Language selection

Search

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.

Figure 3: International patent activity in AI between 1998 and 2017
Publication year Patented inventions Annual growth rate
1998 1,242 -
1999 1,167 -6.04%
2000 1,245 6.68%
2001 1,304 4.74%
2002 1,794 37.58%
2003 1,799 0.28%
2004 1,818 1.06%
2005 1,874 3.08%
2006 2,141 14.25%
2007 2,395 11.86%
2008 2,829 18.12%
2009 3,447 21.85%
2010 3,888 12.79%
2011 3,673 -5.53%
2012 4,568 24.37%
2013 5,584 22.24%
2014 6,889 23.37%
2015 8,602 24.87%
2016 10,669 24.03%
2017 18,216 70.74%

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.

Figure 4: Trend in AI patent activity by assignee’s country of origin between 1998 and 2017
Publication year China United States Japan Republic of Korea Germany Canada United Kingdom Australia
1998 5 337 435 18 97 14 12 2
1999 5 272 443 26 98 10 5 0
2000 11 335 395 53 87 6 8 3
2001 15 406 341 42 66 13 9 2
2002 19 459 363 37 114 32 28 3
2003 54 457 333 42 85 16 15 3
2004 72 434 334 51 91 26 9 4
2005 86 455 420 65 111 18 9 12
2006 116 560 480 72 92 21 24 20
2007 192 716 447 139 103 34 17 17
2008 423 772 496 170 126 43 18 11
2009 697 976 510 165 142 58 19 14
2010 960 1,074 493 127 154 46 21 27
2011 1,059 961 407 178 118 50 22 23
2012 1,932 967 387 197 129 50 35 16
2013 2,631 1,429 484 276 195 61 55 13
2014 3,513 1,832 443 402 163 69 33 19
2015 4,655 2,064 539 385 141 69 41 15
2016 6,180 2,386 622 471 173 81 46 27
2017 11,955 3477 794 690 227 135 71 23

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.

IP concentration

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.

Figure 5: Intellectual Property Concentration Index in AI between 1998 and 2017
Publication year IP concentration index
1998 0.19
1999 0.20
2000 0.17
2001 0.16
2002 0.11
2003 0.10
2004 0.09
2005 0.11
2006 0.12
2007 0.13
2008 0.13
2009 0.14
2010 0.15
2011 0.16
2012 0.22
2013 0.27
2014 0.32
2015 0.35
2016 0.38
2017 0.46

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.

Figure 6: Worldwide international patent activity by assignee's country of origin in AI between 1998 and 2017
Country of origin Patented inventions / GDP
Albania 3.3203E-10
Argentina 9.64366E-12
Australia 1.6548E-10
Austria 9.43528E-11
Belgium 1.39159E-10
Brazil 3.58552E-11
Bulgaria 6.14128E-11
Canada 5.16785E-10
Chile 1.00593E-11
China 2.51019E-09
Colombia 5.45078E-11
Croatia 8.22292E-11
Denmark 2.49959E-10
Egypt 1.59429E-11
Estonia 1.65099E-10
Finland 6.09576E-10
France 2.59943E-10
Gabon 4.11344E-10
Germany 5.72214E-10
Greece 3.66919E-11
Hungary 8.99147E-11
Iceland 1.1591E-10
India 1.3278E-10
Indonesia 2.8786E-12
Iran 8.81032E-12
Ireland 1.64711E-10
Israel 1.20912E-09
Italy 4.29143E-11
Japan 1.84413E-09
Kazakhstan 5.86377E-12
Republic of Korea 2.21869E-09
Kuwait 7.05827E-12
Latvia 2.86952E-11
Lebanon 1.76556E-11
Lithuania 3.75577E-11
Luxembourg 1.29519E-10
Malaysia 9.31287E-11
Mexico 2.61479E-11
Mongolia 1.53733E-10
Morocco 1.77222E-10
Netherlands 5.6914E-10
New Zealand 1.41446E-10
Nigeria 5.03436E-12
Norway 5.29039E-11
Pakistan 3.19928E-12
Panama 6.14863E-11
Philippines 6.04393E-12
Poland 4.77993E-11
Portugal 2.10103E-11
Puerto Rico 9.88817E-12
Romania 8.3489E-12
Russia 6.27431E-11
Saudi Arabia 9.45707E-11
Slovenia 7.37525E-11
South Africa 5.70206E-11
Spain 8.20368E-11
Sri Lanka 1.12485E-11
Sweden 3.57511E-10
Switzerland 5.72642E-10
Taiwan 5.95531E-10
Thailand 3.96045E-12
Tunisia 1.25437E-10
Turkey 2.73969E-11
Ukraine 7.64337E-12
United Kingdom 1.4583E-10
United States of America 9.50664E-10
Uzbekistan 1.9802E-11
Venezuela 8.29257E-12
Vietnam 4.08249E-12

