certainly works protected by copyright law. The technical solutions involved in AI (excluding the so-called "AI-generated inventions") may also be granted invention patents because they meet the conditions for the granting of invention patents under the Patent Law. These intellectual property protections can certainly be a powerful incentive for investment in the AI industry.

5.The Debate on "Unification of Rights and Obligations"

There is a view that in the training of artificial intelligence, it is inevitable that a large number of other people's works will be utilized, and the content generated by the artificial intelligence according to the data model based on other people's works will sometimes be substantially similar to other people's works. At this point, the developer of the AI may be liable for infringement resulting from the dissemination of the content. Denying the developer of the AI copyright in the content generated by the AI would lead to an imbalance of rights and obligations. That is, only the developer of the AI is held liable for the infringement of the AI-generated content, while the developer of the AI is not allowed to enjoy the copyright of the AI-generated content. This violates the concept of unity of rights and obligations.

The above viewpoint is untenable. Take an example to illustrate: a restaurant plays background music during business hours, can the restaurant obtain the copyright of the musical work, the performer's right of the performance and the recorder's right of the sound recording? The answer is no. However, the act of playing background music in the restaurant is a mechanical performance, which should be licensed and remunerated by the music copyright holder for the performance right according to the law, and should be remunerated to the producer of the sound recording at the same time. If the restaurant acted without the permission of the music copyright owner and did not pay remuneration to the producer of the sound recording, it infringed on the performance right of the music copyright owner and the remuneration right of the producer of the sound recording. Obviously, the restaurant has not acquired any rights under the copyright law for playing background music, but it is liable for the act, so is it not a violation of the concept of the unity of rights and obligations?

The key lies in the fact that the "rights" in the "unity of rights and obligations" are not limited to copyright or other legal rights, but refer to benefits in general. In other words, the acquisition of a benefit is usually accompanied by the assumption of a corresponding obligation. In the above example of playing background music in a restaurant, the restaurant will play background music to enhance the dining environment and attract more customers to come to dine, which is certainly a benefit acquired by the restaurant. In line with this, the restaurant should obtain permission from the music copyright owner and pay remuneration to the producer of the sound recording. This reflects the "harmonization of rights and obligations". By the same token, Open AI, a company invested by Microsoft, has invested huge human, material and financial resources in the development of ChatGPT, not so that its employees can let ChatGPT relieve their boredom at home, but rather to develop a corresponding business model for a wide range of commercial applications. Both Microsoft and Open AI knew that the content generated by ChatGPT, no matter how "original" in form, was unlikely to be registered as a work with the U.S. Copyright Office, but they clearly mapped out the commercial future of ChatGPT by embedding it in Office products and the Bing search engine, and licensing it to other commercial organizations. However, it clearly planned ChatGPT's commercial future, which was to obtain commercial benefits by embedding it in Office products and the Bing search engine, and by licensing other commercial organizations to access ChatGPT in order to provide intelligent services. The commercial benefits are huge, both in terms of incentives for Microsoft and Open AI to invest in the development of ChatGPT, and in terms of the "rights" (benefits) in the "harmonization of rights and obligations". In line with this, developers should also have corresponding obligations. If ChatGPT generates content that is substantially similar to someone else's prior work, distribution of that content may still result in infringement of that person's copyright. Recently, the Writers Guild of America, which represents more than 13,000 authors in the United States, added a clause prohibiting the use of an author's work to train an AI without the author's permission to be used by the author when contracting with a publisher, in order to prompt AI developers to seek and pay the author's permission for the extensive use of the work to train an AI. If the AI developer agrees to pay for this, it would be a sign of "harmonization of rights and obligations".

At the same time, attention should also be paid to the prospect of new technologies to prevent AI-generated content from "copying" the work of others. Since data models and algorithms are already capable of making AI-generated content indistinguishable from human-created works in terms of form, it is not impossible to develop new technologies to identify parts of the content that are substantially similar to other people's works, and to ensure that the AI only learns from other people's styles of creation, embodies their ideas, and reasonably quotes from an appropriate proportion of the content. The concept of "harmonization of rights and obligations" is certainly conducive to the healthy development of AI technology by prompting AI developers to invest more in technologies to prevent infringing content.

Conclusion

In the discussion on whether the content generated by AI can be treated as works, there is a view that although the current copyright law cannot be used to protect the content generated by AI, it is only the actual state, not the contingent state, and the content generated by AI should be included in the scope of protection through the amendment of the copyright system in the future. For example, it is proposed that "how to adjust the existing intellectual property theory and legal system to provide a reasonable and necessary rights protection system for AI text or other AI works, including AI art works and musical works, etc., has become a more urgent proposition at present." "Seeking legal protection for AI-generated works under the system of the Copyright Law can minimize the impact and discomfort that may be brought about by legal amendments, and is in line with the dual laws of the times and legal development."

Copyright law, like any other legal system, should evolve with the times and adapt to the development of technology. However, copyright law is not a law that protects all achievements with economic value under the universal sky. Copyright law has its own specific legislative purpose and spirit, and its original heart and mission is to encourage authors to engage in the creation of literary and artistic works by granting them exclusive rights and providing legal protection for those rights. Only human beings can understand the incentive mechanism of copyright law, and can be motivated by the benefits of copyright law in protecting the fruits of creativity. Thus, "people-centeredness" is the basic spirit of copyright law. Works and human authors have never been separated and cannot be separated in the future. As some scholars have pointed out, technology can never replace the human mind and cannot and should not be equated with human creative thinking. Even if the copyright law also has the function of encouraging investment to a certain extent, and in special cases, the civil subject other than the natural person author is proposed to be the author (i.e., the civil subject who is not the author is "regarded as" the author), or to make the original obtaining of the copyright, but also always based on the relevant results by the natural person based on the creation of his or her ingenuity, can be recognized as the basic human work. The basic premise is that the relevant results are created by a natural person based on his or her ingenuity and can be recognized as the work of a human being. Regardless of the advancement of technology and the development of society, the adjustment, amendment or even reform of the copyright system should bear in mind the original intent and mission of the copyright law.

If there really comes a day when artificial intelligence becomes so powerful that no work created by anyone based on his or her mind and emotions can match the content generated by artificial intelligence, so much so that the world no longer needs human creations, then the right choice for human beings is not to incorporate the content generated by artificial intelligence into the scope of works protected by the copyright law, but to recognize that the historical mission of the copyright law has already ended. because it is no longer necessary to encourage anyone's creativity through copyright law. At that point, the discussion of whether AI-generated content is a work will be even more moot than it is today.