wp 32 from prediction to reality · 2 institute for competitiveness & prosperity exhibits...

FROM PREDICTION TO REALITY Ontario’s AI opportunity

32

WP

WORKING PAPERJUNE 2018 32

The Institute for Competitiveness & Prosperity is an independent not-for-profit organization that deepens public understanding of macro and microeconomic factors behind Ontario’s economic progress. Research by the Institute is intended to raise public awareness and stimulate debate on a range of issues related to competitiveness and prosperity. It is the aspiration of the Institute to have a significant influence in increasing Ontario and Canada’s competitiveness, productivity, and capacity for innovation. We believe this will help ensure continued success in creating good jobs, increasing prosperity, and building a higher quality of life. We seek breakthrough findings from our research and propose significant innovations in public policy to stimulate businesses, governments, and educational institutions to take action.

The Institute is advised by Ontario’s Panel for Economic Growth & Prosperity, led by Tiff Macklem.

Comments on this report are welcome and should be directed to the Institute for Competitiveness & Prosperity. The Institute is funded by the Government of Ontario through the Ministry of Economic Development and Growth. The views expressed in this report are the views of the Institute and do not necessarily represent those of the Government of Ontario.

Copyright © June 2018The Institute for Competitiveness & ProsperityISBN: 978-1-927065-29-7

FROM PREDICTION TO REALITY Ontario’s AI opportunity

2 INSTITUTE FOR COMPETITIVENESS & PROSPERITY

EXHIBITS

EXHIBIT 1 Number of academic publications on “Artificial Intelligence” or “Machine Learning,” Canada and select countries, 1982-2017 6EXHIBIT 2 R&D and earnings of companies conducting AI research, 2017 13EXHIBIT 3 Share of labour force with tertiary education, OECD countries and Ontario, 2016 19EXHIBIT 4 Degrees conferred in Computer Science, Ontario, 2005-2016 21

EXHIBIT A Labour share of GDP, Ontario and Canada, 1981-2016 12EXHIBIT B Population with access to the internet, Canada, United States, and China, 2016 17

DISCLOSURE STATEMENT

Christopher Mack, project lead of this Working Paper, has equity in several publicly traded technology companies covered in this Working Papers as part of a broad investment portfolio.

FROM PREDICTION TO REALITY: ONTARIO’S AI OPPORTUNITY 3

CONTENTS

FOREWORD & ACKNOWLEDGEMENTS

EXECUTIVE SUMMARY

CHAPTER 1: WHY IS AI IMPORTANT?

What is Artificial Intelligence? Benefits of Artificial Intelligence Costs of Artificial Intelligence AI raises privacy concerns

CHAPTER 2: THE AI ECOSYSTEM IN ONTARIO AND CANADA

Canada’s global AI strategies Ontario’s strategies for AI success Canada is finding success in AI

CHAPTER 3: HOW CAN ONTARIO AND CANADA WIN?

Create a flexible but sound regulatory framework Orient talent and industrial policy toward the AI ecosystem Incorporate AI into business strategy

END NOTES

PREVIOUS PUBLICATIONS

4

8

6

18

26

22

30


FOREWORD & ACKNOWLEDGEMENTS

Winning in the global AI market

I AM PLEASED TO PRESENT Working Paper 32 of the Institute for Competitiveness & Prosperity, which explores how Ontario can win in the global Artificial Intelligence (AI) market. AI has been part of the public rhet-oric recently, as politicians and policymakers consider how to promote AI as a source of productivity growth and how to support workers who lose their jobs due to automation. This requires an understanding of how AI is likely to disrupt industries and change how we live, work, and learn.

This Working Paper attempts to explain the impact of AI: its benefits and its drawbacks. AI will, in some industries, automate tasks and replace jobs. For example, many customer service and transportation jobs will likely be done by computers instead of humans. However, as with other industrial revolutions that have come before, AI will create new jobs and make us more productive, raising our standard of living.

AI’s strength lies in its predictive ability, learning from patterns found in data. Data is only valuable if we have the capacity to analyze it. AI allows for the analysis of vast troves of structured and unstructured data and can increasingly provide more accurate results than humans. At the same time, this predictive ability raises serious concerns about data privacy and abuse.

AI leaders in China, and to a lesser extent in the United States, are operating in relatively flexible regulatory regimes that allow them to collect and analyze large amounts of user data. The abundance of data due to large customer data-bases and much looser privacy laws gives these companies the upper hand in the global AI market because the more data the AI can access and use, the more accurate its results become.

In Ontario, much has been done to understand AI’s potential, to leverage oppor-tunities, and to allay the public’s fears. The province has a rich AI ecosystem led by some of the world’s best AI scientists and business thinkers, many of them at the University of Toronto, including Geoffrey Hinton, Raquel Urtasun, and Ajay Agrawal. Working with the AI clusters in Montréal and Edmonton can place Canada in a winning position in the global AI market by being an important part of global supply chains.

Tiff MacklemChair, Ontario’s Panel on Economic Growth & Prosperity

Dean, Joseph L. Rotman School of Management


In order to secure this position however, there are several recommendations key to accelerating and sustaining the development and implementation of AI put forward in this Working Paper.

First, it is imperative that the province stay ahead of the curve and support the research and development of this technology, so that we stay at the leading edge of AI innovation. Given the vast investments being made in other countries, this will be hard, but if we do not try we will lose our lead for sure.

Second, support AI companies by creating regulatory sandboxes for them to innovate and test their technologies.

Third, work alongside industry to create sound data privacy regulations that safeguard against misuse while maintaining economic competitive-ness. This is a difficult balance to strike. However, if Canada fails to be a part of the global AI market, then we will face the disruptive effects with few of the economic benefits.

We thank Dr. Dan Trefler at the Rotman School of Management (who is also a member of Ontario’s Panel on Economic Growth & Prosperity) and Dr. Avi Goldfarb for their leadership and guidance on this Working Paper. We also gratefully acknowledge the funding support from the Ministry of Economic Development and Growth. We look forward to sharing and discussing our work, and welcome your comments and suggestions.

The province has a rich AI ecosystem led by some of the world’s best AI scientists and business thinkers.

WHAT IS ARTIFICIAL INTELLIGENCE (AI)?

Artificial IntelligenceIn the broadest sense, AI is the ability of a computer to perform tasks commonly thought of as intelligent or human. AI encompasses a number of techniques to simulate intelligence, which range from traditional, rule-based programs to algorithms that learn and improve to develop an area of expertise.

SOLIDIFYING ONTARIO’S PLACE IN THE GLOBAL ARTIFICIAL INTELLIGENCE ECOSYSTEM

Machine learningFeeds data into algorithms, allowing it to learn from examples and make predictions about new data.

Neural networkSubset of machine learning algorithms. Information passes between branches or layers in a similar way to neurons in the brain. These networks calculate probabilities of results given a set of inputs.

CANADA HAS A GROWING AI ECOSYSTEM BUT IT MUST CONTINUE TO MAKE INVESTMENTS TO OVERCOME THE COSTS AND MAXIMIZE THE BENEFITS OF AI:

THE INCREASED PRODUCTIVITY TO THESE INDUSTRIES:

THESE ARE THE POTENTIAL COSTS OF AI:

Risks of inequality from

job loss and displacement

Large firms may crowd out small players

Ethical issues such as privacy

BUT THEY ARE OUTWEIGHED BY THE BENEFITS:

Health care Financial services

Manufacturing …& more

FOR CONSUMERS, GOODS & SERVICES WILL BE:

Lowerin cost

Morepersonalized

RECOMMENDATIONS ON HOW ONTARIO & CANADA CAN WIN IN THE GLOBAL AI ECOSYSTEM:

Create a regulatory sandbox for innovation

Balance trade policy to protect consumers

Retain skilled talent

Align skill sets learned in post-secondary institutions to AI needs

Understand the trade-offs involved in pursuing AI


EXECUTIVE SUMMARY

WHAT IS ARTIFICIAL INTELLIGENCE (AI)?

Artificial IntelligenceIn the broadest sense, AI is the ability of a computer to perform tasks commonly thought of as intelligent or human. AI encompasses a number of techniques to simulate intelligence, which range from traditional, rule-based programs to algorithms that learn and improve to develop an area of expertise.

SOLIDIFYING ONTARIO’S PLACE IN THE GLOBAL ARTIFICIAL INTELLIGENCE ECOSYSTEM

Machine learningFeeds data into algorithms, allowing it to learn from examples and make predictions about new data.

Neural networkSubset of machine learning algorithms. Information passes between branches or layers in a similar way to neurons in the brain. These networks calculate probabilities of results given a set of inputs.

CANADA HAS A GROWING AI ECOSYSTEM BUT IT MUST CONTINUE TO MAKE INVESTMENTS TO OVERCOME THE COSTS AND MAXIMIZE THE BENEFITS OF AI:

THE INCREASED PRODUCTIVITY TO THESE INDUSTRIES:

THESE ARE THE POTENTIAL COSTS OF AI:

Risks of inequality from

job loss and displacement

Large firms may crowd out small players

Ethical issues such as privacy

BUT THEY ARE OUTWEIGHED BY THE BENEFITS:

Health care Financial services

Manufacturing …& more

FOR CONSUMERS, GOODS & SERVICES WILL BE:

Lowerin cost

Morepersonalized

RECOMMENDATIONS ON HOW ONTARIO & CANADA CAN WIN IN THE GLOBAL AI ECOSYSTEM:

Create a regulatory sandbox for innovation

Balance trade policy to protect consumers

Retain skilled talent

Align skill sets learned in post-secondary institutions to AI needs

Understand the trade-offs involved in pursuing AI



CHAPTER 1

Artificial Intelligence (AI) has been discussed widely as firms develop and apply this technology into its products and services and governments find ways to support companies while protecting citizens from any potentially harmful effects. Understanding what AI is and its costs and benefits is integral to determining how best to leverage this technology to maximize the benefits.

WHY IS AI IMPORTANT?

ARTIFICIAL INTELLIGENCE (AI) is already commonplace and will continue to play a significant role in everyday life. Continued advancements in AI is expected to introduce sweeping changes across the economy and society at large. The effects are likely to be similar to those of the Industrial Revolution, when most livelihoods were disrupted and living standards experienced real and sustained growth. A revolution of this magnitude will reshape how society organizes itself while offering poten-tially unprecedented levels of wealth and prosperity. Therefore, the objective of this Working Paper is to consider how Ontario and Canada can stay competitive in the AI market. It is imperative that new AI-enabled jobs are created domestically, that the labour force is prepared for this disruption, and that individuals are supported as they navigate the new economic landscape.

What is Artificial Intelligence?

Artificial Intelligence is expressed most generally as “the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings.”1 One of the longest running and most popular tests of AI is the Turing Test, set forth by Alan Turing in 1950.2 The Turing Test asks: can an AI fool a human judge into believing it is actually another human carrying on a conversation?

Current AI research is primarily focused on a sub-branch called machine learning, which builds complex statistical models from the data fed into it. These statistical models attempt to optimize their predictive capability on new data. They make predictions given a set of known information.3 This predictive capacity is harnessed to allow AI to excel at complex tasks typically thought of as solely done by humans.


Artificial Intelligence will bring more good than badArtificial Intelligence will come with significant benefits to businesses, consumers, and governments. However, like any technological advancement, it will create short-term disruption, with AI replacing human labour in many tasks. This explains the fears that AI will negatively disrupt the economic well-being of individuals.

As AI applications are developed across the globe, no amount of domestic resis-tance will prevent AI from influencing global demand for higher quality goods and services enabled by AI. Firms that refuse to utilize AI will not be able to compete with AI-enhanced goods and services created by foreign companies that have embraced the technology.

Due to the interconnected nature of the global economy, disruption is coming; a government’s best course of action is to support AI. In embracing the advancement of AI and its supporting ecosystem, Canada can position itself at the forefront of AI and machine learning development. Canada has the potential to shape the outcome and applications of AI globally, and maximize its piece of the economic pie. The alternative is to place exorbitant restrictions on the use of AI in the name of protecting jobs, effectively blocking the progress of AI. This approach would ultimately leave Canadian companies lagging behind the rest of the world, cost jobs, irrep-arably damage the economy, and, still fail to prevent applications of AI from being imported. The risk of obstructing AI development far exceeds the down-side of creating an environment to ensure domestic success. Job loss is

Note: Includes all available document types, including conference papers and books. The top five document-producing countries were selected alongside Canada.Source: Institute for Competitiveness & Prosperity analysis based on data from Scopus.

1982 1987 1992 1997 2002 2007 2012 2017

Publicationsper year

0

1,000

2,000

3,000

4,000

5,000

6,000

7,000

8,000

India

Germany

UnitedKingdom

China

Canada

US

EXHIBIT 1 Number of academic publications on "Artificial Intelligence" or "Machine Learning," Canada and select countries, 1982-2017

unavoidable, therefore Canada must ensure corresponding job gains in other areas to offset this result.

Canada punches above its weight in academic AI researchThe number of academic works published annually on AI has exploded, particularly in the United States and China (Exhibit 1). The world has recog-nized the value AI will hold in the future, and many countries are presently investing heavily in research and devel-opment (R&D) in pursuit of leading the global AI economy.

Canada punches above its weight for its population size in generating academic publications, potentially due to its early role in neural network research.4 This advantage can easily slip away if the surrounding ecosystem does not create opportunities for commercialization.


Benefits of Artificial Intelligence

AI is often touted as a major disruptor of firms and economies, but it also has the ability to generate many positive benefits for customers across all income levels and to raise productivity. Both of these benefits could help Ontario close its prosperity gap.

