Processing Artificial Intelligence: International Importance of IP and AI
From: Canadian Intellectual Property Office
International Importance of Intellectual Property in Artificial Intelligence
AI has started to pave the way for more innovative and futuristic approaches to daily tasks. From autonomous vehicles to smart toys, AI has become synonymous with being on the edge of leading innovation. To protect the new advancements, key players, such as universities, companies and public research organizations, have turned to IP rights to help protect their new technologies. The fast-paced environment of AI development has led to protecting inventions in two ways: (1) a patent, and (2) through scientific publication.Footnote 12
Canada, the U.K., Australia and Germany are leaders in scientific publications in specific applications of AI.Footnote 13 Players may be strategically publishing in scientific journals to put the work into the public domain, thereby preventing others from patenting the invention.Footnote 14 For countries participating in the European Patent Convention, computer programs are not considered to be patentable subject matter. On the other hand, U.S. has no such restrictions.Footnote 15 For Canada, the Canadian Patent Act states that a patent can only be granted for the “physical embodiment of an idea” and hence computer programs are not considered to be patentable subject matter.Footnote 16 When interpreting the analysis in the following sections, it is important to keep in mind that the culture towards patenting for AI is evolving and is currently applied differently between countries which could have some influence on the overall patent activity as measured by patented invention counts.
On this page
Patent filing trend
As seen in Figure 3, worldwide patent filings in AI between 2011 and 2017 have grown considerably, increasing on average by 31% annually. More research should be undertaken to understand the factors responsible for this increase. Other than the fact that there are more researchers actively patenting their AI inventions, it would be useful to understand if this growth is a result of researchers increasingly recognizing the value of patent protection as it relates to AI inventions.
Figure 3: International patent activity in AI between 1998 and 2017
Description of figure 3
Figure 3 consists of a line chart superimposed over a bar chart with each chart. The bar chart shows the number of patented inventions from 1998 to 2017 and uses the axis on the left. The line chart shows the growth rate in percentage over the same time period and uses the axis on the right.
|Publication year||Patented inventions||Annual growth rate|
In Figure 4, we take a closer look at the trend in patented inventions published over time based on the country of origin of the assignee, which includes institutions and researchers who are assigned the rights to an invention. China has made considerable headway in patenting AI inventions and is responsible for most of the growth worldwide over the past decade. Canadian assignees rank sixth overall in terms of absolute counts of patented inventions, ranking behind assignees originating from Germany and ahead of those from the U.K. The surge in filings from assignees originating from the United States (U.S.) is likely a result of the U.S. government policies to promote innovation in this field.Footnote 17
Figure 4: Trend in AI patent activity by assignee’s country of origin between 1998 and 2017
Description of figure 4
Figure 4 is a line chart shows the trend in patent activity by the assignee’s country of origin between 1998 and 2017. The following countries are shown: Australia, Austria, Brazil, Canada, China, Germany, Japan, Republic of Korea, United Kingdom and United States. The line depicting Canadian data is bolded. The lines for China and United States go out of the graph at 2012 and 2014, respectively, due to the high volume of patented inventions from those countries. Labels are included at the top of the graph showing the volume of patented inventions for China for 2012 and 2017, as well as the United States for 2014 and 2017.
|Publication year||China||United States||Japan||Republic of Korea||Germany||Canada||United Kingdom||Australia|
It should be noted that China’s representation in the international dataset pertaining to domestic filings by applicantsFootnote i is incomplete for the timeframe considered in this analysis.Footnote ii As a result, China will not be used to benchmark Canada’s performance in this report. Inventions originating from Chinese applicants that are patented abroad are assumed to be accurately captured by the respective filing offices that administer the international filings.
In order to gauge the predominance of assignees from certain countries in patenting AI inventions over time, we have developed a metric called the Intellectual Property Concentration Index (IPCI). This index can be used to determine the competitiveness of an industry or technology field based on the distribution of patented inventions held by all the countries active in that industry or field (additional details in Annex C). Index values closer to 0 indicate a more competitive global environment consisting of a large number of less-active countries, whereas index values closer to 1 would indicate a more concentrated global environment consisting of a few dominant countries.
In Figure 5, we observe the change in the IPCI value over time at the international level. After the initial dip in the index value in the early 2000s, we notice a gradual increase in the value over the years, synonymous with an increase in the level of concentration of AI patented inventions by country of origin. Interestingly, the index value is approaching 0.5 in 2017, which indicates a near duopoly: a situation where assignees from two countries file predominantly in AI. Based on the trend observed in Figure 4, we confidently establish that these two countries are China and the U.S., since they cumulatively account for 85% of the inventions patented globally in 2017.
Figure 5: Intellectual Property Concentration Index in AI between 1998 and 2017
Description of figure 5
Figure 5 consists of a line chart that shows the index value of the level of IP concentration between 1998 and 2017.
|Publication year||IP concentration index|
Figure 6 provides a representation of influential countries in AI. A fractional counting approach of assignees assigned to patented inventions is used to better represent the distribution of assignees across jurisdictions from which they originate. This approach avoids double counting assignees while accurately accounting for patented inventions involving multiple assignees, sometimes from different countries. For example, in the case where an invention is patented by an American researcher and two Canadian researchers, Canada would be assigned two-thirds of the patented invention count, whereas the U.S. would be assigned the remaining third. The counts are normalized by gross domestic product (GDP) in order to account for countries of different economic size. Even after accounting for this, China and the U.S. remain the leaders in this field, followed by Japan and Republic of Korea.
Figure 6: International patent activity by assignee's country of origin in AI between 1998 and 2017
Description of figure 6
Figure 6 is a choropleth map of the world showing the number of patented inventions per country. The graph is shaded with blue. The darker shades represent a higher number of patented inventions.
|Country of origin||Patented inventions / GDP|
|Republic of Korea||2.21869E-09|
|United States of America||9.50664E-10|
AI patent classification
Taking a closer look at the international data, in Figure 7 we notice that 88% of patented inventions categorized to AI Techniques are related to Machine Learning (ML). Due to the significant difference in the magnitude of volumes of patented inventions between ML and the other AI Techniques sub-categories, a different legend, represented by different shades of orange, is used to depict the change in the volumes for this sub-category. ML inventions grew annually by 63% between 2011 and 2017. Even though China is responsible for most of the ML patented inventions, the same trend can be observed for the eight leading AI patenting countries presented in Figure 4. Of note, 37% of the patented inventions are not being categorized to the predefined sub-categories of AI Techniques. This undefined sub-grouping of the traditional AI Techniques categories experienced a growth rate of 23% between 2011 and 2017 in patented inventions. It will be interesting to monitor the growth in this category to see if new AI techniques emerge as this technology area evolves.
Figure 7: Growth in international AI patent activity by AI Techniques
Description of figure 7
Figure 7 is a heatmap showing the growth in worldwide international AI patent activity by AI Techniques (sorted from most activity to least: Machine Learning, Logic Programming, Fuzzy Logic, Probabilistic Reasoning, Ontology Engineering, and Search Methods) from 1998 to 2017. All AI Techniques were represented with shades of blue except for Machine Learning, which is shaded in red and uses a different scale from the rest due to the exceptionally high activity in this field.
Turning to the applications of AI, it can be observed that AI was increasingly applied over the twenty-year period in patented inventions covering Computer Vision. It wasn’t until 2006 that Natural Language Processing (NLP) started to emerge as a key area of AI innovation. It is also important to indicate that 66% of patented inventions are not being categorized to the designated sub-categories of AI Applications. This could imply that more work is needed to identify additional groupings for AI Applications or, alternatively, new applications of AI are being patented that are not yet defined. This is a challenge that presents itself when studying a technology area that is evolving very rapidly and that is difficult to accurately define.
Figure 8: Growth in international AI patent activity by AI Applications
Description of figure 8
Figure 8 is a heatmap showing the growth in worldwide international AI patent activity by AI Applications (sorted from most activity to least: Computer Vision, Natural Language Processing, Control Methods, Distributed Artificial Intelligence, Planning and Scheduling, Knowledge Representation and Reasoning, Speech Processing, Predictive Analytics, and Robotics) from 1998 to 2017. All AI Techniques were represented with shades of blue except for Computer Vision and Natural Language Processing, which were shaded in red and use a different scale from the rest due to the exceptionally high activity in these fields.
|Distributed Artificial Intelligence||6||13||5||4||8||11||10||10||9||8||19||38||65||82||136||232||325||426||437||723|
|Knowledge Representation and Reasoning||14||11||10||15||35||33||62||79||78||94||98||120||126||84||102||127||132||129||186||312|
|Natural Language Processing||23||25||35||54||40||81||65||74||88||92||102||149||137||130||168||265||321||425||610||12,00|
|Planning and Scheduling||19||19||18||23||33||28||31||23||30||25||65||76||99||91||110||181||231||313||367||680|
With respect to AI Fields over the 1998-2017 period, Life and Medical Sciences remains one of the main fields in terms of generating AI patented inventions. Physical Sciences and Engineering is another field where AI is being significantly patented. With the growing interest in autonomous vehicles, it is interesting to note the rise in AI patented inventions categorized to the Transportation field, especially over the past six years. In an economy where the transportation industry is responsible for a significant number of jobs, it will be important to keep pace in the creation and use of AI technologies for transportation. As is the case with the two aforementioned groupings, we notice that 34% of the patented inventions are not being categorized to the pre-defined sub-categories of AI Fields.
Figure 9: Growth in international AI patent activity by AI Fields
Description of figure 9
Figure 9 is a heatmap showing the growth in worldwide international AI patent activity by AI Fields (sorted from most activity to least: Life and Medical Sciences, Transportation, Physical Sciences and Engineering, Telecommunications, Networks, Computing in Government, Industry and Manufacturing, Personal Computers and PC Applications, Businesses, Banking and Finance, Security, Energy Management, Education, Document Management and Text Processing, Arts and Humanities, Entertainment, Agriculture, Cartography, Military, Publishing, and Law, Social and Behavioral Sciences) from 1998 to 2017. All AI Techniques were represented with shades of blue except for Life and Medical Sciences, Transportation, and Physical Sciences and Engineering, which were shaded in red and use a different scale from the rest due to the exceptionally high activity in these fields.
|Arts and humanities||5||16||11||11||24||23||19||28||27||47||40||49||44||43||37||44||59||59||71||136|
|Banking and finance||33||30||42||42||48||59||41||54||53||59||83||123||132||132||170||167||215||277||327||601|
|Computing in government||63||69||66||69||117||110||98||114||112||145||169||195||229||229||249||373||472||570||729||1,238|
|Document management and text processing||5||5||6||22||22||28||24||44||34||35||29||61||51||35||36||64||55||68||95||202|
|Industry and manufacturing||56||35||38||53||72||84||77||49||57||56||85||133||119||84||156||220||249||352||481||1,086|
|Law, social and behavioral sciences||0||0||2||1||2||0||1||0||1||1||3||5||8||3||7||4||4||13||13||11|
|Life and medical sciences||120||157||170||205||284||330||298||343||304||403||417||451||602||545||698||776||945||1,182||1,338||2,159|
|Personal computers and pc applications||51||46||66||74||118||77||78||112||112||127||159||163||184||162||180||223||263||302||342||628|
|Physical sciences and engineering||193||180||189||187||262||277||305||268||202||224||263||309||319||300||425||443||538||665||917||1,467|
Considering 12.4% of patented inventions do not fall into any of the designated AI categories, we decided that it was appropriate to look into this area more closely. Consistent with the growing trend in AI inventions being patented, the trend in inventions not being categorized is also growing, as seen in Figure 10. Deep diving into the data to identify the IPC and CPC codes, we notice that many of the patented inventions not being classified belong to the G06F, G06K, G06N, G06Q subclasses that are related to data processing and related systems, recognition of data, and computer systems based on specific computational models. Additional research into these patented inventions could be conducted to identify new emerging AI technologies that are not being captured according to the definition used in this report.
Figure 10: Growth in international AI patent activity for patented inventions not categorized to a pre-defined AI category
Description of figure 10
Figure 10 is a heatmap showing the growth in international AI patent activity for patented inventions not categorized to a pre-defined AI category from 1998 to 2017. The heatmap is shaded in blue.
- Date modified: