|
|
|
| Technology, Theory, Problems and China's Response to Artificial Intelligence Algorithm Patents |
| Yao Ye1,2 |
| (1.Center for Studies of Intellectual Property Rights, Zhongnan University of Economics and Law, Wuhan 430073, China；2.Max Planck Institute for Innovation and Competition,Munich 80539,Germany) |
|
|
|
|
Abstract Artificial intelligence algorithms are involved in all aspects of human life and are used to determine loan rates and manage human resources. They can track, tag, predict and manipulate human behaviors. Analyzed from a technical point of view, algorithms themselves imply that solving a problem should follow structured steps. Historically, algorithm is considered as a natural process. Later, it is regarded as an algorithm written by programmers. At present, it is taken as an algorithm derived from machine learning technique. Subsequently, algorithms combined with certain technical features is considered as patented inventions in some countries and regions such as the European Union and the USA and China.#br# Nevertheless, artificial intelligence algorithms differ from traditional algorithms in that data is fed into them, the formation process is complex, and other factors contradict the established rules of patent law relating to algorithm examination. Many scholars have focused on how the current law should be changed to address the problems brought by artificial intelligence algorithms, lacking the justification for the patentability of artificial intelligence algorithms. Therefore, the purpose of this paper is to explore the distinctive characteristics of artificial intelligence algorithms from traditional algorithms and explore how the examination rules of these algorithms and their theoretical basis may be reformualted and established.#br# This paper argues for the justification of patentability of artificial intelligence algorithms from the perspective of promoting innovation by cumulative innovation theory and disclosure innovation theory. Due to the relatively low fixed costs and short time to market associated with artificial intelligence algorithm products, and the rapid and incremental nature of innovation in the artificial intelligence industry, it makes sense to grant patents to artificial intelligence algorithms and encourage developers to disclose technical solutions. Furthermore, patenting algorithms will enhance their transparency. In addition, this paper uses a comparative approach as a research method. In this paper, the patent laws and guidelines of the USA, China and the European Union are compared horizontally. In the Patent Examination Guidelines of the European Union, the artificial intelligence algorithms are excluded from the exclusion list because they are related to mathematical methods, ang when technical problems are involved, they can be regarded as patents. The USA has also updated its Patent Examination Guidelines to include a two-step test to address patent law issues for artificial intelligence algorithms. The Patent Examination Guidelines of China make additional provisions for the examination of artificial intelligence algorithms by amending the provisions It emphasizes the overall consideration principle of the claims, specifying that claims containing technical and algorithmic features should not be excluded from patentability, and that all claims should be considered as a whole in the process of examination of the patent office.#br#After analysis, this paper finds out that China encounters many obstacles in examining artificial intelligence algorithms. first, the algorithmic black box. Secondly, it is unknown whether artificial intelligence algorithms can be replicated, or if they can be implemented successfully with positive social effects. Third, in terms of creativity, it is difficult to determine the technical field and to judge whether the technical solution is positively progressive. Fourth, with respect to novelty, artificial intelligence algorithms' arithmetic power and data storage capabilities make it possible to bypass the examiner's database or search methods and produce seemingly novel but outdated solutions.