AI will boost productivityIncreased industry adoption of AI will reshape the economy and has the poten-tial to unlock unprecedented levels of productivity, defined as output per hour of human labour. By 2030, AI could contribute an additional $19.8 trillion or 14 percent to global GDP.5

Productivity will be increased through two main channels. First, human labour will be replaced or significantly reduced. Companies are already working on commercial AI applications to replace routine tasks that use codifiable patterns detectable by an algorithm over time.6 Jobs at significant risk of near-term automation include tellers, telemar-keters, production workers, and truck drivers.7

The value created remains consistent or even grows, but spread over fewer employees. There are already examples of automation replacing human labour in today’s economy: fast food restau-rants replacing cashiers, with self-serve kiosks.8 These kiosks follow rules and are incapable of thinking or prob-lem-solving like human cashiers. A more complex example is a call centre that uses machine learning to understand clients’ questions, and return accurate information to assist with their queries. To do this, a machine learning algorithm must identify speech as accurately as a human, interpret the received instruc-tions, and formulate the response sourced from a supplied database, or from the internet more broadly.9 It is

predicted that by 2020, 85 percent of customer service interactions will be handled by AI without the assistance of a human worker.10

Second, human labour for some tasks will be enhanced by AI in much the same way it is enhanced by capital investment in machinery and equip-ment. Complementary AI will directly increase the value creation potential of workers in jobs that are less likely to be automated, including elementary school teachers, lawyers, social workers, photographers, financial analysts, and software developers.11 Similar to the Industrial Revolution, labour will find itself working alongside the capital investments that corporations make. During the industrial revolution the method of weaving fabric, the loom, was improved with the invention of the flying shuttle in 1733 and the power loom in 1785, advancements which greatly augmented the productive capacity of a single weaver.12 Lower costs and greater demand for cloth domestically and internationally are attributed to the increased productivity of the power loom, which eventually created jobs through higher volumes.13 When a task is complemented by a new technology, costs are lowered, efficiency and accuracy are improved, total output is increased, and innovative solutions to problems not previously solvable are created. Additional examples of AI enhancing industrial activities can be found in the Institute’s blog by clicking: Augmenting high value industries with Artificial Intelligence.

AI will benefit consumersFor individuals, there are two distinct benefits. First, the additional produc-tivity unlocked by AI will drive down the cost of existing goods and services, increasing the purchasing power of consumers.14 Second, consumers will benefit from more personalized prod-ucts and services.

As AI continues to integrate into regular economic activities, individuals will be able to afford the same lifestyle for less. As has been observed after the introduction of other productivity-en-hancing technologies to society, the cost of producing existing goods and services will decline. Food became far more abundant with the adoption of specific technologies, allowing fewer farmers to produce a greater quantity of crops. Irrigation, selective breeding, hybrid-ization, and genetically-modified crops have all unlocked new levels of supply while reducing the cost to consumers.15

The predictive capacity of AI can reduce the time consumers require to deter-mine the ideal product or service for their specific needs. As businesses begin inferring more about each user through data collected, they can offer more sophisticated and personalized clothing, travel, real estate, internet searches, and media recommendations.

By 2030, AI could contribute an additional $19.8 trillion or 14 percent to global GDP.

https://www.competeprosper.ca/blog/augmenting-high-value-industries-with-artificial-intelligence


Furthermore, the widespread adop-tion of AI could even facilitate price differentiation based upon preferences and ability to pay. At first glance, this sounds undesirable, but it already exists. For example, the children and seniors’ separate pricing for movie tickets, where tickets sold at different pricing levels are for the same seat. This may be advertised as well-intentioned by the company but actually exists to ensure that a group that might otherwise avoid the service partakes in the economic exchange, raising total sales.16 The adoption of AI coupled with e-commerce could allow pricing based upon a predic-tion of each individual’s willingness to pay. This would lead some individuals to consistently pay more than others, but would allow a cohort of low-in-come individuals to access products or services that are otherwise unafford-able, as long as the price was above the absolute minimum point where a firm was willing to conduct an exchange.17

Wealth and income primarily exist to facilitate current or future consumption, and so reducing this gap also decreases actual inequality.

Marketing firms also use AI to target ads based on browsing habits or search histories. Alphabet (Google) and Facebook, the world’s biggest digital advertising providers, optimize who sees specific ads based on the likelihood of translating a pageview into a genuine interaction or purchase.

Additionally, AI has the ability to improve existing product offerings rather than personalize them. Alphabet’s email service blocks 99.9 percent of junk and phishing emails sent to users, saving the consumer time, and protecting from potentially damaging phishing emails.18 Also recently implemented is Google’s smart reply AI, allowing Gmail users to send quick replies like “Sounds good to me” based on the contents of the email.19

Artificial intelligence could be harnessed to minimize unconscious bias. Hiring or university program acceptances could be candidates for AI assistance. Some companies already utilize blind hiring, which works to ensure selected candidates are based solely on relevant indicators. Artificial intelligence could be similarly applied, utilizing internal company performance data to evaluate a new candidate’s fit in the role being filled.

Costs of Artificial Intelligence

Despite the benefits of AI, there are also many costs that can generate fear and uncertainty. However, by understanding these costs, the province can prepare to curb the negative effects.

Risk of inequality The Institute recognizes that advance-ments in AI will not benefit everyone in society equally. Some people, indus-tries, and jobs will be more negatively affected, and the effects are unlikely to be offset by the broader economic advancements and rising prosperity. The effects may be similar when a new trade deal is implemented: some industries will expand and their workers will benefit but there will also be workers who face job loss or wage declines in industries opened up to foreign competition.20

The overall benefit of AI to the economy is decidedly positive, as the majority of people will be at least marginally better off, but those experiencing job loss without reinstatement will experi-ence acute negative effects. Despite the disruption, the long run expectation is for human labour to find gainful employ-ment alongside automated systems or in tasks ill-suited to automation. (See People vs machines).

In the short run, skill sets easily substituted by AI are likely to face displacement. This cohort may struggle to find new employment at a comparable income after job loss. If the province fails to adjust its skills training and retraining programs for the economy of the future, a subset of Ontarians could be permanently disadvantaged. Older individuals in particular may face barriers re-entering the labour market after displacement. If their formative years were prior to the widespread adoption of computers, and their jobs did not require using digital interfaces, they are unlikely to transition well even in many minimum wage positions that still require the ability to interact with a computerized system. For example, even restaurants orders are entered into a digital point of sale machine. The ability to work alongside a digital display or input device will grow in importance with the widespread adoption of AI.


Note: Gross mixed income and taxes are omitted from total GDP.Source: Institute for Competitiveness & Prosperity analysis based on data from Statistics Canada CANSIM Table 384-0037.

1981 1986 1991 1996 2001 2006 2011 2016

Compensation/(compensation+

operating surplus)

0

Canada

Ontario

60

62

64

66

68

70

72

74

76

78%

EXHIBIT A Labour share of GDP, Ontario and Canada, 1981-2016

SOME PAINT A PICTURE of a future where the capitalist system has failed, almost all human labour has been automated, and only those who own the production-creating capital reap the benefits. Instead of this bleak future, the Institute looks to the past for indications of what will come to pass.

Professors Daron Acemoglu and Pascual Restrepo argue that, while significant displacement effects are likely to occur in the labour market as a result of AI, automation, and machines, the countervailing effects of increased productivity, capital accumulation, deepening of automation, and reinstatement will offset these advancements and ensure demand for human labour.a

Productivity will increase due to the automation of tradi-tionally labour-intensive tasks. Each worker will produce more output with the help of AI, and the economy will subsequently increase in size. The growth in economic activity will be coupled with increased demand for consumption. As people become wealthier, their demand for goods and services rises, which will spur demand for labour in tasks not currently automated.

Second, as capital is continually invested in automation, the total stock of capital available to workers grows. Worker productivity is complemented by capital investment, and if employment levels stay constant then wages will grow to reflect the increases in productivity due to the capital investment. The complementary nature of capital and labour as described in the Cobb-Douglas production function ensures that as capital accumulates and becomes more expensive, the demand for relatively cheaper labour will balance out production.b

Continuing automation not only affects current tasks done

by people, but also improves tasks that have already been automated. This effect causes no additional displacement of human labour but raises overall productivity, and as a result the productivity attributable to each worker. Continuing automation will increase the demand for labour as the economy expands and the cost of capital increases.

Finally, technological advancements shift production from labour to capital, which might suggest a significant decoupling of wages and output (GDP). In general, this has not been the case: labour still makes up 68 percent of compensation and operat-ing surplus, despite the recent economic downturn observed recently (Exhibit A). Operating surplus can be thought of as “income… accruing to the capital factor of production from the production of goods and services.”c A decline in this ratio represents a greater portion of income flowing to capital invest-ments businesses, rather than employees working alongside it.

Over years of advancement, it would be expected that labour’s share of GDP would be significantly lower if there were not a powerful countervailing force creating new, valuable tasks for human labour. However, the creation of new tasks may be the most significant factor counteracting the displacement effect of automation. The reinstatement effect is created when automation reduces the labour share of production and proportionally increases the cost of additional automation. The creation of new labour tasks is then incentivized, as they are relatively cost efficient. Ontario should take comfort knowing that, through this reinstatement effect, human labour will never be replaced but will continue to adapt to the new economic landscape.

People vs machines


Market power of large firmsCurrently, top tech multinationals in the US such as Amazon, Alphabet, Intel, Microsoft, Apple, Facebook, IBM, NVIDIA, as well as Chinese companies such as Tencent, Alibaba, Baidu, and JD.com are investing heavily in AI (Exhibit 2). These corporations have significant influence on the development of AI (what tasks will be automated) and the reasons for their use (how moral and ethical dilemmas are resolved).

These firms operate in their own best interests, which may not always align with those of the broader society. Governments have a responsibility to ensure products and services do not negatively impact society and consumers, and stop firms from utilizing their size and market power to exert undue control. In fact, some of these firms may become so integral to society that they cannot be allowed to act freely. For example, internet service providers

(ISP) are subject to net neutrality rules that restrict their ability to treat different types of data or users differ-ently.21 Without this restriction, ISPs could slow down, block, or charge addi-tional fees to access specific websites or online services such as Netflix.22

Incumbency advantageIt is unclear if the rise of machine learning will create significant barriers to productive operations, or if it will enable small firms to compete with larger players or specialized tasks. Instead, machine learning is likely to have two distinct effects working counter to one another.

Substantial investments in technology and expert labour required to set up a working AI or machine learning algo-rithm could drastically raise upfront costs for new firms.23 This could in turn restrict small firms from utilizing AI in their production, and further expand

the market power of existing large firms capitalizing on their size to capture additional market share. Large AI firms may even expand horizontally to enter other industries.

Conversely, the offsetting factor is the possibility of pre-built AI becoming readily available, requiring minimal additional training to improve a firm’s operations. The cost to the firm could be further reduced by offloading hardware requirements to cloud-based services offered by companies such as Amazon, Google, and Microsoft.24 Removing the need to invest in Graphical Processing Units (GPUs) or Tensor Processing Units (TPUs) spreads the cost of running AI over the lifespan of its usefulness, similar to amortization, rather than frontloading the entire cost.25

Note: Amazon does not separate the costs associated with GAAP R&D activities (ASC 730-10-55-1) from routine or periodic alterations to existing products (ASC 730-10-55-2). USD and HKD were converted to Canadian dollars using December 2017 exchange rates.Source: Institute for Competitiveness & Prosperity analysis based on data from The Wall Street Journal.

0 10 20 30 5040 $7060R&D expenses and earnings (C$ 2017 billions)

R&D expenseEarnings

EXHIBIT 2 R&D and earnings of companies conducting AI research, 2017

Amazon

Alphabet

Intel

Microsoft

Apple

Facebook

IBM

Tencent

Alibaba

Baidu

NVIDIA

JD.com


but as accuracy improves, the service would increase in value, allowing Amazon to capture an enormous share of total retail sales (currently around four percent of the US retail market).31 The consumer’s desire to seek products offered by competitors would shrink, as most of their consumption needs would already be satisfied.

Despite the widespread efforts by current tech multinationals to drive down prices, increase convenience and improve user experience, these firms likely crowd out smaller entrants with potential to offer more innovative products. As such, the best interests of society may eventually be served by controlling or even breaking up these large firms.

Traditionally, and especially in the US, antitrust has acted to prevent firms from using monopolistic power to extract additional profits through price fixing and collusion. The Sherman Antitrust Act of 1890 (known as the Sherman Act) prevents the extraction of consumer surplus by a monopoly through artificially inflated prices, but does not prevent natural formation of a monopoly through standard busi-ness practices and value offering.32 Working to protect a legitimately-earned monopoly through uncompetitive or illegal means is barred. In general, the Sherman Act protects consumers from exploitation while allowing aggressive competition among firms. Seen through this lens, many of today’s tech giants pass this anticompetitive scrutiny: they often drive prices down instead of up, and advance value offerings for consumers. Consumers’ best interests may be used as a guise by companies acting primarily to protect market share. Additionally, even if antitrust rules do not prevent the domination of large firms, there is still an argument that technology companies can be easily

It is currently unclear which effect will be the most dominant in Ontario’s future, but the potential exists for incumbent firms to benefit from an undeserved advantage, due to their earlier entry and associated size.

Incumbent firms also benefit greatly from detailed data they have accumu-lated on millions of users. This contrasts with new market entrants who begin training AI using limited, low quality, or public datasets. Larger datasets are more likely to contain rare occurrences or exceptional circumstances.26 The ability for AI to handle these extraor-dinary circumstances can translate to avoiding a false diagnosis on a medical imaging exam or may be the difference between retaining or losing a client to a search engine competitor.27 In the case of search engines, customers stay with a search provider because of accurate performance on rare searches.28

Even the largest multinationals are unlikely to have “enough” data, as the predictive power of more data will continue to advance and unlock new business strategies that did not previ-ously exist.

Rotman School of Management Professors Ajay Agrawal, Joshua Gans, and Avi Goldfarb put forward a thought experiment highlighting the predic-tive power of data for a business.29 Amazon currently recommends items, but its relatively low accuracy inhibits customers from buying them. If this algorithm’s predictive power improved to the point where most recommended items translated into sales, then Amazon’s business model would dras-tically transform.30 Imagine if Amazon automatically shipped items selected by AI to customers, which could be returned for free. At low levels of predic-tive accuracy this process would annoy the consumer and add substantial costs,


displaced if they spurn consumers or make a critical mistake that changes public sentiment, due to the “one click away” nature of internet competition.33 Examples include Yahoo, AOL, and Myspace, all of whom dominated their respective markets until a more nimble player ousted them.

Despite the appearance that consumers are unaffected, there is mounting evidence that some companies utilize their size to undermine new entrants and prevent superior offerings from reaching the market.34 In 2017, the European Union levied a $3.5 billion fine against Alphabet (Google) for favouring its own shopping results over competitors.35 The amount of the fine was determined not only by damages caused by Alphabet, but also its size and profitability. The EU concluded that the amount was necessary to sufficiently impact Alphabet’s financials to dissuade them from further uncompetitive actions, as well as those of other tech-nology firms.

In Canada, antitrust rules were origi-nally enacted in 1889, one year prior to the US Sherman Act, and have evolved to cover many of the same antitrust categories, namely: mergers, price fixing, quality or quantity agreements among competitors, bid rigging, trade associations, resale price maintenance, price discrimination, promotional allowances, predatory pricing, refusal to deal, exclusive dealing, tied selling, market restriction, and abuse of domi-nant position.36

Canada can follow the EU by levying significant fines on anticompetitive behaviour in order to ensure domestic firms have the ability to reach the market and compete fairly on their own merits. This strategy, of course, is not without risk. Enforcing overly protectionist antitrust regulations could limit the investment of foreign companies within Canada, as well as decrease service offerings to Canadian consumers; firms may be reluctant to bring offerings to Canada for fear of overt penalties. Finally, strict antitrust rules that only target foreign compa-nies could be seen as a tool to skirt trade agreements. They can act as a protectionist policy to unfairly benefit domestic firms over foreign firms.

Even the largest multinationals are unlikely to have “enough” data, as the predictive power of more data will continue to advance and unlock new business strategies that did not previously exist.


AI raises privacy concerns

Advancements in data collection, storage, manipulation, and the predic-tive capacity of AI brings legitimate privacy concerns. Protection of data privacy can be compared to environ-mental regulations.37 The demand for environmental protection grows with income level, resulting in wealthier countries imposing carbon restrictions while less wealthy countries forego environmental protections in favour of access to traditional goods and services.38 Data privacy appears to follow a similar pattern where wealthier countries have a greater desire to ‘purchase’ privacy.39 Environmental regulation and data privacy also share the risk of a “race to the bottom” as less regulation is preferred when the focus is on growing the economy.40 (See Is China a threat?)

Individuals have an inherent desire for privacy, but appear to undervalue their privacy when making an “exchange” of data for a service or good, especially when substantial personal informa-tion are traded for a nominal reward, payment, or service.41 Individuals enable this paradox because they do not fully understand how their data is being used. This occurs due to three properties of data: persistence, ability to be repur-posed, and spillovers.42

When individuals willingly offer up personal data, they may not be aware of the length of time the data will be useful to the collector. Additionally, there is evidence that individuals’ preferences surrounding data change over time, with the result being that some individ-uals are likely exchanging their data too easily or cheaply.43 This coincides with an individual’s inability to determine future use of the same piece of data. There may be limited uses for the data within today’s technology framework, but future algorithms may be capable of deriving undesirable predictions from the data.44

The second issue concerns repurposing long-lasting data for predictions outside of the intended use. Unexpected predic-tive power is sometimes achieved when combining large datasets, predictions of personal preferences, and extracted information an individual did not intend to divulge.45

Finally, the creation or sharing of certain types of data impacts more than the direct user. A user’s friends or family are likely to have information revealed by a consumer’s actions, as more comprehen-sive datasets link individuals and their preferences together. Even strangers risk their data being distributed by the original user. As an example, a photo could include other individuals who did not consent to having their information

shared.46 With facial recognition software this data could supplement a company’s profile on these third-party individuals.47 Given these three char-acteristics, the question that should be asked is not, “what do large firms know about us?” but rather, “what do they not know?” and “what are they doing with this information?”

In contrast, personal information is still required for a company to deliver tailored services to customers, as well as drive innovation, so an optimal data privacy policy should not be completely restrictive.48 There is evidence that highly restrictive data policies have significant negative effects on business operations. For example, the EU’s ePri-vacy Directive has been attributed with decreasing the effectiveness of digital advertising by 65 percent.49

The EU is taking further steps to protect citizens’ data from misuse. It adopted a new regulatory standard called the General Data Protection Regulation (GDPR) in 2016, which became enforce-able on May 25, 2018. This regulation applies uniformly across all EU coun-tries, and the companies that operate within them, in order to “protect and empower all EU citizens’ data privacy and to reshape the way organizations across the region approach data priva-cy.”50 These regulations, enforceable with fines of up to $29 million, provide citizens the following rights: breach notification, the right to access, and the “right to be forgotten.”51 These rights work to protect domestic consumers, but all rights come with a corresponding obligation. The cost required to comply with these regulations will impose some penalty on operating and innovating within the EU, which may be worth-while for society, but may impact the competitiveness of the region.

The question that should be asked is not, “what do large firms know about us?” but rather, “what do they not know?” and “what are they doing with this information?”


AS OF 2016, China’s population was nearly 1.38 billion, and 53.2 percent of its population had access to the internet (Exhibit B).d This means that almost 650 million people are currently without access to the internet and the digital economy in China. There is still a tremendous market for new internet users, given the number of people who will join China’s digital economy over the coming years.

While China currently lags behind the US in AI technology and research, its government-led focus places it on a winning trajectory. In 2017, the State Council issued the New Generation AI Development Plan, which highlights the initiatives that government, business, and academia will be undertaking together.e

Rotman School of Management Professors Avi Goldfarb and Dan Trefler argue that the Chinese government has the ability to shield domestic firms from foreign competition, ensure access to consumer data, and provide financial and regulatory support.f China has already demonstrated a willingness to shield its companies from foreign competition through a variety of direct and indirect blockades. Alphabet attempted to provide its popular Google search engine to China, but ran into regulatory backlash due to its refusal to censor internet searches. In 2010, Alphabet began redirecting Chinese users to Google Hong Kong, which does not face the same censorship rules as China. This action was met with a block: Google search has been inaccessible in China since 2014.g

China has also imposed substantial barriers on electronic entertainment companies. The most popular video game makers in the world, such as Activision Blizzard ($9 billion revenue in 2017), which produces World of Warcraft, StarCraft, Hearthstone, and Overwatch, has to utilize a domestic publishing company, NetEase, to localize their games.h Localization involves turning over all assets to a foreign company that makes necessary changes to appease the Chinese government. The regulations put up implicit trade barriers that essentially bar foreign competition in this technology-heavy industry unless they transfer revenue and intellectual property – and hence control – to a Chinese firm.

By carefully controlling the actions of foreign firms through protectionist regulations, while simultaneously promoting intense R&D with unfettered access to consumer information and minimal regulation, China’s technology firms have the opportunity to overtake technology giants in the West.i

Is China a threat?

Note: China's projected population with access to the internet is based on Canada's 2016 usage rate.Source: Institute for Competitiveness & Prosperity analysis based on data from World Bank.

Access givenCanada's rate

Totalpopulation

Actual access

Population with access to the

internet (millions)

0

200

400

600

800

1,000

1,200

1,400

Canada ChinaUS

EXHIBIT B Population with access to the internet, Canada, United States, and China, 2016


CHAPTER 2

For Ontario and Canada to benefit from AI, it must leverage its existing assets. Canada has a lot to offer: a highly educated workforce, ties to countries around the world, leading researchers, and concerted government support.

THE AI ECOSYSTEM IN ONTARIO AND CANADA

CENTRAL TO ANY SUCCESSFUL AI ECOSYSTEM is talent - or those individuals who develop, apply, and commercialize AI. While small in size, Ontario is building the skill sets necessary to be a world leader in the development and commercialization of AI. The province has a strong base to build upon as Canada has a highly-educated working age population. As of 2016, 56 percent of Canada’s working age population (between 25-64 years) had a tertiary education, the highest among OECD countries (Exhibit 3).52 Ontario is Canada’s most educated province, further extending this lead by seven percentage points. Within this broadly educated labour force are the skill sets necessary to create, facilitate, sell, advertise, and govern AI.

On the other hand, Canada has thus far produced few world class AI or technology companies that can compete globally. In fact, the best and brightest in computer science and Artificial Intelligence have long left Ontario or Canada for the US or countries abroad who offer exciting jobs with six to seven figure starting salaries. Recognizing this, there are many national efforts to stem the “brain drain.”

Canada’s global AI strategies

The federal government has invested heavily in the AI ecosystem, from funding an AI-focused supercluster to executing a strategy that will also benefit provinces.

Pan-Canadian Artificial Intelligence (AI) StrategyThe Canadian Institute for Advanced Research (CIFAR) manages $125 million of federal funds for the execution of the Pan-Canadian Artificial Intelligence Strategy.53 CIFAR works directly with three interconnected regional initiatives in Edmonton, Toronto and Montréal to achieve its goals. The Strategy aims to bolster the output of AI research and the collaborative capacity of distant geographies; attracting new


actors, across a variety of related clus-ters. Due to their scope, a formalized non-profit organization (which the Institute calls a supercluster organi-zation) is needed to ensure consistent collaboration as member organizations implement cluster initiatives. These initiatives convene industry players, with collaborative organizations, funding, and research institutions to enable projects that individual compa-nies would be incapable of tackling alone due to a lack of funds, ideas, or specific skill sets.

industries) that actively feed into one another through both collaborative and competitive means.57 Their close proximity encourages the diffusion of knowledge, benefitting all members of the cluster. The power of clustering that created the world’s most iconic indus-trial regions such as Wall Street in New York, or Silicon Valley in California.58 Competitors have conglomerated near one another, attracting the necessary supporting industries, and developing a deep pool of talent from which to draw.

A supercluster builds upon this by expanding the industrial and geographic scope to a larger group of firms and

talent and commercializing for nation-wide economic benefit. In addition to funding and working with the three AI Institutes, CIFAR also funds AI research chairs, researching societal impacts, and holding educational seminars.54

Innovation Superclusters InitiativeCanada has recently earmarked $950 million over five years to fund the Innovation Superclusters Initiative.55 Clusters are “geographically proximate groups of interconnected companies, suppliers, service providers, and associ-ated institutions.”56 Put simply, a cluster is a group of companies (along with its employees, suppliers, and associated

Note: Provincial data from CANSIM was appended onto the OECD country level data. Tertiary education consists of certificates or diplomas from college, as well as certificates below a Bachelor’s degree, Bachelor’s, Masters, and Doctoral degrees, excluding trades certificates.Source: Institute for Competitiveness & Prosperity analysis based on data from OECD OECD.Stat Table Educational attainment and labour-force status, and Statistics Canada CANSIM Table 477-0135.

Share of 25 to 64-year-olds

with tertiary education (%)

0 10 20 30 40 50 60 70%

MexicoItaly

TurkeySlovakia

Czech RepublicHungaryPortugal

GermanyPolandGreece

SloveniaAustria

LatviaFrance

SpainNetherlands

New ZealandBelgium

DenmarkEstoniaIceland

SwedenSwitzerlandLuxembourg

NorwayFinland

AustraliaUSA

United KingdomSouth Korea

IsraelJapan

CanadaOntario

EXHIBIT 3 Share of labour force with tertiary education, OECD countries and Ontario, 2016


The Innovation Superclusters Initiative announced five finalists in February 2018.59 The AI-powered Supply Chains Supercluster (SCALE.AI) based in Québec was one of the finalists.60 This supercluster will work with “retail, manufacturing, transportation, infrastructure, and information and communications technology sectors together to build intelligent supply chains.”61 SCALE.AI is expected to augment the supply chains of Canadian firms, and increase global competitive-ness through exports.

Ontario’s strategies for AI success

In addition to the federal initiatives, there are also a number in Ontario, primarily focused on education.

Vector InstituteVector Institute is funded by CIFAR, government of Canada, government of Ontario, and corporations.62 The Vector Institute is also affiliated with the University of Guelph, University of Toronto, and University of Waterloo.63 The Vector Institute has multiple goals in the domestic AI ecosystem: increasing the supply of available talent with an AI skill set, ensuring professors groom an ever-increasing pool of talent, and attracting world leading companies and jobs conducting leading-edge research and product design.64

A critical mass of talent is necessary to attract world-leading companies that will shy away from domestic invest-ments if there is a dearth in talent for their projects, or will face overly fierce competition for the limited pool of talent available. A virtuous cycle of talent and industry attraction should estab-lish itself as more companies continue to conglomerate along the Toronto-Waterloo corridor. However, Canada faces the challenge of competing with

the US for talent, and salary is a major factor in retention. Canadian firms are uncompetitive compared to US companies.65 Vector resolves this by topping up existing researcher salaries to provide a competitive compensa-tion package. To retain talent, Vector leverages Toronto’s selling features: its liveability and lifestyle. Canada’s high quality of life, coupled with a welcoming society comparatively free from travel restrictions, gives the country a major advantage.

This applies also to university profes-sors. Vector helps to prevent technology firms from making the short-sighted mistake of hiring professors instead of graduating students. By removing a professor who imparts knowledge and grooms upcoming talent, the talent pipeline can be irreparably damaged. Vector provides top-up compensation for professors as they retain their scholarly duties, while still engaging with real-world projects.

