Rising Patent Applications – And Challenges – For New Technologies, Artificial Intelligence 22/02/2018 by Catherine Saez, Intellectual Property Watch Leave a Comment Share this:Click to share on Twitter (Opens in new window)Click to share on LinkedIn (Opens in new window)Click to share on Facebook (Opens in new window)Click to email this to a friend (Opens in new window)Click to print (Opens in new window)The steady increase in innovations relating to new digital technologies, in particular technologies using artificial intelligence, is matched by an upward patenting trend. The European Patent Office recently issued a study on the subject and is preparing a conference in May, while the World Intellectual Property Organization is working on its own in-depth study. However, the current patent system might not be ready for artificial intelligence-related inventions, according to a global standards-setting body. The European Patent Office is organising a conference on 30 May on the patenting of artificial intelligence, in Munich. The conference will look into the relevance of patents for artificial intelligence, the challenges and opportunity of patenting this new area of technology. Registration for the event opens on 3 April. EPO Study: Patent Applications Rising According to a a study [pdf] “Patents and the Fourth Industrial Revolution – The inventions behind digital transformation,” published by the EPO in December, innovation in the Fourth Industrial Revolution is taking off. As explained by the study, the Fourth Industrial Revolution (4IR) describes the “full integration of information and communication technologies (ICT) in the context of manufacturing and application areas such as personal, home, vehicle, enterprise and infrastructure.” The 4IR is more than a “mere continuation or even acceleration of the development of ICT,” the study says, as its “truly disruptive nature” originates in the combined use of a wide range of new technologies. Over 5,000 patent applications for inventions relating to autonomous objects were filed at the EPO in 2016, the study said, adding that in the last three years, the rate of growth of 4IR patent applications was 54 percent. The largest number of such patent applications were filed for 3D systems, artificial intelligence, and user interfaces. At the core of the 4IR, according to the study, is the massive deployment of the Internet of Things. The trend is also fed by a number of other technologies, such as cloud computing and artificial intelligence, according to the study. By 2025, it is estimated that between 26 to 30 billion devices in the home and workplace will be equipped with sensors, processors and embedded software, and connected to the Internet of Things. Once these objects are combined with other technologies, such as cloud computing and artificial intelligence, “they enable the automation of entire business processes, including repetitive intellectual tasks previously performed by human beings.” The study classified the 4IR inventions into three main sectors: Core technologies (hardware, software and connectivity), enabling technologies (such as analytics, security, artificial intelligence, position determination, power supply), and application domains (such as home, personal, manufacturing, vehicles). The top patent applicants at the EPO for Fourth Industrial Revolution inventions between 2011 and 2016 were Samsung, LG, Sony, Nokia, Huawei, Qualcomm, Blackberry, Koninglijke, Philips, and Intel. Europe, the United States and Japan are “established leaders,” while China and South Korea, “are rapidly catching up,” the study found. According to the study, the top European, US and Japanese applicants include large companies with strong patent positions in application domains and enabling technology fields. The leading applicants in Europe are Germany and France. Germany is specialising in vehicles, infrastructure and manufacturing, while France is focused on enabling technologies, such as artificial intelligence, security, and 3D systems. Innovation mostly happens in the enhancement of products and processes in the “virtual layer of software, rather than in any hardware components,” and a “large proportion” of current inventions are based on software implementation. Artificial intelligence and 3D systems are the smallest and most recent fields in the enabling technologies sector, the study found, but have also been the fastest-growing 4IR fields, with average growth rates of 43 percent for artificial intelligence, and 56 percent for 3D systems between 2011 and 2016. There are however unresolved technical obstacles in the field of artificial intelligence, linked to the computing capacity that artificial intelligence calculations require, “that need to be overcome before AI can achieve its full potential.” Patent Eligibility for AI Inventions in US, Canada According to January 2018 blog by US law firm Rowand, in Canada a patent awarded to Amazon.com for their “one click” ordering technology after a challenge clarified the Canadian Patent Office’s policy on computer-implemented inventions. After that ruling, the patent office published new guidance which “broadly speaking” considers computer-related subject matter as patentable, the blog said. The “one click” Amazon US patent expired in September 2017. In the US, according to the blog, the US Patent and Trademark Office follows “the two-part test,” which was set out in a Supreme Court ruling Alice v CLS Bank. Following that test, an invention can only be patent eligible if it amounts to “significantly more” than subject matters that are not eligible, such as law of nature, natural phenomena, or abstract ideas. AI innovations can be divided into two categories, Rowand said: 1) new or improved artificial intelligence techniques, such as refinements to machine learning techniques; and 2) applications for known artificial intelligence techniques. The blog said in Canada the situation is “relatively favourable to both categories,” but in the US, the situation varies between the two categories. Favourable factors include consideration of a new or improved AI technique as an improvement in computer technology. It also includes that the innovation is about automating a task previously performed by humans using a new, distinct process. Last month, two Microsoft executives gave a presentation at the Max Plank Institute for Innovation and Competition on IP on considerations for industry 4.0 and artificial intelligence. Sonia Cooper, senior patent attorney in the IP Policy and Strategy Group at Microsoft, and Nicolas Schifano, senior director and assistant general counsel at Microsoft, presented a set of slides [pdf] on global AI patent activity, and specific challenges on patenting AI technology. According to their presentation, between 2010 and 2017, 549 deep-learning patents were granted in the US, followed by China with 168, Japan 68, South Korea 67, and Germany 44. The unique challenges faced by AI-related inventions are multiple, according to the slides. They include differences in legal systems, and the fact that AI inventions are algorithmic, where patenting pure algorithms is not permitted. In the US, according to the slides, the top AI patent holders for the past ten years are Microsoft, IBM, Alphabet, Samsung, Canon, Sony, Nuance Communications, Fujifilm, and Intellectual Ventures Management. For Global Standards Organisation, AI a Different Animal Separately, the Institute of Electrical and Electronics Engineers (IEEE), a global standard-setting organisation, considers that the quality of patents is of utmost importance, according to Clara Neppel, senior director of the IEEE European Business Operations. “If the technology incorporated into a standard is covered by patents, their validity and scope must be clear for all stakeholders,” she told Intellectual Property Watch. When it comes to AI, it is possible that the patent system meets some limitations, and some issues are specific to AI and differ from traditional technology, she said. For example, from the point of view of the European patent system, patents are granted for technological inventions, she said. In the context of computers, issues have arisen on whether computer programmes were patentable or not. AI brings new uncertainties in the sense that AI systems are not stable, she explained. “AI is a learning system, something that for instance in the context of autonomous systems relies on data collection, and it poses the fundamental question of knowing if monopoly rights can be granted to something which is evolving in time.” The traditional patent system is based on the disclosure of the invention before applying for a patent, but for AI “you can only disclose what happens at the application time but you cannot possibly cover everything happening after that,” she said. “Who guarantees that the technical effect claimed is actually achieved and maintained over the entire life span of a patent?” she asked. “It is not clear how to cover this kind of evolving behaviour in AI with intellectual property rights, and in particular patents.” For its part, the World Intellectual Property Organization got an early start at the issue with a 1991 symposium [pdf] on the intellectual property aspects of artificial intelligence, at Stanford University (US). According to WIPO sources, an in-depth study of international AI patenting trends is expected to be published before the end of the year. Image Credits: Flickr – Howard Lake Share this:Click to share on Twitter (Opens in new window)Click to share on LinkedIn (Opens in new window)Click to share on Facebook (Opens in new window)Click to email this to a friend (Opens in new window)Click to print (Opens in new window) Related Catherine Saez may be reached at csaez@ip-watch.ch."Rising Patent Applications – And Challenges – For New Technologies, Artificial Intelligence" by Intellectual Property Watch is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.