Processing Artificial Intelligence: Analysis from a Canadian Perspective

From: Canadian Intellectual Property Office

Analysis from a Canadian Perspective

This section provides a snapshot of Canadian involvement in AI based on researchers who patented inventions between 1998 and 2017. Canadian participation in this field is likely much larger based on those who use non-formal forms of IP, such as trade secrets, and involvement in scientific publications, among other activities. Based on the information extracted from the patent data, this report provides a partial indication of the overall activity level in AI using the detailed information that is captured from this source.

As organizations turn their focus towards the ever-evolving world of AI, researchers, academics and experts are coming into higher demand. This has led to a greater push by academic institutions to fund AI-related disciplines, such as computer science, electrical and electronic engineering, mathematics and statistics, and neuroscience to name a few.^{Footnote 18} In turn, retaining those trained in a country while simultaneously attracting those trained in other countries has become an initiative many governments have taken on. In their annual Global AI Talent Report,^{Footnote 19} Element AI has found that, "...the AI talent pool is highly mobile, with about one-third of researchers working for an employer based in a country that was different from the country they received their PhD." Of this talent pool, Canada is one of the leaders in high-impact research. Element AI defines high-impact research as a country where a "...higher-than-average percentage of the local talent pool is making high-impact contributions to the field" of AI. This attributes to Canada’s label as a platform country: a country which sees both an inflow and an outflow of talent. Our research in this section validates the aforementioned statement.

Patent filing trend

Patent filing activity by Canadian researchers in AI has experienced waves of increased activity over the past 20 years, with a significant increase over the past five years. The 31% average annual growth over this short period is similar to the growth rate experienced at the global level. As seen in Figure 11, inventions patented by Canadian researchers increased gradually between 1998 and 2001, before experiencing a significant surge in 2002. Canada experienced a slightly higher average annual growth rate of 24% between 2005 and 2010 compared to the global 20% growth rate observed during this timeframe.

Figure 11: Patent activity by Canadian researchers in AI between 1998 and 2017
Publication year	CIPO	USPTO	WIPO (PCT)	Other	Annual growth rate
1998	4	4	4	1
1999	8	2	2	0	-7.69%
2000	9	6	5	0	66.67%
2001	4	20	5	0	45.00%
2002	17	31	10	0	100.00%
2003	14	29	5	1	-15.52%
2004	12	17	10	1	-18.37%
2005	13	15	5	1	-15.00%
2006	12	28	9	4	55.88%
2007	10	45	12	1	28.30%
2008	16	40	20	3	16.18%
2009	31	52	14	0	22.78%
2010	14	64	18	2	1.03%
2011	22	45	11	0	-20.41%
2012	16	40	18	1	-3.85%
2013	11	40	16	4	-5.33%
2014	21	43	24	5	30.99%
2015	16	66	23	6	19.35%
2016	20	65	36	13	18.92%
2017	25	138	34	14	58.33%

Relative specialization of Canadian researchers

In order to gain a better understanding of a country’s performance in terms of AI patent activity, we use the Relative Specialization Index (RSI) (additional details in Annex D). This measure uses patenting intensity to allow for technology sectors to be compared between countries of different sizes. The index provides a measure of each country’s share of patented inventions within the AI field as a share of the country’s total patented inventions produced within a given timeframe. For countries where the value is greater than zero, they are seen as being relatively specialized compared to the rest of the world. Conversely, countries with an RSI value of less than zero are deemed to not have a specialization. In Figure 12, we incorporate a time dimension to present the change in degree of specialization by splitting the dataset in two 10-year periods. It is interesting to see that the index scores for Canada, along with Germany, Japan, and the U.S., have decreased over the index scores for the first decade.

Figure 12: Relative Specialization Index by researcher’s country of origin in AI
Publication year	1998-2007	2008-2017
Republic of Korea	-0.56	-0.44
Germany	-0.06	-0.27
Japan	-0.07	-0.18
United Kingdom	-0.14	-0.03
Australia	0.12	0.22
Canada	0.34	0.30
United states	0.40	0.35

The index can be further broken down at an AI Applications grouping level to determine areas in which Canadian researchers are specialized. As observed in Figure 13, Canadian researchers are highly specialized in Knowledge Representation and Reasoning, and NLP. It is interesting to compare RSI values between Canada and the U.S., as Canada appears to hold its own against a country that is a world leader in terms of patent activity in AI more broadly. The question then arises: how can Canadian researchers’ talent be leveraged to advance innovation in this field so as to allow this specialization to increase further?

Figure 13: Relative Specialization Indices by AI Applications for American and Canadian researchers
AI application	RSI Canada	RSI United States
Distributed artificial intelligence	-0.55	-0.65
Planning and scheduling	-0.17	-0.28
Control methods	-0.02	0.12
Speech processing	0.09	-0.06
Predictive analytics	0.08	0.32
Robotics	0.42	0.34
Knowledge representation and reasoning	0.48	0.29
Computer vision	0.43	1.72
Natural language processing	1.32	2.41

From the perspective of absolute patented invention counts, it is reassuring to observe that Canadian researchers are prolific filers in areas in which they are deemed specialized. This is not always the case, as seen with Robotics, where Canadian researchers do not hold many patented inventions but are considered specialized relative to researchers from other countries who file a proportionately lower number of patented inventions.

The leading Canadian researchers patenting in AI are presented in Figure 15 below. The institution associated with a particular researcher and the researcher’s primary area of expertise are shown in blue and red respectively in the figure below. Interestingly, 93% of Canadian institutions include at least one Canadian researcher on patented inventions included in this dataset. However, only 51% of Canadian researchers in the AI field are associated with Canadian institutions or a Canadian subsidiary of a multi-national company. Focusing on the top ten researchers, only two are associated with international institutions. Canada’s top patenting researchers have a significant presence in Canada, working for established companies, such as Primal AI and D-Wave Systems.

Figure 15: Top Canadian researchers along with their associated institutions (in blue) and their main area of expertise (in red)
Researcher name	Patented inventions	Name of institution	AI technique
Boyer, John M.	5	IBM, US	Natural language processing
Hardjasa, Amelia	5	Pulse Energy, CA	Predictive analytics
Ka-Hing, Lin	5	Rockwell Automation Technologies Inc, CA	Control methods
Macready, William G.	6	D-Wave Systems Inc, CA	Knowledge Representation and Reasoning
Francis, Ilyas Ihab	9	Primal AI, CA	Knowledge Representation and Reasoning
Bertrand, Benoit Patrick	10	Accenture LLP, CA	Knowledge Representation and Reasoning
Hatami-Hanza, Hamid	10	Science Counter Inc, CA	Knowledge Representation and Reasoning Natural Language Processing
Edson, Tirelli	10	Red Hat, US	Other
Joseph, Sweeney Peter	15	Primal AI, CA	Knowledge Representation and Reasoning
Voon, Gerard	27	Ecco Projects Inc, CA	Computer vision

Beyond using patented invention data to identify trends and top filers, it is useful to further explore the types of technologies being created. A patent landscape map is presented in Figure 16, which was generated using an algorithm that relies on the keywords from patent documentation to cluster patented inventions according to shared language. The patented inventions are organized based on common themes and grouped as "contours" on the map to identify areas of high and low patent activity. The white peaks represent the highest concentrations of patented inventions, and each peak is labelled with key terms that tie the common themes together. The distance between keywords helps to illustrate the relationship between peaks, where shorter distances indicate that the patented inventions they represent share more commonalities relative to those that are further apart. Words located close together may be part of similar systems or technologies, whereas keywords that are further apart likely have less of a relationship.

Superimposing the names of the leading Canadian researchers is a useful way to highlight who is working in what space. For example, we see that Peter Joseph Sweeney and Ihab Francis Ilyas—both associated with Primal AI and patenting in the Knowledge Representation and Reasoning application field—are both captured in the middle-right side of the map and are associated with patented invention characterised by keywords such as "Atomic knowledge representation model", "Complex knowledge representation", "Concept", "Structure" and "Program". Perhaps as expected, we see that Amelia Hardjasa, who is the only leading researcher patenting in the Predictive Analytics application field, has patented inventions on the map that are located at the bottom centre on an island separated from the mainland on the map. This type of landscape map can also be useful to institutions in order to target talent. For example, in the case where a researcher’s patented invention portfolio overlaps with that of another researcher, it can indicate that the two are patenting in similar areas, as in the case of Gerard Voon and Patrick Benoit Bernard in the upper right quadrant of the map.

Figure 17 shows the distribution of the patent activity by Canadian researchers by province across the country. Each province is shaded in orange, with darker shades representing a higher number of patented inventions. The number of researchers residing in prominent Census Metropolitan Areas (CMA) with prolific patenting populations is highlighted on the map using red location pointers. It should be noted that patented invention volumes were calculated using the fractional counting approach and have been normalized by population size in this figure. The provinces that have a higher number of patented inventions, such as Ontario, British Columbia and Quebec, remain the leading provinces after being adjusted. The provinces with more patent activity tend to be associated with clusters of researchers around large cities.

Number of patented inventions per population
Province	Patented inventions / population
Northwest Territories	0
Nunavut	0
Alberta	1.48735E-05
Yukon	0
Saskatchewan	8.16868E-06
Newfoundland and Labrador	3.80515E-06
Ontario	4.14448E-05
British Columbia	3.86343E-05
Manitoba	1.04921E-05
Quebec	1.60857E-05
New Brunswick	9.08005E-06
Nova Scotia	5.94059E-06
Prince Edward Island	0

Geographical clusters of inventive activity by Canadian researchers
Census metropolitan area (CMA)	Number of researchers
Abbotsford-Mission	1
Barrie	7
Belleville	1
Calgary	44
Edmonton	41
Greater Sudbury	1
Guelph	16
Halifax	8
Hamilton	18
Kelowna	1
Kingston	12
Kitchener-Cambridge-Waterloo	162
Lethbridge	3
London	21
Moncton	1
Montréal	171
Oshawa	11
Ottawa-Gatineau	273
Peterborough	2
Québec	29
Regina	8
Saint John	3
Saskatoon	9
Sherbrooke	4
St. Catharines-Niagara	9
St. John's	2
Thunder Bay	1
Toronto	496
Trois-Rivières	4
Vancouver	238
Victoria	19
Windsor	8
Winnipeg	22

Gender analysis: Female participation in AI patent activity

In 2014, 59% of graduates aged 25 to 34 in Canadian science and technology programs were female. However, only 23% of engineering graduates of that same age group were female.^{Footnote 20} Understanding that female representation in science, technology, engineering, and mathematics (STEM) fields is lower than desired, this observation is expected to also be reflected in female participation in patenting. To measure female participation in AI, we leverage WIPO’s comprehensive name dictionary to assign genders to the names of researchers listed on patented inventions.^{Footnote 21}

At the international scale, there was one female identified for every three males involved in AI patenting. By comparison, for patented inventions containing at least one Canadian researcher, that ratio decreases to one female for every six male researchers.

To better understand female participation in this technology field, Figure 19 effectively shows the evolution of females involved in AI over the twenty-year period. The following figure shows new researchers by gender entering the field based on their involvement as captured by the first patent application published. Unlike the trend observed for women in Canada, which is relatively steady over the twenty-year period, the international trend is much different, with female representation as a share of the total number of researchers decreasing over time.

International
Publication year	Male	Female	Not assigned	Total
1998	1,067	1,037	210	2,314
1999	943	1,024	203	2,170
2000	1,087	900	210	2,197
2001	1,226	1,102	198	2,526
2002	1,936	1,265	310	3,511
2003	2,052	1,300	295	3,647
2004	2,106	1,347	304	3,757
2005	2,312	1,342	419	4,073
2006	2,767	1,492	466	4,725
2007	3,333	1,353	491	5,177
2008	4,173	1,636	682	6,491
2009	5,224	1,967	731	7,922
2010	5,716	2,032	889	8,637
2011	5,312	1,954	890	8,156
2012	6,474	2,106	1,072	9,652
2013	8,488	2,544	1,403	12,435
2014	10,197	2,833	1,724	14,754
2015	12,281	3,219	2,028	17,528
2016	14,422	3,639	2,593	20,654
2017	27,031	80,52	3,883	38,966

Canada
Publication year	Male	Female	Not assigned	Total
1998	14	4	0	18
1999	19	2	1	22
2000	16	2	1	19
2001	91	13	6	110
2002	128	11	7	146
2003	116	15	5	136
2004	102	19	5	126
2005	73	10	4	87
2006	75	9	5	89
2007	105	19	6	130
2008	118	18	8	144
2009	109	21	5	135
2010	128	14	13	155
2011	117	21	10	148
2012	101	23	7	131
2013	82	17	5	104
2014	138	25	19	182
2015	170	32	18	220
2016	199	36	23	258
2017	269	62	29	360

Non-traditional sectors, such as mining, forestry, electricity and the skilled trades among others, tend to experience a drop in women’s participation five to ten years after graduation.^{Footnote 22} This lack of retention could possibly be due to the systemic barriers faced by women who already work in the field, a theory more closely examined in a report CIPO had published titled "Women’s Participation in Patenting: An Analysis of PCT Applications Originating in Canada".^{Footnote iii} Increasing this retention rate is one of the values set forth by the Department for Women and Gender Equality. A result of this would be an expected increase in women’s representation in patented inventions across STEM fields, which would include AI. In 2012, Germany’s Institute of Labour Economics published an article that looked into the causes for the lack of participation by women in patenting, and the results coincide with the observations made for Canada.^{Footnote 23}

To continue growing in AI, Canada must work to retain our best talent and also attract new talent into the country. In Element AI’s 2019 Global AI Talent Report, Canada cements itself as a platform country.^{Footnote 24} Interestingly, the U.S., Germany and China are anchored countries: able to retain more talent but unable to attract talent into the country. Australia, however, finds itself as an inviting country; meaning it is able to draw in a large portion of international talent without losing much of its domestic talent. Canada is among the top countries in the world consisting of high-impact researchers, joined by the U.S., China, the U.K. and Australia.

Canadian institutions

After gaining a better understanding of the AI patent landscape from the perspective of Canadian researchers, this section examines AI from the perspective of Canadian institutions. A Canadian institution for the purpose of this report includes a corporate firm, an academic institution, or a government establishment. Overall, corporate firms were responsible for 82% of the patented inventions in this section, whereas academic institutions and government departments accounted for the remaining 15% and 3% of the patent activity, respectively. To better understand innovation in the field by Canadian institutions and where their specialization lies, this section examines the associated patent activity in Canada and abroad. To limit the overall data cleaning effort in order to capture patented inventions by institutions, CIPO only focused its attention to cleaning the institution data for select countries that were used to benchmark against Canada. These countries include Australia, Germany, the U.K., Japan, Republic of Korea, and the U.S., and were selected based on the fact that they are among the leaders in terms of rankings by the WIPO Technology Trends report and their inclusion in the UKIPO’s AI report.

Innovation Superclusters Initiative

The Innovation Superclusters Initiative, introduced in 2017, is an investment of $950 million, matched dollar for dollar by the private sector, to support "superclusters" across Canada, which bring together "small, medium-sized and large companies, academic institutions and not-for-profit organizations" to "transform regional innovation ecosystems".^{Footnote 25} While the Scale. AI supercluster is largely focussed on AI, each of the superclusters plan to support projects that will promote the use of AI in their respective industries. The superclusters include the Digital Technology Supercluster, the Protein Industries Supercluster, the Next Generation Manufacturing Supercluster, SCALE.AI Supercluster, and the Ocean Supercluster.^{Footnote 26} The funding is allocated to the superclusters to help support advances in each sector and to help build and support emerging start-ups.^{Footnote 27}

The second investment is a joint venture between Canada and the U.K. known as the Canada-UK AI Initiative. This is a funding opportunity that requires interdisciplinary research between three major domains: social sciences and humanities, health and biomedical sciences, and natural sciences and engineering. The initiative looks to support the responsible development of AI while creating partnerships between researchers in Canada and the U.K.^{Footnote 28}

Finally, in 2017, the Government of Canada appointed the Canadian Institute for Advanced Research (CIFAR) to develop and lead the Pan-Canadian Artificial Intelligence Strategy.^{Footnote 29} This strategy is a $125 million initiative in partnership with the Alberta Machine Intelligence Institute (Amii), the Montréal Institute for Learning Algorithms (Mila) and the Vector Institute.^{Footnote 30} This funding initiative looks to do several things:

Increase the number of AI researchers and skilled graduates in Canada
Connect three of the major centres for AI in Canada
Invite conversation surrounding economic, ethical, policy and legal implications of advances in AI
Support the national research community

These initiatives work to further AI development in Canada ethically and fairly. Many of these initiatives focus on improving the lives of Canadians by funding and supporting researchers and developers as they explore the world of AI.

Relative specialization of Canadian institutions

Prior to exploring the AI Canadian institution data more closely, it is useful to begin by establishing where Canadian institutions rank in terms of relative specialization. In comparison to the six other countries presented in Figure 20, Canada is seen to be specialized because it holds a positive index score. When breaking down the institution dataset over the two decades covered, it is interesting to observe that the specialization for all seven countries under consideration has decreased during the second decade. Later in this section, we will explore the AI fields in which Canadian institutions are predominantly patenting and their respective RSIs.

Figure 20: Relative Specialization Index by institution’s country of origin in AI
Publication year	1998-2007	2008-2017
Republic of Korea	-0.21	-0.26
Germany	0.00	-0.24
United Kingdom	-0.08	-0.13
Japan	0.62	0.11
Canada	0.26	0.22
Australia	0.31	0.22
United states	0.38	0.34

Patent filing trend

Figure 21 shows the general trend in AI patent filing activity by Canadian institutions. The annual growth rate of 8% between 1998 and 2010 for Canadian institutions is similar to the growth rate experienced at the global level during this timeframe. Although the 21% growth rate observed in patent activity by Canadian institutions over the 2011 and 2017 period is significantly higher, this rate is lower than the 31% growth rate observed at the international scale. Even though Canada has assumed a leadership role in defining a policy framework for AI,^{Footnote 31} their overall influence is limited considering Canadian institutions account for less than 1% of the total number of inventions patented by institutions globally.

Figure 21: Patent Activity by Canadian institutions in AI between 1998 and 2017
PUBLICATION YEAR	CIPO	USPTO	WIPO (PCT)	OTHERS	ANNUAL GROWTH RATE
1998	3	2	6	1
1999	3	1	1	1	-50%
2000	1	1	0	0	-67%
2001	5	4	3	0	500%
2002	15	2	7	0	100%
2003	10	2	3	0	-38%
2004	8	4	7	0	27%
2005	7	0	2	0	-53%
2006	7	1	4	4	78%
2007	11	12	7	1	94%
2008	13	3	15	2	6%
2009	12	10	13	0	6%
2010	6	12	10	2	-14%
2011	12	14	9	1	20%
2012	11	13	13	2	8%
2013	7	14	7	1	-26%
2014	13	12	12	3	38%
2015	13	18	17	2	25%
2016	19	19	30	1	38%
2017	20	53	33	7	61%

Understanding in which countries Canadian institutions are seeking protection for their inventions internationally provides an indication of which markets they are strategically targeting. Not surprisingly, Canadian institutions, apart from filing at CIPO, file predominantly in the U.S. because of its large market size, and have priority filings in that country for each year over the 20year timespan. The USPTO and CIPO administered 64% of all inventions patented by Canadian institutions. Inventions patented via the Patent Cooperation Treaty (PCT) system account for 32% of the filings. Other IPOs targeted by Canadian institutions, but to a significantly lower degree, include the IP Australia (IPA), China National Intellectual Property Administration (CNIPA), EPO, Japan Patent Office (JPO) and the French National Institute of Industrial Property (INPI).

Distribution of patented inventions

Since the early 1990s, IP has emerged as an important asset class to the corporate sector, be it in terms of protecting the value of its inventions resulting from significant investments in R&D or, alternatively, if there is interest in acquiring a targeted firm, or even simply if institutions are looking to exit the market and reap the benefits of their efforts.^{Footnote 32} Large acquisitions, resulting in significant transfers of IP, can create a monopoly-like condition by realizing economies of scale and driving out smaller players from the market. In this section, the IPCI is used to understand the distribution of patented inventions held by Canadian institutions actively patenting in AI and benchmark it against institutions from other leading countries. An index value closer to 0 in this section would indicate a country having a higher number of less-active institutions patenting in AI, whereas an index value closer to 1 would indicate a country has a few dominant players which patent extensively in AI. In reference to Figure 22, Canada has one of the lowest index values, thereby indicating that Canadian institutions operate in a highly competitive patenting AI environment.

Figure 22: IP Concentration Index for institutions from leading countries in AI
Country of origin	IP concentration index	Number of institutions
Germany	0.1877	479 institutions
United Kingdom	0.0577	166 institutions
Republic of Korea	0.0539	613 institutions
Japan	0.0430	541 institutions
United States	0.0279	4,136 institutions
Canada	0.0152	284 institutions
Australia	0.0111	144 institutions
International	0.0039	14,277 institutions

Figure 23 shows the distribution of inventions patented by Canadian institutions and their overall representation in the Canadian AI institution dataset. The size of each pie depicts the proportion of Canadian institutions responsible for the inventions in each of the five groups, whereas the angle prescribed by each pie in the figure represents the proportion of patented inventions in each group. As expected, a vast majority of the Canadian institutions hold between one and four patented inventions. On the other hand, institutions holding five or more patented inventions account for 43% of the Canadian institution AI dataset. Nevertheless, simply holding more patented inventions does not reflect the importance of such inventions.

Figure 23: Distribution of Canadian institutions by patented inventions and their overall representation
Patented invention group	Number of Canadian institutions	Total number of patented inventions
1	199	199
2-4	62	156
5-9	13	70
10-19	6	86
20+	4	117
Total	284	628

Figure 24 shows the breakdown of the data by application fields in order to gauge the industrial applications of AI inventions patented by international and Canadian institutions, respectively. As observed in Figure 24, institutions globally have been consistently filing for patented inventions pertaining to Life and Medical Sciences, and Physical Sciences and Engineering between 1998 and 2017, whereas patented inventions pertaining to Transportation gained prominence post 2011. On the other hand, Canadian institutions seem to specialize in the field of Telecommunications, apart from Life and Medical Sciences, and Physical Sciences and Engineering. However, there is a need for Canadian institutions to be more specialized in the field of Transportation. This fact is corroborated by Figure 25, in which Canadian institutions seem to have a low specialization index in this field.

In reference to Figure 25, it is important to note that there are certain AI Fields, such as Banking and Finance, Military, Law, social and behavioural sciences, etc., that have not been included in the figure because they have not attracted considerable amounts of patent activity globally. As a result, the specialization indices for such AI fields may be biased.

Figure 25: Relative Specialization Index by AI Fields for American and Canadian institutions
AI field	Canada	United States
Transportation	-0.36	-0.14
Networks	-0.13	-0.12
Energy management	-0.11	-0.51
Document management and text processing	-0.08	0.11
Industry and manufacturing	-0.04	-0.18
Computing in government	-0.03	-0.13
Banking and finance	-0.01	-2.11
Education	0.02	0.13
Business	0.04	0.02
Personal computers and PC applications	0.17	0.10
Telecommunications	0.21	-0.01
Security	0.22	0.06
Life and medical sciences	0.24	0.02
Physical sciences and engineering	0.25	0.03

Canada’s leading institutions patenting in the AI field over the 1998 to 2017 period are presented in Figure 26. Many of the institutions listed are prominent patenting entities that patent in a variety of technology fields while others, such as D-Wave and Primal AI, specialize in this field. It is also interesting to note that Nortel Networks, a now defunct business entity, is among the leading filers. The fact that owners of patented inventions are not required to update the information held in the patent database reflects a challenge when using patented invention data to identify owners of such rights.

Figure 26: Top Canadian institutions and their associated patented invention counts
Institution name	Patented inventions
University of British Columbia	11
1QB Information Technologies Inc.	12
Alcatel-Lucent Canada Inc.	13
IBM Canada Ltd.	13
Maluuba Inc.	14
Primal AI	15
Nortel Networks Corporation	19
D-Wave Systems Inc.	25
Schlumberger Canada Ltd.	25
BlackBerry Ltd.	40

Presenting the patent landscape map for Canadian institutions is another way to visualize patented invention data. In Figure 27, we superimpose the names of the top ten leading Canadian institutions to highlight who is working in which space. Unlike the landscape map for Canadian researchers, there is not much overlap in areas where multiple institutions are patenting, the exception being D-Wave Technologies and 1QB, which are both operating in the quantum computing technology field. One thing that is striking is that Primal AI’s patented inventions are captured in a peak on a peninsula secluded from the rest in the bottom-centre of the map. This peak is characterized by keywords such as "Knowledge Representation", "Concept", and "Data Structure". This placement on the map may suggest the company is operating in a niche area, far different than technologies related to those captured in the larger AI technology space.

Geographical clusters

In Figure 28, Canada’s CMAs associated with more than 10 patented inventions by AI institutions are presented in a geographic map. It should be noted that patented invention volumes were calculated using the fractional counting approach and have been normalized by GDP in this figure. In 2018, 650 startups were created "across all cluster cities". Many of these start-ups found support in local investors who helped give credibility to the ecosystem, leading to interest from international investors. The attention received by foreign investors has raised AI related deals by 41% and acquisition rates by 50% each year on average.^{Footnote 33} There are many benefits for institutions to cluster together, including increased productivity, faster innovation through collaborative research, and the creation of small institutions to cater to the niche needs of the industry.

Number of patented inventions
Province	Patented inventions / GDP
Northwest Territories	0
Nunavut	0
Alberta	1.68208E-10
Yukon	0
Saskatchewan	1.48738E-10
Newfoundland and Labrador	1.20333E-10
Ontario	3.71479E-10
British Columbia	3.63912E-10
Manitoba	5.50297E-11
Quebec	2.03698E-10
New Brunswick	1.62311E-10
Nova Scotia	1.35275E-10
Prince Edward Island	1.4298E-10

Geographical clusters of inventive activity by Canadian institutions
Census metropolitan area (CMA)	Number of institutions
Calgary	19
Edmonton	4
Guelph	2
Halifax	5
Hamilton	4
Kingston	3
Kitchener-Cambridge-Waterloo	20
Lethbridge	1
London	2
Moncton	1
Montréal	42
Oshawa	2
Ottawa-Gatineau	30
Québec	4
Regina	1
Saskatoon	4
Sherbrooke'	1
St. Catharines-Niagara	1
St. John's	1
Toronto	78
Trois-Rivières	2
Vancouver	30
Victoria	4
Winnipeg	3

There are six institution clusters, each comprised of ten or more institutions, that emerge as key areas leading innovation in the Canadian AI sector. Patenting from these clusters accounts for 83% of AI patented inventions nationwide. Similar to the geographic map in the Canadian Researchers section, the provinces are shaded in orange and normalized by their GDP. Most of the institutions that have inventions patented in AI are located in Ontario, which holds three main CMA clusters: Toronto, Ottawa–Gatineau, and Kitchener–Cambridge–Waterloo. The largest is the CMA of Toronto, which boasts 78 patenting AI institutions. Other CMAs with large clusters include Montréal (42 institutions), Vancouver (30 institutions) and Calgary (19 institutions).