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.

Figure 7: Growth in international AI patent activity by AI Techniques
AI technique 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Fuzzy logic 130 105 90 109 94 70 90 66 65 66 93 94 112 111 109 112 138 148 137 192
Logic programming 113 117 122 124 144 162 157 126 133 155 175 225 227 192 187 238 272 364 476 785
Machine learning 582 519 555 585 777 778 790 825 969 1,060 1,344 1,774 2,138 2,088 2,615 3,178 4,026 5,301 6,832 12,859
Ontology engineering 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 1 0 0
Probabilistic reasoning 1 0 0 0 1 2 1 3 0 2 3 1 7 7 5 4 3 4 4 12
Search methods 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0

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.

Figure 8: Growth in international AI patent activity by AI Applications
AI application 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Computer Vision 134 142 145 138 215 223 217 259 299 325 392 444 534 501 687 864 1,081 1,551 2,056 3,830
Control Methods 104 113 97 86 80 87 83 75 59 67 83 95 103 65 71 94 118 189 380 568
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
Predictive Analytics 1 7 6 8 15 15 25 22 18 19 21 36 37 36 61 69 87 88 121 197
Robotics 3 1 3 5 9 6 19 8 9 9 12 18 14 7 15 16 23 41 18 58
Speech Processing 49 38 45 63 73 49 47 52 36 44 56 46 48 48 62 56 84 107 139 258

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.

Figure 9: Growth in international AI patent activity by AI Fields
AI field 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Agriculture 7 6 9 7 13 11 12 13 14 8 16 15 31 19 28 59 58 74 92 210
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
Business 11 6 24 27 45 39 30 74 43 60 81 155 189 190 163 198 233 284 317 612
Cartography 6 0 10 6 10 13 6 11 12 16 16 24 32 25 50 57 46 51 80 147
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
Education 21 20 61 37 43 52 32 53 49 43 31 58 60 50 158 83 97 108 172 311
Energy management 24 33 13 13 37 38 37 30 30 23 38 52 54 88 127 180 229 252 362 533
Entertainment 14 11 14 27 37 30 30 31 18 37 34 49 42 25 39 40 46 62 65 112
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
Military 3 3 8 2 2 6 15 10 8 16 17 14 19 16 18 25 31 41 39 48
Networks 84 97 87 76 83 68 74 93 117 111 131 143 169 181 274 305 377 554 761 1,752
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
Publishing 0 2 3 2 4 2 3 7 7 4 3 6 3 5 11 7 12 13 13 33
Security 18 18 26 44 56 51 56 72 85 75 89 117 123 94 133 167 158 240 297 466
Telecommunications 104 81 106 123 184 187 156 144 169 178 222 226 278 272 424 398 437 640 753 1,272
Transportation 86 82 112 103 135 148 177 195 192 207 277 299 303 285 420 573 728 1,134 1,684 3,363

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.

Figure 10: Growth in international AI patent activity for patented inventions not categorized to a pre-defined AI category
Other 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Other 201 188 192 207 249 229 283 272 395 432 501 573 569 535 630 736 898 936 1,079 1,420
Date modified: