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.

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
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
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 applicants^{Footnote 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
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: 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.

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.

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.

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.

Endnotes

Footnote 12

World Intellectual Property Organization. [2019]. WIPO Technology Trends 2019 - Artificial Intelligence. [p. 21]. Retrieved from: https://www.wipo.int/edocs/pubdocs/en/wipo_pub_1055.pdf

Return to footnote 12 referrer

Footnote 13

Ibid, 12

Return to footnote 13 referrer

Footnote 14

World Intellectual Property Organization. [2019]. WIPO Technology Trends 2019 - Artificial Intelligence. [p. 23]. Retrieved from: https://www.wipo.int/edocs/pubdocs/en/wipo_pub_1055.pdf

Return to footnote 14 referrer

Footnote 15

World Intellectual Property Organization. [2010]. Experts' Study on Exclusions from Patentable Subject Matter and Exceptions and Limitations to the Rights. Retrieved from: https://www.wipo.int/edocs/mdocs/scp/en/scp_15/scp_15_3-annex2.pdf

Return to footnote 15 referrer

Footnote 16

Innovation, Science and Economic Development Canada. A guide to patents. Retrieved from: https://www.ic.gc.ca/eic/site/cipointernet-internetopic.nsf/eng/h_wr03652.html

Return to footnote 16 referrer

Footnote 17

United Kingdom Intellectual Property Office. [2019]. Artificial Intelligence: A worldwide overview of AI patents and patenting by the UK AI sector. [p. 14]. Retrieved from: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/817610/Artificial_Intelligence_-_A_worldwide_overview_of_AI_patents.pdf

Return to footnote 17 referrer

Language selection

WxT Search form

Processing Artificial Intelligence: International Importance of IP and AI

International Importance of Intellectual Property in Artificial Intelligence

On this page

Patent filing trend

IP concentration

AI patent classification

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

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

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