Researchers are also attracted by the immense talent at the Vector Institute. Dr. Geoffrey Hinton, a pioneer of the world’s dominant AI technique and neural networks, is the Chief Scientific Advisor at Vector. Dr. Hinton’s exper-tise is a substantial draw for Vector, the University of Toronto, and Toronto’s AI ecosystem more broadly.

Finally, the Vector Institute has attracted corporate partners who want access to its talent pool and resulting research, but recognizes the value of cultivating it for the future. This ensures that the ecosystem thrives and rewarding research positions are being created both within Vector and the surrounding industry. While many of these firms who partner with Vector are direct compet-itors, they often have similar goals and regulatory restrictions. For example, Canada’s big banks can collaborate

on fraud, anti-money laundering, and anti-terrorist financing detection.66 Few consumers differentiate financial institutions based upon their prowess in these areas. By convening through the Vector Institute, a single AI solution can be created despite a limited competitive edge to be gained.

Investment in STEM and AI-related educationThe fuel of a thriving AI economy is talent. At the present time, Ontario graduates a surprisingly small cohort in computer science, given this skill set’s importance. In 2016, only 2.1 percent of Bachelor’s degrees, 1.9 percent of Master’s degrees, and 3.2 percent of Doctoral degrees were conferred in computer science.67

The Ontario government announced that it aims to have 1,000 annual Master’s level graduates in applied AI-related fields within five years.68 It is assumed the majority of AI-related fields are within the broader discipline of computer science, albeit making up a small portion of Master’s students. In 2016, only 413 students graduated with a Master’s degree in computer science – a broader category than AI. A drastic program expansion will be necessary to reach 1,000 annual graduates (Exhibit 4). The Vector Institute is looking to involve “universities across the province” along with current leaders in the field in order to reach this goal.69

Additionally, the province aims to “increase the number of postsecondary students graduating in the Science, Technology, Engineering and Math (STEM) disciplines by 25 percent over the next five years.”70 This increased rate of mathematically-inclined backgrounds add to the AI and technology ecosystem with both direct and supplementary skill sets (eg. marketing).


Creative Destruction LabThe University of Toronto’s Rotman School of Management is home to the Creative Destruction Lab (CDL).71 The CDL brings high tech ideas in AI and quantum machine learning together with Rotman’s business acumen. Companies at the pre-seed stage of funding are assisted by access to Rotman’s top students, and funding opportunities from local venture capital organizations.72 The infusion of top talent into innovative firms helps strengthen the high tech business ecosystem surrounding the University of Toronto.

Canada is finding success in AI

The efforts of Canadian businesses and researchers are starting to pay off. The country is being recognized on the world stage for creating products and services that are leading the field.

Recently, an 18-month-old Toronto company, 16 Bit, entered and won the Radiology Society of North America

(RSNA) pediatric bone age compe-tition.73 The competition focused on identifying the skeletal maturity of a child’s hand to determine if their bones are growing too fast or too slow. 16 Bit surpassed the accuracy of a panel of pediatric radiologists, and highlighted the potential impacts AI will have on the medical imaging field. They expect to expand their technology to assist breast cancer diagnosis through mammogram imaging.

Layer 6 is another successful AI company in Canada. They developed AI that can transform financial banking data into more personalized services for consumers. The venture was origi-nally funded by more than 30 unique companies as well as the governments of Ontario and Canada.74 As of 2018, TD Bank acquired Layer 6 and integrated its capabilities in the hopes of providing more directed services for customers.

The machine learning company, Acerta Analytics Solutions of Kitchener, Ontario developed a quality control solution

for the manufacturing industry.75 By feeding sensor data obtained at any stage in the manufacturing and oper-ating process into their AI solution, they can detect abnormalities and predict equipment failure in real time. Acerta’s AI machine learning solution is capable of identifying problems undetectable by traditional methods, saving clients time and money. They are already assisting major international car manufacturers such as Daimler (Mercedes-Benz) and Volkswagen.76

A Vancouver company called Finn.ai won the Best of Show at the Finovate conference in New York in 2017.77 Finn.ai is a personal banking assis-tant providing AI solutions to improve customer satisfaction while reducing customer care costs.78 As of 2018, they will supply Bank of Montréal with a personal chatbot to directly engage consumers.79 The chatbot will provide information through naturally flowing conversation, and continue to develop its range of skills throughout its operation.80

Source: Institute for Competitiveness & Prosperity analysis based on data from Council of Ontario Universities, Table A3 Degrees Conferred by Program.

2005 2006 2007 2008 2010 2012 2014 20152009 2011 2013 2016

Number ofgraduating

students

0

500

1,000

1,500

2,000

2,500

Master'sDegree

DoctoralDegree

Bachelor'sDegree

EXHIBIT 4 Degrees conferred in Computer Science, Ontario, 2005-2016


CHAPTER 3

HOW CAN ONTARIO AND CANADA WIN?

Ontario has an opportunity to win in the global AI market. But it must know where it can compete and how to support firms developing AI and related technology. When firms and government collaborate and operate on their respective strengths, they can realize a bright economic future that will be shared by all.

CANADA AND ONTARIO ARE FACED with the monumental task of becoming a world leader on AI development and commercialization. The two levels of government will need to work together to ensure the domestic regulatory environment can achieve this goal, and that financial investments are being made to facilitate this high value added industry.

Create a flexible but sound regulatory framework

When working properly, the economy propels companies with the best product or service offering to the forefront, while redistributing the assets of failed firms. However, government interference through regulatory barriers, rules, or even investment can act as a distraction and add costs and uncertainty to the economic environment. The well-intentioned act of picking winners and making educated bets on firms within the economy can cause unforeseen consequences that may lead to suboptimal outcomes. This is especially problematic as government is less equipped than the private sector to provide venture capital or other investment. Limited government interaction can sometimes be the best thing for an industry.

Instead, when companies are provided with a clearly defined agenda by govern-ment, their uncertainty is reduced and they can make investments with the confidence they will not be knocked off course by non-market forces.

Create a regulatory sandbox to incentivize domestic innovationMost AI applications are aligned with the wants and needs of society because only in-demand services can be profitable. Government should recognize that some existing regulations may stand in the way of a domestic AI reaching market. This is where a regulatory sandbox can help. By creating a regulatory-light framework for AI companies to experiment and grow in, before facing more stringent regulations,


a greater number of successful domestic companies should emerge. Companies acting within the sandbox would not need to contend with heavy regulatory burdens while still in their infancy and potentially unprofitable. A sandbox is not a complete removal of rules and regulations—instead, it merely facili-tates a flexible environment conducive to innovation, and may reveal areas in which regulations must change for the broader industry.