Figure 29 takes a closer look at the relative specializations for each of the CMAs by AI Field. It is interesting to note that most of the prominent CMAs are specialized in a different AI. The Toronto CMA specializes in Networks, and Life and Medical Sciences, whereas the Kitchener–Cambridge–Waterloo CMA specializes in Computing in Government, and Personal Computers and PC Applications. Toronto has become an integral part of the national ecosystem, partly owing to the Vector Institute, a "not-for-profit corporation" that works with start-ups, the marketplace, incubators and accelerators to help drive AI research.^{Footnote 34} Toronto has attracted a lot of attention from foreign investors because of its reputation as Canada’s financial capital.^{Footnote 35} The city has also teamed up with Waterloo to gain the title of Silicon Valley of the North.^{Footnote 36}

Telecommunications AI field
CMA	RSI
Montreal	0.4575
Kitchener-Cambridge-Waterloo	0.1139
Ottawa-Gatineau	-0.0305
Toronto	-0.1806
Calgary	-0.3035
Vancouver	-0.3848

Physical sciences and engineering AI field
CMA	RSI
Calgary	0.6229
Vancouver	0.1583
Montreal	-0.0165
Toronto	-0.1998
Ottawa-Gatineau	-0.6056
Kitchener-Cambridge-Waterloo	-0.7325

Personal computers and PC applications AI field
CMA	RSI
Kitchener-Cambridge-Waterloo	0.2006
Toronto	0.1829
Montreal	0.0215
Ottawa-Gatineau	-0.1314
Calgary	-0.1546
Vancouver	-0.3360

Networks AI field
CMA	RSI
Toronto	0.2703
Ottawa-Gatineau	0.0065
Kitchener-Cambridge-Waterloo	-0.0089
Vancouver	-0.1400
Montreal	-0.1804
Calgary	-0.5607

Life and medical sciences AI field
CMA	RSI
Toronto	0.1752
Montreal	0.0434
Vancouver	0.0151
Ottawa-Gatineau	-0.0982
Calgary	-0.2684
Kitchener-Cambridge-Waterloo	-0.8570

Computing in government AI field
CMA	RSI
Kitchener-Cambridge-Waterloo	0.3425
Ottawa-Gatineau	0.1479
Calgary	0.0853
Montreal	0.0790
Vancouver	-0.0961
Toronto	-0.8374

Montréal was identified as a cluster specializing in Telecommunications. Relying on Mila, Montréal has a research focus on a wide range of topics, including deep learning, recurrent neural networks and generative models.^{Footnote 37} In addition to Mila, Montréal has another institution called IVADO, which focuses more on industrial research.^{Footnote 38} This combination has made them a target for international investors, leading to a wide range of deals and generous research funding. Last but not the least, Calgary was found to be specialized in Physical Sciences and Engineering.

Profile of Canadian institutions patenting in the field of artificial intelligence

While analysis of the patent data itself already yields substantial information on innovation in the area of AI by Canadian institutions and researchers, a more complete picture of the environment can be obtained when this data is linked to other data sources. With the support of Global Affairs Canada and CIPO, Statistics Canada linked the patented inventions for Canadian institutions from the EPO-PATSTAT database to data on the characteristics of institutions held at Statistics Canada. This linked data source sheds lights on the industry, size, and ownership characteristics of Canadian institutions patenting in AI.

Industry

A large percentage of patented inventions in the field of AI by Canadian institutions come from the Professional, Scientific, and Technical Services industry. Over the entire 2001 to 2016 period, 41% of AI patented inventions came from this industry. This is far higher than the 23% accounted for by the next largest source, the Manufacturing industry. When patenting across all fields is considered, the situation is reversed. The Manufacturing industry accounts for 47% of the patented inventions in all fields, while Professional, Scientific, and Technical Services industry accounts for 17%.

Figure 30: Breakdown of patent activity by industry sector across all sectors (left) and in AI (right)
NAICS	All sectors	AI
Manufacturing	38,373	81
Professional, Scientific and Technical Services	13,853	146
Others	30,012	129

Institution size

Innovation occurs in institutions of all sizes. While innovation is often associated with younger, smaller institutions, larger institutions have more specialized resources and finances to sustain development efforts. The data on patent activity in AI is broadly consistent with this. Although 59% of the patented inventions in AI were made by institutions with less than 100 employees over the 2001 to 2016 period, the share of AI patented inventions accounted for by large institutions with 500 or more employees was also substantial at 27%. Compared with patenting in all fields, where 44% of patented inventions come from small institutions and 33% come from large institutions, patenting in AI is more concentrated in smaller institutions.

Figure 31: Breakdown of patent activity by institution size across all sectors (left) and in AI (right)
Size	All sectors	AI
Large (500 or more employees)	27,079	98
Medium (100 to 499 employees)	18,671	49
Small (99 or fewer employees)	36,488	209

Ownership

The share of patented inventions in AI by Canadian institutions that are foreign controlled is lower than for overall patented inventions. Over the 2009 to 2016 period, 10% of the patented inventions in AI were made by foreign-controlled Canadian institutions. This is compared to 15% for patented inventions overall.

