SF Technotes

WIPO: U.S. and China Lead the World in AI Innovation

By Michael Castelluccio
February 6, 2019

A new study from the United Nations World Intellectual Property Organization (WIPO) documents “a massive recent surge in artificial intelligence-based inventions, with U.S.-based companies IBM and Microsoft leading the pack as AI has moved from the theoretical realm toward the global marketplace.”


The first publication in the WIPO Technology Trends series was released Thursday, January 31, 2019, in Geneva, Switzerland. The study considered more than 340,000 AI-related patent applications and 1.6 million scientific papers published since AI arrived on the scene in the 1950s. The agency notes in its summary that what had been a gathering storm of applications before 2013 has since become a blizzard.



More than 20 leading global AI experts contributed to the study, and Director General Francis Gurry explains in the preface, “This new report aims to shed light on trends in innovation in AI. I hope that we will help to shift debate away from speculative interpretation and toward evidence-based projections, thereby informing global policymaking on the future of AI, its governance, and the IP framework that supports it.” Filed patents not only provide a fact-based method for mapping the current state of AI, but by identifying current trends in AI, can also help outline where the revolution is heading.


The report begins its analysis of AI patent data with applications made in the 1950s. It starts there because the first AI conference that initiated research into AI was held at Dartmouth College in 1956. At the time, computers were learning strategies for playing checkers. The WIPO report offers an overview of 70 years of AI development since then, with this startling figure emerging from the numbers: “Fifty percent of all AI patents have been published in just the last five years.” Today, after machine defeats of the best human chess and Go players, thousands of researchers worldwide are collectively branching out, exploring autonomous cars, robotic surgeons, and apps that can recognize your relatives in your photos to help you sort your digital albums.


Sorting the data to find the top 30 AI patent applicants identified two general tracks: companies and academic enterprises—universities and public research organizations. For now, companies hold 26 out of the top 30 positions for applicants, and the remaining four spots are assigned to schools and research organizations.


The top two companies are IBM, with the largest portfolio of AI patent applications (8,290 at the end of 2016), and Microsoft, with 5,930. The next three companies are Japan-based Toshiba, Corp. (5,223); Samsung Group of the Republic of Korea (5,102); and NEC Group of Japan (4,406). The rest of the field looks like this:


Click to enlarge.

Image courtesy of World Intellectual Property Organization


The second track is where China excels. “Out of the top 20 universities and public research organizations in the AI field, the vast majority (17) are in China, and the remaining three are in the Republic of Korea.” The two largest portfolios in this category are the Chinese Academy of Sciences (CAS), number 17 in the top 30 list, and Korean Electronics and Telecomm Research Institute (ETRI).


Click to enlarge.

Image courtesy World Intellectual Property Organization




Looking at the trends in AI techniques, machine learning far outpaces all others with an overwhelming “89% of filings mentioning this AI technique and 40% of all AI related patents.” Machine learning grew by 28% from 2013 to 2016. For comparison, other important AI techniques showed a 16% growth for fuzzy logic and a 19% growth rate for logic programming. (NOTE: Boolean logic supplies computers with true or false answers, 0s or 1s, while fuzzy logic deals with degrees of truth. For example: Is it dark out yet? Boolean=Yes or No. Fuzzy=All the shades of twilight from sunlit to dark. Logic programming produces programs built on the formal rules of logic. See the link to a four-page glossary of AI terms from WIPO at the end of this article.)


Within the general category of machine learning, deep learning is the fastest-growing technique in AI with a 175% increase between 2013 and 2016. Deep learning describes the machine searching for answers or successful paths, without step-by-step programming instructions. Deep learning, also known as deep neural learning, is a division of machine learning in which neural networks can learn from unstructured and even unlabeled data, unsupervised by human operators.




Where the progress becomes most obvious are in the AI functional applications where the public can see and often purchase the amalgamations of software, hardware, and machine intelligence.


Leading the growth in AI applications are those enabling computer vision, which includes image recognition. “Computer vision,” according to the report, “was mentioned in 49% of all AI-related patents and grew by 24% during 2013-2016.” Close behind are natural language processing (14% of all AI-related patents) and speech processing, mentioned in 13% of the patents.


Although these very human functions lead the functional applications list, growth in many other areas are highlighted in the study. These include increases in robotics and control methods (+55%), and applications for planning/scheduling (+37%). The top AI application fields by industry include: transportation (15% of all patents); telecommunications (12%), and life and medical sciences (12%). Those with the most active growth of at least 30% in the 2013-2016 period are transportation, agriculture, and computing in government.



As the phrase “artificial intelligence” gets a little ragged at the edges by a combination of exaggerated marketing and general misunderstanding, the promised WIPO updates will become even more important.


It would be a serious mistake to assume artificial intelligence is just another bullet item to be added to a product listing. Many have already categorized AI as an invention not unlike electricity in its wide-ranging impact. But in the long term, it might be even more, coupled with the advances in computer hardware that have brought us to the near horizon of quantum computing.


Machine learning, and deep learning are more than just two recent components of the Fourth Industrial Revolution. In fact, we really haven’t seen anything of their kind unless you reach back to two other epic inventions—writing and reading. We sometimes forget that we continue to evolve as a species, and now, so do some of our machines. Especially noteworthy are those that can learn on their own.

Read the full WIPO Report, WIPO Technology Trends, here: www.wipo.int/tech_trends/en/artificial_intelligence


Download a glossary of Selected AI Categories and terms here:



Michael Castelluccio has been the Technology Editor for Strategic Finance for 24 years. His SF TECHNOTES blog is in its 21st year. You can contact Mike at mcastelluccio@imanet.org.

0 No Comments