The impact of AI-generated content in the field of patent law and its response

"The so-called technical documents generated by text editing technology shall not be protected as technical solutions under the Patent Law, nor should the novelty and inventive step of the patent application be evaluated as prior art in the examination of patent applications."

Author | Meng Rui is a partner and patent agent of Grandall Law Firm (Nanjing).

Edit | Bruce

1. Artificial intelligence generated content in the field of patents: formalizing patent texts

As AI is increasingly applied to content creation, the concept of AIGC is on the rise. Artificial Intelligence Generated Content (AIGC) refers to the technology of generating text, pictures, sounds, videos, codes, etc. based on algorithms, models, and rules. AIGC can generate content in different styles and forms according to different scenarios and goals, or combine different types of content to generate richer and more diverse content. At present, the main application areas include media, e-commerce, film and television, entertainment, etc. [1] In fact, in the field of patents, the application of intelligent technology to automatically generate formal patent texts has long existed. In practice, websites such as Cloem, All The Claims.com, and All Prior Art.com use artificial intelligence algorithms to generate formal patent text. The Cloem website claims to assist patent applicants in drafting various claim texts. [2] Cloem's automated claim drafting system uses a variety of algorithms such as NLP, semantic techniques, automated reasoning, and text mining to draft thousands of patent claims according to predetermined rules to cover potentially novel inventions. All The Claims.com is a sister website to All Prior Art.com, which independently generates thousands of patent claims and descriptions through linguistic analysis and random combination of patent texts published in the database of the United States Patent Office, and then timestamps the generated claims and description texts and publishes them on the website. All The Claims.com is All Prior Art.com so much purpose is to make the technical solutions disclosed in the formal patent text become prior art, and to prevent patent trolls (NPEs) from preemptively appropriating these technologies in a patent manner. [3] There is also an AI-driven service called Specifio that provides clients with the service of drafting patent specifications, and customers can simply email the claims to a specialized automated system to obtain a complete patent application text in just a few minutes, even including numbers and diagrams. [4] Specifio produces patent application documents through analysis, extraction and sorting, and the completeness of patent application documents can reach about 90%, and lawyers only need to spend another hour to revise and polish them as formally submitted patent application documents. Similar technologies have emerged on the mainland that can automatically draft patent application documents. The invention patent application with application number 201810200739.5 records "a patent drafting method and drafting system based on artificial intelligence", which identifies and analyzes mechanical drawings and identifies multiple mechanical parts and their connection and position relationships; According to the technical disclosure document, a number of mechanical parts are screened and the necessary technical characteristics are marked; Mechanical components marked as necessary technical features and their connection and position relationships are generated according to the neural network writing module to generate independent claims or dependent claims. Through the recognition of mechanical drawings, the neural network writing module automatically generates independent claims, which greatly reduces the workload of patent agents.

The text generated by the above-mentioned various content generation application scenarios is not an officially published patent application document or patent document, but a natural language readable text generated by the content generator according to a certain patent file format, and has a certain patent file format, which is referred to in this article as a formal patent text.

2. The impact of formalized patent texts on patent applications and protection

It has been concluded that the mass generation of patent texts by artificial intelligence machines has the following negative effects on patent application and protection:

First, it brings a high degree of uncertainty to patent rights, increases the external burden of patent examination, and reduces the quality of patent examination. If AI machines generate patent texts in large quantities, it will expand the number and content of prior art, further reducing the comprehensiveness of prior art searches in patent application examination. [5] The result is to affect the stability of the rights of granted patents and increase the uncertainty of patents. In addition, patent uncertainty can also affect inventors' enthusiasm to use patents to protect inventions. [6] In addition, if a formal patent text is filed for a patent application, it will lead to an explosion of patent applications, and the explosive growth of prior art literature and patent applications will increase the work pressure of the patent examination department, and eventually lead to a decline in the quality of patent examination. [7]

Second, the excellent innovation ability of AI machines will lead to the failure of current examination standards such as novelty, inventive step, and practicality. [8] Artificial intelligence machines have super text replacement and data mining functions, which can accurately compare the technological achievements generated by themselves with existing technologies with the help of patent databases, so as to effectively circumvent prior technologies and meet patent granting conditions such as novelty and inventive step.

