From The Field

December 31, 2019

Earlier this year, the World Intellectual Property Organization (WIPO) initiated a new series of reports, WIPO Technology Trends.  The first report in the series, entitled Artificial Intelligence, looks at artificial intelligence (AI) patenting trends and issues across industries and around the world.

With 27 contributing authors, one might expect the more than 150 page report to be disjointed and to read like a collection of articles, related only by a common topic area.  To the contrary, the report is not just informative, but its parts stand together as a consistent whole.  The directors of the report made sure that it employed a common set of definitions and statistical assumptions throughout, and that each chapter was written with the contents of other chapters in mind.  It certainly merits study by anyone interested in AI and patents.

The report highlights that patenting activity around AI inventions has increased dramatically since 2013, after being relatively low for decades prior.  Notably, of all the AI patent families published since 1960, more than half were published after 2013. 

Using patent office technology classifications, the report defines several technologies as falling under the AI category.  The “machine learning” category is by far the most statistically significant, accounting for almost 90% of all AI patents that address a specific AI technique.  Within the machine learning category, neural networks and deep learning show the most patent growth in recent years.  According to the report, filings of new patent families on deep learning inventions increased 175% per year during the period 2013-2016, growing from 118 in 2013 to around 2,400 in 2016.  In the broader category of neural networks, applications increased 46% annually over the same period and over 6,500 patent families were initiated in 2016.

United States, Japanese, and Korean companies, led by IBM, Microsoft, Toshiba, and Samsung have obtained the most AI patent families.  However, Chinese entities dominate AI patent filings by university and public research organizations. 

The report’s statistics do not break down total filings by nationality of the applicant.  However, the report does identify which patent offices receive the initial patent filing in a family.  The report explains that, with a few exceptions, entities generally initiate filings in the country in which they are located.  Therefore, looking at which country’s patent offices receive the most initial filings for AI patent families is a reasonable proxy for which country’s entities are doing the most filings in total.  China and the U.S. lead the world in patent filings, followed by Japan and Korea.  The trends for each of these four countries differ noticeably.  In recent years, U.S. AI patent filings have increased significantly, but at a substantially linear rate.  By contrast, China’s AI patent filings have increased exponentially over the past 10 years and the annual number of new AI patent families filed in China was more than double the number in the U.S. during the most recent years for which the report cites available data (2016 for China data and 2015 for U.S. data).  

Japan has shown a steady rate of AI filings dating all the way back to 1985 and, until the late 1990s, regularly exceeded all other countries.  But Japan’s filings are now a fraction of those in the U.S. and China.  Korea had a negligible number of filings until the mid-1990s, but has had a consistent, if modest, increase in AI filings each year since then.

The full report is available for download at: https://www.wipo.int/tech_trends/en/artificial_intelligence/

Author: Michael Mauriel