#br#Therefore, this paper argues that the examination guidelines of China should be appropriately modified to address these issues like the provisions in China's patent examination guidelines and the the disclosure rules. It is essential to disclose not only the basic structure of the artificial intelligence model, but also the training method and the training data. Third, when conducting the "novelty" review, the "implementation effect" of the technical solution should be examined, and the model and training data of the artificial intelligence algorithm should be disclosed within a reasonable and legal scope. Fourth, when examining "inventiveness," attention should be paid to the concept of its technical field.#br#Although it is reasonable and gradually accepted to protect artificial intelligence algorithms under patent law, artificial intelligence algorithms may also be acquired and controlled by investors because they consist of underlying code. Therefore, artificial intelligence algorithms may also be the object of copyright law and the object of trade secret protection. The protection of patents on artificial intelligence algorithms cannot extinguish the possibility of their protection in other legal systems.#br#
|
|
Received: 20 December 2021
|
|
|
|
|
|
| [1] ABBOTT R.The reasonable robot: artificial intelligence and the law[M]. Cambridge University Press, 2020:28-30.[2] USPTO releases report on artificial intelligence and intellectual property policy[EB/OL]. (2020-10-07)[2020-10-23]. https://www.uspto.gov/sites/default/files/documents/USPTO_AI-Report_2020-10-07.pdf.[3] 蒋朔.2020年《专利审查指南》第二部分第九章修改解读[EB/OL]. (2020-02-10)[2022-03-11]. http://ip.people.com.cn/n1/2020/0210/c179663-31579809.html.[4] KUR, ANNETTE, THOMAS DREIER, et al. European intellectual property law: text, cases and materials[M]. The United Kingdom: Edward Elgar Publishing, 2019:156-157.[5] United States Patent and Trademark Office, Commerce.2019 revised patent subject matter eligibility guidance[EB/OL]. (2020-10-07)[2020-10-23]. https://www.federalregister.gov/documents/2019/01/07/2018-28282/2019-revised-patent-subject-matter-eligibility-guidance.[6] HILTY R, HOFFMANN J, SCHEUERER S. Intellectual property justification for artificial intelligence[J].Journal of Economic Issues, 2020,15(6):21-35.[7] BURK DAN L, MARK A LEMLEY. Policy levers in patent law[J]. Virginia Law Review,2003,17(8): 95-114.[8] MAZZOLENI R, NELSON R R. Economic theories about the benefits and costs of patents[J]. Journal of Economic Issues, 1998, 32(4): 1038-1039.[9] NLO SHIELDMARK. AI inventions and sufficiency of disclosure-when enough is enough [EB/OL]. (2019-10-03)[2022-03-14]. https://www.iam-media.com/ai-inventions-and-sufficiency-disclosure-when-enough-enough.[10] KIM D, ALBER M, KWOK M W, et al. Ten assumptions about artificial intelligence that can misleadpatent law analysis[J]. Max Planck Institute for Innovation & Competition Research Paper, 2021 ,36(21):56-69.[11] 狄晓斐. 人工智能算法可专利性探析——从知识生产角度区分抽象概念与具体应用[J]. 知识产权, 2020，34(6):81-96.[12] 哈工大大数据产业有限公司.一种基于深度学习的缺失值填充方法及系统: CN201710358297.2[P]. 2018-04-13.[13] 叶茂.一种基于卷积神经网络的车辆检测方法: CN 201410299644.5[P]. 2014-09-10.[14] 东北大学.一种基于深度学习的无人驾驶物流车: CN 201710146233.6[P]. 2017-06-20.[15] ANDREW GRIFFIN. Facebook shuts down AI Robot after it creates its own language[EB/OL]. (2017-02-08)[2022-03-11]. https://www.huffpost.com/entry/facebook-shuts-down-ai-robot-after-it-creates-its-own-language_n_61087608e4b0999d2084f6bf.[16] SENI G, ELDER J F. Ensemble methods in data mining: improving accuracy through combining predictions[J]. Synthesis Lectures on Data Mining and Knowledge Discovery, 2010, 2(1): 1-126.[17] 姚叶. 人工智能算法的可专利性问题研究[J]. 创新科技, 2021, 21(9):70-78.[18] 陈全真, 徐棣枫. 论人工智能生成发明的可专利性及权利归属[J]. 科技进步与对策，2022，39（9）：114-121.[19] 洪凌啸. 误区与正道:法律人工智能算法问题的困境、成因与改进[J].四川师范大学学报(社会科学版), 2020, 47(1):5 8-70.[20] 张洋. 论人工智能发明可专利性的法律标准[J]. 法商研究, 2020,37(6): 181-192.[21] 吴汉东. 人工智能生成发明的专利法之问[J]. 当代法学, 2019,33(4): 24-38.[22] YANISKY-RAVID S, JIN R. Summoning a new artificial intelligence patent model: in the age of pandemic[J]. Social Science Research Network, 2020，32（8）：105-126.[23] DREW HIRSHFELD. Artificial intelligence tools at the USPTO[EB/OL]. (2021-03-08) [2022-03-11]. https://www.uspto.gov/blog/director/entry/artificial-intelligence-tools-at-the?utm_campaign=subscriptioncenter &utm_content=&utm_medium=email&utm_name=&utm_source=govdelivery&utm_term=.[24] 中国知识产权资讯网. 专利无效宣告程序的制度功能探究[EB/OL]. (2019-05-24) [2022-03-11]. http://www.iprchn.com/cipnews/news_content.aspx?newsId=116235. |
|
|
|