Third, the ability to invent and create AI machines will be used as a tool for improper industrial competition, jeopardizing innovation. Filing a formalized patent text for a patent application exacerbates the patent jungle, with potential innovators with little freedom to operate or prohibitive licensing fees. [9]

This paper argues that problems such as the difficulty of achieving comprehensive prior art search, the high degree of uncertainty of patent rights, the large external burden of patent examination, and the hindrance of innovation by patent jungles have existed in patent practice for a long time, and are not problems that only appeared after the development of artificial intelligence content generation technology. Patent authorities in various countries are constantly adapting patent policies and measures to address these issues. For example, the Invention Act introduced in the United States in 2011 focused on solving these problems to some extent. Of course, artificial intelligence content generation technology makes these problems more prominent. In addition, text search technology is an important application field of artificial intelligence technology, and patent offices in various countries, including China, have applied artificial intelligence-based search technology to patent examination practice. It is foreseeable that with the continuous development of artificial intelligence technology, the search of existing technologies will develop in the direction of becoming more and more comprehensive, rather than more difficult and incomplete. AI-based search technology can also free examiners' energy from prior art literature search work, and focus more on the examination of patent application documents, thereby promoting the improvement of patent examination quality. As for the AI machines that are concerned about existing research, they accurately compare their technological achievements with existing technologies to effectively avoid the problem of existing technologies improving the possibility of authorization, confusing the relationship between invention and patent application layout and other patent legal affairs. Comparing technological achievements with prior art to effectively avoid prior art and enhance the possibility of authorization, involving the interpretation of prior patent claims and the judgment of novelty and inventive step, these activities have gone beyond the scope of technological invention and creation, and belong to typical patent legal service matters, usually completed by professional legal service personnel such as patent attorneys or patent agents. Discussing relevant issues on the basis of advocating that artificial intelligence machines have patent legal affairs capabilities such as patent application layout and inventive step judgment is too advanced. This paper argues that how to treat patent texts formed by people based on text mining and editing techniques in patent examination and protection has more urgent practical significance.

3. Be cautious about the patent law attributes of formal patent texts

For the AI-machine-generated patent texts introduced in the first part above, they can be divided into two categories based on the presence or absence of technological invention-creation activities. In the first category, human inventors have obtained technical solutions for which patents are to be applied for through creative labor, and artificial intelligence machines are used as application document writing tools to assist in drafting patent application texts, such as the aforementioned Specifio and "an AI-based patent drafting method and drafting system". The second category is completely composed of artificial intelligence machines using big data mining and analysis methods, using text editing technology to generate so-called patent texts, without actual invention and creation activities, such as the aforementioned All The Claims.com and All Prior Art.com websites. This paper argues that for the above-mentioned patent application text generated by the first type of artificial intelligence machine, the role of the artificial intelligence machine is only to assist the inventor to convert the technical solution that has been proposed by the inventor into a patent document, so the disclosed technical solution still belongs to the technical solution under the patent law, and can be applied for as a real technological research and development result, and the patent application text can be used as prior art to evaluate the novelty and inventive step of other patent applications after publication. For the second type of patented text generated based on text mining and editing techniques, should the so-called technical solution recorded in it be granted a patent even if it meets the novelty and inventive step? Can it be used as a judgment of novelty and inventive step in other patent applications? It is still necessary to analyze the nature of technical solutions in patent law.

According to the needs of human beings to continuously improve their own living and development conditions, technical solutions are summarized by people from practice, or invented by people under the guidance of scientific theories, and confirmed and applied through practical tests, so as to use the resources of the external world to expand people's own capabilities. Based on the occurrence mechanism of science and technology, the expansion of science and technology to human cognition and ability mainly focuses on "obtaining information, refining knowledge, generating intelligent strategies and transforming into intelligent behaviors under the conditions of given problems, constraints and goals". For now, the most fundamental difference between human intelligence and machine intelligence is that humans have purpose and creativity (pioneering power), that is, the ability to detect problems and establish goals in any environment; Machines can only perform their talents within the framework of "problems, constraints, goals" given by humans. [10] Specific to the technical solution mentioned in the Patent Law, it refers to the collection of technical means adopted to solve the technical problem to be solved that use the laws of nature, which can obtain beneficial technical effects, and reflects the creative intellectual labor results made by the inventor based on the accumulated technical knowledge in the face of the technical problem to be solved. The technical solution to be protected by the patent application or the prior art solution in the examination of the patent application shall be a collection of specific technical means to solve a specific technical problem, that is, it should be the result of human creative intellectual labor, reflecting the ability of human beings to discover and solve problems.

