In recent years, the evolution of computer software and hardware has led to a new trend of Artificial Intelligence (hereinafter “AI”) causing a rage in many fields, and the number of AI patent applications is significantly increasing. Hereunder is our overview of how to evaluate whether AI patent applications are in accordance with the definition of invention of Patent Act in Taiwan.
An invention patent must have a technical character, that is, utilizing the laws of nature (Article 21 of Taiwan’s Patent Act) to create something, in order to meet the definition of invention under Patent Act. Notably, Taiwan’s Patent Examination Guidelines clearly define that mathematical methods per se do not utilize the laws of nature. Further, when it comes to an invention regarding computer program, said Guidelines state that the invention of computer program must overcome technical difficulties and achieve technical effect(s) rather than only replace labor work with computer, for said computer program invention to be deemed to possess technical character.
However, taking Artificial Neural Network in AI field for instance, its essence is to emulate biological neural networks, especially animal brains, by mathematical methods to allow computer working similar to human brains to learn, draw inference, and make judgment or decision. In other words, Artificial Neural Network is essentially a mathematical model, and the main function of its many applications is to replace human mental processes.
As such, in local practice, the difficulty for Artificial Neural Network to meet the requirement of invention definition is that Artificial Neural Network itself is a mathematical method that does not utilize laws of nature, and implementing Artificial Neural Network is an ordinary ability of computer to replace human work. Therefore, neither Artificial Neural Network itself nor computers that implement Artificial Neural Network can confer technical characters to the invention.
Under such circumstances, the key points to determine whether the Artificial Neural Network invention patent application has technical characters include, among others, what field the invention would be applied in, what the additional technical function the invention could be in combination with, and whether the invention has technical effect. Hereunder is the flowchart that shows the process to determine whether an invention patent application comprises technical characters.
This flowchart was recently provided by the Taiwan Intellectual Property Office (hereinafter “TIPO”) in a couple of occasions. 1 For the current examination practice, the examiner would follow the foresaid flowchart to determine whether an invention has technical character when the patent claims may include both “abstract ideas” and “technical component.” For further explanation, said “abstract ideas” mean matters which do not utilize laws of nature, such as mathematical algorithms and formulas, business method, computer program, and certain methods of organizing human activity, while said “technical component” indicates the remaining elements after excluding abstract ideas in patent claims.
Step 1 of the aforesaid flowchart is to determine whether the invention involves any technical information generated from some technical component, as well as any interaction of collaboration among technical components. It is worth noting that, the technical information indicates, such as, physical, chemical, biological, or electrical information that is rendered by sensing elements, which possesses technical characters in nature. If the information is merely the history record or the business information input and processed by human, it tends to be regarded as non-technical information as it does not comprise technical character.
Subsequently, Step 2 provides a macroscopic viewpoint to examine whether the invention solves a technical problem and achieves technical effect. For example, if an invention merely gives prediction by means of Artificial Neural Network, it could not serve any technical purpose. However, if the invention provides more description regarding the way on how other technical components solve technical problems by using the prediction result from Artificial Neural Network, such invention is more likely to be deemed as a technical application serving specific technical purpose.
Lastly, Step 3 adopts a microscopic viewpoint to examine whether the invention discloses substantive technical means. For example, an invention would be deemed as simply implementing the computer if the invention does not present any special technical nature about the computer hardware and software, but merely implements Artificial Neural Network on an ordinary computer. On the other hand, it is more possible to define the invention as having specific technical application if such invention comprises specific technical means in the technical field.
From the statistics released by TIPO2, out of 517 applications, there are 21 that received rejection office actions (OAs) for their failure to meet invention’s definition. Although the percentage of the applications receiving this kind of rejection OA is not high, an application may still be unable to overcome the rejection if the essence of the application is not to solve technical problem (for example, the application is only to enhance the efficiency of business marketing.) With the increasing AI patent applications, TIPO puts emphasis more than before on whether an application complies with the definition of invention during the examination procedures. Therefore, it is recommended that, for filing AI patents and other patents that are related to computer software, the applicants could use the pathway described in this article to evaluate the patent disclosure before submitting the patent applications.
We may further take into account the practices in examination of AI patents applied by other Patent Offices. The European Patent Office (hereinafter “EPO”) released Guidelines for Examination on “AI and machine learning” invention in October 2018. Said Guidelines stipulate that it is not sufficient to provide technical character if the invention “applying AI itself,” unless the invention would apply AI to achieve technical contribution for technical purpose. Meanwhile, the United States Patent and Trademark Office (hereinafter “USPTO”) released its “2019 Revised Patent Subject Matter Eligibility Guidance” in January 2019, adopting a new examination process to determine whether the invention is integrated into a practical application. China Intellectual Property Office (hereinafter “CNIPA”) amended new examination rules regarding AI patent application in its China Guidelines for Patent Examination in December 2019, which manifest that the invention would not be deemed as a technical solution and a patentable subject matter, unless the invention solves the technical problem, applies technical means, or achieves technical effect.
In conclusion, similar to EPO, USPTO and CNIPA, TIPO’s patent examination practices show that the determination on technical character depends on whether the invention provides technical contribution. Thus, along with the development of technology, merely implementing AI is not sufficient to fulfill the “technical character” requirement. As to whether the judgment of technical character among TIPO and other worldwide Patent Offices will become more consistent, this remains to be seen in the future.
||Ching-Chi LI, The Development And The Response In The Field Of AI Under Taiwan Patent Examination Practice, Seminars On International AI Technical Development, AI-related Patent Portfolio And Its Trend & The Development And The Response In The Field Of AI Under Taiwan, EU and US Patent Examination Practice, p.17 (2019)
||TIPO, Overview On AI-related Patent Application And The Analysis On The Common Reasons For Rejection, October 2019 version, p.34 (2019)