周伯翰

中文摘要

因AI可用於眾多科技領域，如使用AI的心臟功能監測設備、協助醫師診斷癌症的AI設備、使用AI控制交通工具、使用AI為核心的理財機器、使用AI設計工業產品、結合AI與機器視覺用於檢測異常產品、將AI分類演算法的數位影像技術用於語言辨識與數位遊戲及專家執業的數位系統。醫界已廣泛運用卷積神經網路作為「深度學習」網路，並將其用於疾病的影像判讀與分類，如用於皮膚科、眼科、病理學、侵入性放射學等，故AI相關發明越來越受各界重視。 由於將AI用於特定領域以解決特定問題係出於人類構思，即使AI能自行蒐集資料、選用參數、深度學習、生成結果，其主要仍依人類設計的原則運作。雖然南非專利局認定AI系統DABUS係發明人，且澳洲聯邦法院裁定AI系統DABUS為專利發明人，惟專利制度發達的主要國家與地區皆未承認AI為專利發明人，更遑論授予AI專利權；台灣智慧財產及商業法院亦駁回以AI系統DABUS為專利發明人之主張。 隨「邊緣運算」技術出現，AI可降低資料取得成本並增強資料隱私性能，故可使AI更能普遍用於物聯網，而更融入日常生活，所以專利制度發達的主要國家與地區越來越重視AI專利相關法制。因 此，本文將研究美國、歐洲與台灣有關AI發明的「專利適格性」與「進步性」之法制，並歸納其法律原則，俾法院與智慧局在審理AI專利爭議案件時能有較清晰且有系統的參考依據。

The Study of the Patent Eligibility and Inventive Step of Inventions of Artificial Intelligence

Bo-Han Jou

abstract

Because AI can be applied to many fields of technology, for example, heart function monitoring equipment using AI, AI equipment to assist physicians in diagnosing cancer, using AI to control means of transportation, the financial management machine using AI as the core, using AI to design industrial products, combining AI and machine vision to detect abnormal products, or digital imaging technologies using AI classification algorithms applied to language recognition, digital games, and digital systems related to the practice of professionals. The medical field has widely used convolutional neural networks as "deep learning" networks, and has applied the AI systems with "deep learning" networks to the image interpretation and classification regarding diseases, for instance, used in dermatology, ophthalmology, pathology, interventional radiology, etc.; therefore, AI-related inventions have received increasing attention from all social circles. Because the application of AI to specific areas to solve specific problems is derived from human ideas, even if AI can collect data, select parameters, deeply learn, and generate results by itself, it still mainly operates according to the principles designed by human. Although the South African Patent Office identifies the AI system DABUS as the inventor, and the Australian Federal Court rules that the AI system DABUS is the patent inventor, but major countries and regions with developed patent systems have not recognized AI as the patent inventor, let alone granting AI patent rights; Taiwan’s Intellectual Property and Commercial Court also rejects the claim that the AI system DABUS is the patent inventor. With the emergence of "edge computing" technology, AI can reduce the cost of data acquisition and enhance the performance of data privacy, so it can make AI more universally applied to the Internet of Things, and more integrated into daily lives; therefore, the AI patent-related legal systems in major countries and regions with developed patent systems is becoming more and more important. Consequently, this article will study the legal systems of the "patent eligibility" and "inventive step" of AI inventions in the United States, Europe, and Taiwan, and summarize the legal principles thereof, so that courts and the Intellectual Property Office can have clearer and more systematic reference bases when hearing AI patent dispute cases.