Figure 32: Breakdown of patent activity by ownership across all sectors (left) and in AI (right)
Foreign Ownership	All sectors	AI
Domestic	34,027	207
Foreign	5,965	23

AI In health

Canada is hailed as one of the leaders in healthcare AI. One initiative that is being heavily leveraged by Canada is the Equitable AI Initiative. This initiative focuses on using AI to provide opportunities in the area of public health not only to analyze complex data but also to design and provide impactful solutions that draw on a wider range of insights. This is done through the initiative’s goals to fund, train and promote knowledge transfer across platforms.^{Footnote 39} At the University of Alberta, scientists are researching and creating prototypes of a bionic arm that can learn and anticipate the movement of its wearer to allow for better and smoother use of the prosthetic.^{Footnote 40} At Humber River Hospital, affectionately dubbed the first all-digital hospital, resides a command centre that tracks the flow of patients from their intake to their discharge, analyzes the data and reports back where there are slowdowns in the process.^{Footnote 41} This allows those working in the hospital to know where and why delays are caused and allows them to fix the problem before it arises. Humber River has taken it a step further and moved from electronic health records to a fully automated hospital.^{Footnote 42} From robots sorting medicine to machines delivering blood samples from patients to the lab, Humber River has made use of AI in almost all areas of the hospital.^{Footnote 43}

Opportunities and challenges for integrating AI at intellectual property offices

AI and IP Policy

The world of IP is changing at an incredible pace. This change is driven by a number of factors, including rising value of intangibles, emergence of new countries as IP powerhouses, convergence of science and technology, increasing volumes of IP filing, increasing complexity of applications, interdisciplinary nature of innovation, and changing nature of work as a result of the integration of new technology.

CIPO is committed to ensuring that our IP system supports transformative technologies, such as AI. As we consider the implications of AI for the Canadian IP regime, Canada is adopting a whole-of-government approach to ensure that the Canadian IP system is well equipped to support the emergence of transformative technologies. Although we are still in the early stages of considering the implications of AI for the Canadian IP regime, CIPO is working with leading thinkers, policy makers, academics, practitioners and international partners to analyze the implications of AI for IP policy and law. In particular, we are engaged in discussion on:

the policy questions being raised for IP, creation and innovation, such as authorship and inventorship;
the copyright considerations in using copyright protected work to train AI algorithms and data; and
the best practices to address the implementation of AI in a manner compatible with core administrative law principles, such as transparency, accountability, legality, and procedural fairness.

AI operations

Transformative technologies have the potential to benefit the administration of the IP system. CIPO is looking to use AI technologies to gain efficiencies both in the delivery of timely and quality IP rights and the provision of a modern service experience. Toward this end, CIPO is launching and managing multiple AI projects.

Endnotes

Footnote 18

Gagne, JF. [2019]. Global AI Talent Report 2019. Retrieved from: https://jfgagne.ai/talent-2019/

Return to footnote 18 referrer

Footnote 19

Element AI. [2019]. 2019 Global AI talent report. Retrieved from: https://www.elementai.com/news/2019/2019-global-ai-talent-report

Return to footnote 19 referrer

Footnote 20

Status of Women Canada. [2014]. Building Canada’s Innovation Economy: Best Practices for Supporting Women in Non-Traditional Sectors. Retrieved from Government of Canada website: https://cfc-swc.gc.ca/abu-ans/wwad-cqnf/wesp-sepf/ntes-en.html

Return to footnote 20 referrer

Footnote 21

Lax, Gema, et al. [2016]. Identifying the gender of PCT inventors. Retrieved from World Intellectual Property Organization website: https://www.wipo.int/publications/en/details.jsp?id=4125

Return to footnote 21 referrer

Footnote 22

Ibid, 20

Return to footnote 22 referrer

Footnote 23

Hunt, Jennifer, et al. [2012]. Why Don’t Women Patent? Retrieved from: https://www.nber.org/papers/w17888.pdf

Return to footnote 23 referrer

Footnote 24

Ibid, 19

Return to footnote 24 referrer

Footnote 25

Innovation, Science and Economic Development Canada. Canada's new superclusters. Retrieved from:https://www.ic.gc.ca/eic/site/093.nsf/eng/00008.html

Return to footnote 25 referrer

Footnote 26

Ibid, 25

Return to footnote 26 referrer

Footnote 27

Innovation, Science and Economic Development Canada. Innovation Superclusters Initiative. Retrieved from: https://www.ic.gc.ca/eic/site/093.nsf/eng/home

Return to footnote 27 referrer

Footnote 28

Social Sciences and Humanities Research Council. [2019]. Canada-UK Artificial Intelligence Initiative. Retrieved from: http://www.sshrc-crsh.gc.ca/funding-financement/programs-programmes/canada-uk_ai/index-eng.aspx

Return to footnote 28 referrer

Footnote 29

CIFAR. About Us. Retrieved from: https://www.cifar.ca/about

Return to footnote 29 referrer

Footnote 30

CIFAR. CIFAR Pan-Canadian Artificial Intelligence Strategy. Retrieved from:https://www.cifar.ca/ai/pan-canadian-artificial-intelligence-strategy

Return to footnote 30 referrer

Footnote 31

United Nations Educational, Scientific and Cultural Organisation. [2018]. Canada first to adopt strategy for artificial intelligence. Retrieved from: http://www.unesco.org/new/en/media-services/single-view/news/canada_first_to_adopt_strategy_for_artificial_intelligence/

Return to footnote 31 referrer

Footnote 32

Bryer, Lanning G. et Scott J. Lebson. [2001]. Intellectual Property Assets in Mergers & Acquisitions. John Wiley & Sons, Inc. New York, United States.

Return to footnote 32 referrer

Footnote 33

Gagne, JF. [2018]. Canadian AI Ecosystem 2018. Retrieved from: https://jfgagne.ai/canadian-ai-ecosystem-2018-en/

Return to footnote 33 referrer

Footnote 34

Vector Institute. About Us. Retrieved from:https://vectorinstitute.ai/about/

Return to footnote 34 referrer

Footnote 35

McKinsey. [2016]. Primer on technology superclusters and a fact base on Canada's Toronto-Waterloo Innovation Corridor. [p. 9]. Retrieved from: https://www.mckinsey.com/~/media/McKinsey/Featured%20Insights/Americas/Tech%20North/Toronto-Waterloo%20Innovation%20Corridor%20white%20paper%20-%20fact%20base-20161213.ashx

Return to footnote 35 referrer

Footnote 36

Canadian Broadcasting Corporation. [2016]. Silicon Valley North? Toronto, Waterloo position themselves as next tech hub. Retrieved from: https://www.cbc.ca/news/canada/kitchener-waterloo/silicon-valley-toronto-waterloo-1.3519032

Return to footnote 36 referrer

Footnote 37

Montréal Institute for Learning Algorithms. Why study at Mila? Retrieved from:https://mila.quebec/en/why-study-at-mila/

Return to footnote 37 referrer

Footnote 38

Deveau, Denise. [2017]. Revolution AI: Why everyone wants in to Montreal's deep-learning hub. Retrieved from Financial Post website: https://business.financialpost.com/entrepreneur/revolution-ai-everyone-wants-in-to-montreals-deep-learning-hub

Return to footnote 38 referrer

Footnote 39

Canadian Institutes of Health Research. Equitable AI activating opportunities. Retrieved from: http://www.cihr-irsc.gc.ca/e/51204.html

Return to footnote 39 referrer

Footnote 40

Zabjek, Alex. [2018]. Researchers create ‘smart’ bionic limbs. Retrieved from:https://www.ualberta.ca/newtrail/research/researchers-create-smart-bionic-limbs.html

Return to footnote 40 referrer

Footnote 41

Humber River Hospital Foundation. Humber River Hospital Command Centre. Retrieved from: https://www.hrhfoundation.ca/commandcentre/

Return to footnote 41 referrer

Footnote 42

Ibid, 41

Return to footnote 42 referrer

Footnote 43

Ibid, 39

Return to footnote 43 referrer

Language selection

WxT Search form

Processing Artificial Intelligence: Analysis from a Canadian Perspective

Analysis from a Canadian Perspective

On this page

Canadian researchers

Patent filing trend

Relative specialization of Canadian researchers

Gender analysis: Female participation in AI patent activity

Canadian institutions

Innovation Superclusters Initiative

Relative specialization of Canadian institutions

Patent filing trend

Distribution of patented inventions

Geographical clusters

Profile of Canadian institutions patenting in the field of artificial intelligence

Industry

Institution size

Ownership

AI In health

Opportunities and challenges for integrating AI at intellectual property offices

AI field	1998	1999	2000	2001	2002	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
Agriculture	6	5	6	2	8	4	10	7	11	6	11	8	21	14	22	53	50	56	72	166
Arts and Humanities	4	15	10	6	12	13	16	24	26	42	31	30	37	32	30	36	43	48	63	118
Banking and Finance	31	22	35	30	33	38	25	36	40	38	54	86	95	103	138	143	198	239	289	556
Business	7	4	19	24	27	19	15	44	27	39	59	101	136	134	129	172	205	254	296	557
Cartography	4		6	3	7	4	3	8	9	14	12	19	24	19	39	49	41	44	72	134
Computing in Government	59	61	56	52	81	71	63	77	82	93	132	155	184	186	211	329	424	508	685	1,141
Document Management and Text Processing	5	5	5	16	16	17	10	30	28	29	21	51	43	33	33	53	48	63	87	185
Education	15	14	27	23	26	22	24	36	31	27	16	35	42	26	48	61	72	85	143	257
Energy management	21	29	11	11	30	27	28	18	24	16	35	42	43	71	105	158	220	235	340	506
Entertainment	12	10	12	19	29	21	20	20	12	29	21	23	25	16	16	32	32	44	53	101
Industry and Manufacturing	50	31	31	41	48	54	53	36	45	49	60	103	97	75	130	187	226	315	431	996
Law, Social and Behavioral Sciences			1	1	1					1	1	1	5		4	4	1	11	9	8
Life and Medical Sciences	94	129	129	152	185	189	199	224	215	276	304	349	452	418	566	639	802	1,002	1,150	1,899
Military	2	3	4	2	1	3	2	6	6	7	13	7	15	14	13	18	22	30	25	37
Networks	64	85	67	56	52	40	52	70	89	77	98	110	113	135	176	234	319	491	686	1,617
Personal Computers and PC Applications	40	37	54	46	66	45	46	68	78	77	111	120	125	110	134	167	220	252	291	551
Physical Sciences and Engineering	155	139	145	137	168	158	167	177	145	156	200	250	245	230	297	368	467	561	782	1,264
Publishing		2	3	2	2	2	2	3	6	3	3	4	3	5	7	7	11	11	13	27
Security	13	16	21	29	32	24	35	49	53	52	68	89	87	74	107	137	140	198	250	394
Telecommunications	76	63	75	87	97	99	90	87	114	122	161	159	201	203	225	311	359	542	651	1,130
Transportation	65	67	93	75	76	87	108	138	140	149	221	236	239	228	347	505	659	1,021	1,523	3,108