The Claims.com and All Prior Art.com websites generate patent text in a completely different way from the aforementioned technical solution. The Claims.com and All Prior Art.com websites use artificial intelligence algorithms to generate a large number of claims and their permutations that are novel in form by rearranging phrases or replacing terms with alternative definitions, synonyms or antonyms. Cloem claims that its algorithm is more than random, and that it can be substituted with antonyms to describe a unique technical solution. Some researchers believe that the vast majority of the so-called technical solutions produced in these ways are nonsense. [11] In fact, the websites themselves are aware that "most of the technical solutions that result are ridiculous and meaningless"[12]. This paper argues that there is no real invention and creation activity in the process of using text editing technology to generate formal patent text by artificial intelligence machine with the help of big data mining and analysis means. It is only based on a large number of published patent texts, and a large number of formal texts can be obtained by random arrangement and combination through algorithms, machine learning, and decision-making mechanisms. The process of producing these formal patent texts and the content recorded do not reflect the mechanism of scientific and technological genesis, there is no invention idea and real invention and creation activity, but it is only the product of text editing of published patent texts. In the process of generating the content of these formal patent texts, artificial intelligence machines do not recognize the prior art and its defects, do not propose technical solutions to solve specific problems from the perspective of innovative knowledge generation, and cannot reflect the creative intellectual labor made based on the accumulated technical knowledge in the face of the technical problems to be solved. It is also not suitable for use as prior art in patent examination.

In summary, the patent text and its content based on the content generated by artificial intelligence need to be specifically analyzed according to its generation process whether it can be patented as a technical solution or as prior art in patent examination. The so-called technical documents generated by text editing technology should not be protected as technical solutions under the Patent Law, nor should the novelty and inventiveness of the patent application be evaluated as prior art in the examination of patent applications.

exegesis

[1] China Academy of Information and Communications Technology, JD Exploration Research Institute, Artificial Intelligence Generated Content (AIGC) White Paper (2022).

[2] Victoria Slind-For, Cloem, PepsiCo, Coca-Cola, In Dyne: Intellectual Property, Bloomberg Tech. (3 October 2014).

[3] See https://allpriorart.com/about/

[4] See Specifio, https://specif.io.

[5] LIU Qiang,ZHOU Yicheng. Research on patent examination standards for artificial intelligence invention[J].Jingyue Journal,2018,(3): 76-85.)

[6] A Marco, The Value of Certainty in Intellectual Property Rights: Stock Market Reactions to Patent Litigation",

(2005) Vassar College Department of Economics Working Paper No 82, at 1.

[7] Ben Hattenbach & Joshua Glucoft．Patents in an Era of Infinite Monkeys and Artificial Intelligence[J]．

Stanford Technology Law Review，2015，(1): 32-51．

[8] LIU Xin. Patent Regulation System for Artificial Intelligence Generation Technology Solutions: Theoretical Controversies, Practical Problems and Legal Countermeasures[J]. Legal Science(Journal of Northwest University of Political Science and Law),2019,(5):82-92.

[9] Erica Fraser, Computers as Inventors - Legal and Policy Implications of Artificial Intelligence on Patent Law,13 SCRIP Ted 322 (2016).

[10] Zhong Yixin. Principles of Machine Knowledge-Action—Unified Theory of Artificial Intelligence[M]. Beijing:Beijing University of Posts and Telecommunications Press,2014.

[11] Erica Fraser, Computers as Inventors - Legal and Policy Implications of Artificial Intelligence on Patent Law,13 SCRIP Ted 305 (2016)，at 308.

[12] All Prior Art, “About”，available at http://allpriorart.com/about/.

(This article only represents the views of the author, not the position of intellectual property)

Image source | Internet