November 2023
Luke Hung and Albert Yen
Since the introduction of the Transformer model architecture by Google in 2017, the adoption of the Self-Attention Mechanism has significantly enhanced the processing efficiency of sequential input data. This advancement has led to major breakthroughs of deep learning, particularly in NLP (natural language processing) tasks, ushering in the era of LLM (large language models). In the field of artificial intelligence (AI), the most notable landmark event recently is the release of OpenAI’s ChatGPT in November 2022, built upon GPT-3.5, the previously trained and developed large language model. Due to its remarkable interactivity and text generation capabilities, ChatGPT quickly became the fastest-growing consumer application in history, sparking a global AI sensation. This has prompted discussions not only in the technological realm but also across various societal, economic, legal, and educational aspects, with the patent law system naturally being one of the affected areas.
This article’s main discussion can be broadly categorized into two aspects: “AI related inventions” and “inventions or creations involving AI’s participation.” The former primarily focuses on the examination of patent requirements, a practical challenge that has been faced for many years but has recently required adjustments due to the evolution of AI. The latter pertains to the identification of inventors or creators, representing a relatively new challenge for the existing patent law system.
1. AI Related Inventions Aspect
The term “AI related inventions” encompasses any invention whose purpose is achieved through AI, meaning that AI is used as a means to achieve purposes and generate functions. This includes not only inventions whose claims are based on AI related algorithms. As a subordinate concept to “computer software related inventions, “AI related inventions” similarly encounter the challenges that traditional computer software related inventions face and the issues may become even more complex due to the technical characteristics of AI.
For example, the issue of “patent eligibility” in computer software related inventions is a contentious area in practical examination. According to Article 21 of the Patent Law, “invention” means the creation of technical ideas, utilizing the laws of nature. The examination guidelines further clarify that an invention must possess “technical character,” involving technical means in solving problems. Cases identified as purely mathematical methods or methods requiring human mental activities for execution are excluded from patent eligibility. However, as AI’s essence is rooted in mathematical methods, merely “utilizing AI” does not sufficiently confer technicality. When “AI related inventions” combine “mathematical methods” with “hardware resources,” determining the overall technicality and establishing specific examination steps for maintaining consistency in results become challenging issues.
Moreover, Article 26, paragraph 1 of the Patent Law states that the description shall fully disclose the invention in a manner clear and sufficient for it to be understood and carried out by a person ordinarily skilled in the art. However, AI models, especially large-scale neural network models trained using deep learning, are often perceived as “black boxes” by humans. It is challenging to understand or explain the complex parameter relationships as well as the computational logic behind the model’s output. Determining how much disclosure is “sufficient” for “AI related inventions” in the specification, balancing the goals of “enablement requirement” and practical feasibility, presents a dilemma.
As early as 1998, the Intellectual Property Office of the Ministry of Economic Affairs (the “IP Office”) formulated the initial version of the “Examination Guidelines for Computer Software Related Inventions.” With the booming of information and communication technology, these guidelines underwent four revisions, and the latest version (the “2021 Guidelines”) became effective on July 1, 2021. To cope with the related technology evolution and the challenges arising from “AI related inventions,” in the “patent eligibility” aspect, the 2021 Guidelines removed content related to “further technical functions” and “simple computer utilization,” Instead, clearer judgment steps and process were established[1], including an initial assessment of whether it conforms to the “clearly conforms/does not conform to the invention definition.” If not, a further assessment of whether it complies with the requirement of “whether information processing through computer software is concretely implemented using hardware resources.” In the “enablement requirement” aspect, the 2021 Guidelines also incorporate explanations using AI related cases[2]. Furthermore, in January 2022, the IP Office released the “Compilation of Information Technology Patent Examination Cases,” based on the 2021 Guidelines. It includes cases related to information technology, incorporating AI, such as training methods for deep learning systems, neural network systems, neural network chips, providing reference for the judgment of patent requirements.
2. Inventions or Creations Involving AI’s Participation Aspect
“Inventions or creations involving AI’s participation” refer to situations where AI plays a certain role in the invention or creation process. The most direct question arising from “inventions or creations involving AI’s participation” is: How to identify the inventors or creators? A somewhat intuitive answer is to consider the extent of AI’s participation in the invention or creation process. If AI is used as a tool by humans, assisting in the invention or creation process (referred to as “AI assisted creation”), the human using AI should be considered the inventor or creator. If AI independently generates inventions or creations unforeseen by humans (referred to as “AI’s autonomous creation”), AI itself should be considered the inventor or creator. However, this approach may face the following two challenges.
Firstly, while “AI assisted creation” and “AI’s autonomous creation” can be conceptually distinguished, in practice, there is a considerable gray area, making it challenging to draw clear boundaries. As aforementioned, current large-scale neural network models are basically like a “black box” to humans. When humans provide the model with certain instructions and conditions, with some degree of anticipation for the model’s output, but understanding and explaining its principles are very difficult, it becomes challenging to determine whether this scenario falls under “AI assisted creation” or “AI’s autonomous creation.”
Secondly, according to the current judicial practice, inventors or creators must be “natural persons,” but AI is not a natural person (or even a legal person), thus cannot be considered an inventor or creator. This is consistent with international mainstream views, as seen in the notable “DABUS case.” Dr. Stephen Thaler, in the United States, filed patent applications in multiple countries, asserting that an AI system named DABUS should be recognized as the inventor. However, this argument did not receive support from patent offices or courts in most jurisdictions. In Taiwan, the IP Office rendered a “decision of case not entertained” for the DABUS case, and the decision was upheld in subsequent administrative litigation by the Intellectual Property and Commercial Court[3]. Consequently, even in cases of “AI’s autonomous creation,” under the current patent law system, AI still cannot be recognized as an inventor or creator.
As noted above, given the current state, achieving “AI’s autonomous creation” without controversy is still a distant goal, and most jurisdictions do not allow AI to be recognized as an inventor or creator. Therefore, issues related to “inventions or creations involving AI’s participation” have not yet caused practical difficulties in the law system, except for a few cases such as DABUS. However, as AI technology continues to evolve rapidly, the question arises: When the number of model parameters approaches the order of magnitude of human brain synapses, could a “quantum leap” occur, unleashing AI’s creativity? If AI develops the ability to autonomously contribute to inventions or creations in the future, will the law systems that do not recognize AI as inventors or creators inhibit incentives to file patent applications for such inventions or creations (e.g., users of AI might prefer to protect inventions or creations as trade secrets)? Would it be necessary to create a new intellectual property right type for “inventions or creations involving AI’s participation”? All of these issues await further discussion.
This article briefly reviews the evolution of AI and its impact on the patent law system. On the aspect of “AI related Inventions,” the technical characteristics of AI make the challenges faced by traditional computer software related inventions more complex during examination. To address this, the new version of the “Examination Guidelines for Computer Software Related Inventions” provides clearer examination guidance and case explanations. The IP Office also published the “Compilation of Information Technology Patent Examination Cases” for external reference. On the aspect of “inventions or creations involving AI’s participation,” while there are currently only a few cases in practice, the potential development of higher creativity in AI raises questions about how the patent law system should be adjusted in response, becoming a focal point of academic discussion.
[1] Pages 2-12-13 to 2-12-21 of the 2021 Version of Examination Guidelines for Computer Software Related Inventions.
[2] Pages 2-12-35 to 2-12-38 of the 2021 Version of Examination Guidelines for Computer Software Related Inventions.
[3] Intellectual Property and Commercial Court, 101 Xin Zhuan Su No. 3 Administrative Decision.
The contents of all materials (Content) available on the website belong to and remain with Lee, Tsai & Partners. All rights are reserved by Lee, Tsai & Partners, and the Content may not be reproduced, downloaded, disseminated, published, or transferred in any form or by any means, except with the prior permission of Lee, Tsai & Partners.
The Content is for informational purposes only and is not offered as legal or professional advice on any particular issue or case. The Content may not reflect the most current legal and regulatory developments. Lee, Tsai & Partners and the editors do not guarantee the accuracy of the Content and expressly disclaim any and all liability to any person in respect of the consequences of anything done or permitted to be done or omitted to be done wholly or partly in reliance upon the whole or any part of the Content. The contributing authors’ opinions do not represent the position of Lee, Tsai & Partners. If the reader has any suggestions or questions, please do not hesitate to contact Lee, Tsai & Partners.