The financial industry in Canada has already set up its own regulatory sandbox. The Canadian Securities Administrators (CSA) has constructed the CSA Regulatory Sandbox to allow Fintech companies freedom to explore innovative product and service offerings.81 The sandbox is open to companies of all sizes, and involves working directly with regulators to find areas where regulations can be acceptably softened in order to facilitate innovative exploration.82

The regulations involved with tech-nology companies developing AI will likely fall under the jurisdictions of both federal and provincial governments. A single joint sandbox is preferable to the creation of separate federal and provin-cial programs. A joint effort would put most of the onus on government regulators to make the program under-standable, and simplify the process for companies largely preoccupied with development. By working together to create flexible regulations, Canada and Ontario can mould future regulations for the broader industry to be less administratively burdensome.

Balance societal protection with innovation performance through regulationThe regulatory burden companies face outside of a sandbox must adjust to fit both the current technological land-scape and the concerns of society. In general, stricter regulation limits the output and innovative capacity of an industry. This is not to say regulations are unnecessary. Some amount of regulation is certainly superior to an unregulated economic environment. Striking the balance between what is best for society and what will encourage world leading AI development is a crucial task for Canada and Ontario. Keeping this balancing act top of mind, government should continue to review and update regulations surrounding AI.

Modernize trade policy and trade agreementsTrade agreements were once documents predominantly concerned with how goods are treated as they cross borders. Agreements were used to reduce tariffs levied on incoming goods for both sides, in the hopes that expanded trade would yield mutual benefits.83 By exposing producers to a larger pool of both consumers and competitors, regions specialize in products they are good at, trade for products they are compara-tively less good at, and the overall pie is expanded. Today, trade agreements still cover these same rules but pay greater attention to other factors that have an impact on the development of AI.

Discussion surrounding intellectual property, investments, and data have grown in importance in trade agree-ments such as the recent Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP).84 This is because countries may attempt to use trade agreements or restrictions in order to advantage their domestic technology firms by placing limitations on data

Both the federal and provincial government have datasets that they could use to advantage Canadian firms in the pursuit of global AI dominance.

sharing, data use, or data movement across borders. By limiting the use of their citizens’ data to domestic firms they can gain a substantial advantage. The advantage corresponds to the size of the domestic population and the ability of firms to retain access to foreign data.

Countries may also impose implicit trade barriers that advantage their domestic firms by restricting access to certain datasets. The types of datasets a country might want to protect includes medical documents and records, tax filings, military information, law enforcement or criminal records, financial records, and real estate records. In obtaining information held in any of these sensi-tive datasets, a company could achieve a competitive advantage against other firms, especially in the domestic market where data can produce the most relevant predictions. A population’s private data could drastically improve the AI’s predictive power, increasing the value offering of the company. Both the federal and provincial government have datasets that they could use to advan-tage Canadian firms in the pursuit of global AI dominance. This is a relatively costless strategy from a trade retaliation perspective. Domestic firms were never going to be granted access to sensitive foreign data, so a retaliation would be of limited consequence.

The danger of retaliation is higher for restrictions in data localization and could lead to a race to the bottom


scenario, disadvantaging Canada due to its relatively small population. Data localization rules would limit the ability for data to travel between borders.85 Cloud computing currently handles many business operations, and this trend will likely accelerate.86 Cloud computing involves data passing from the user’s or company’s device to a server, which may not be located in the same country as where the data origi-nated. A restriction on where data can traverse would limit the ability for global firms to offer cloud computing services in another country without making a new domestic capital investment.

Data localization legislation can limit how consumer data collected in a country are utilized by multinational firms. The majority of world leading companies in AI development currently operate in multiple countries and some also operate across multiple distinct industries. For example, Alphabet operates their Google search engine along with products such as email, web browser, maps, cellphone operating system, social media, mobile payment platform, and video in many coun-tries globally.87 All data collected from these distinct and varied users can be exported back to the parent company for use in AI training and development.

Companies with access to geographi-cally and demographically dispersed data are at an advantage when devel-oping AI for commercial applications as there is more total data, it fits their consumers better, and the variation across geographies highlights rare cases allowing a company to successfully differentiate itself from the compe-tition.88 If countries restrict the flow of consumer data across borders, the advantage incumbent multinationals possess would be greatly diminished.

Orient talent and industrial policy toward the AI ecosystem

In the Artificial Intelligence ecosystem, government investment can ensure industry players have access to the factors necessary for success. Investments in education, skills, training, research entities, accelerators, and cluster or business organizations directly benefit domestic AI firms. These initiatives are best handled by a central funding and organizational institution rather than industry players.

Teach AI-related skillsAs education is a provincial file, it falls to Ontario to ensure students are graduating with the skills necessary to succeed in the future workplace. A continuous supply of graduates with hard skills such as computer science will be necessary to fulfill the needs of the economy.

Skills that do not yet exist will need to be taught by post-secondary institu-tions. Equipping people with the ability to evaluate and manage rather than create AI may also become increas-ingly useful as the industry progresses. Humans need to ensure that the “black box” algorithms of AI achieve their intended goals. Additionally, reframing these goals to remain aligned with the changing objectives of companies or society as a whole will be a valuable skill set.

Educational institutions will also be responsible for teaching skills that are complementary to (rather than substi-tuted by) AI and machine learning. The curriculum revision will be necessary to ensure that the four or more years of an individual’s time and money spent in receiving post-secondary education result in the best chance for economic success. In the medium to long term, high tech and hard skills are likely to be

among the most relevant. In 2016, 22.3 percent of Ontario’s Bachelor’s degrees and 25.4 percent of Master’s degrees were awarded in STEM programs.89 Encouragingly, the proportion of STEM Doctoral degrees is even higher, at 42.8 percent.90 It is highly probable that the current distribution among educational fields is not aligned with the industries of today, and it is almost certainly not aligned with the skill set required in the future.

Prevent the loss of talent to foreign jurisdictionsCanada and Ontario have lost some of their best talent to jurisdictions, including the US and China, willing to pay more for in-demand skill sets. A society that pours significant resources into training and educating its youth, without retaining them for commer-cialized applications will eventually stagnate, as it shifts to being an importer (purchaser) of technology rather than an exporter (supplier). Supplying high quality education is a costly endeavour in Canada, and having the most educated students leave as they reach their prime innovation and value creation potential is an unsustainable business model.

Other jurisdictions have made efforts to curtail this subsidization of foreign economies at the expense of their own. Singapore successfully reduced its brain drain and realigned costs with those benefiting from education by altering the requirements of their student loan program. Singapore now requires “student loans to be repaid if the student does not work in Singapore for a minimum number of years.”91 For example, the Public Service Commission (PSC) Scholarship is a full university ride with a four to six year domestic work bond obligation.92 A similar program could partially return the costs of education to the individual, which


would influence their decision between domestic and foreign job opportuni-ties. Firms that aim to poach graduates would have to pay outstanding educa-tion costs.

If a similar policy were enacted in Ontario it might incentivize graduates to remain in the country long enough to pay off their student loans. Once individuals obtain employment in their field, they become increasingly likely to stay as they put down roots—friends, housing, significant others, and starting a family can all anchor them to a region. By retaining individuals post-gradu-ation, Ontario could reduce the brain drain of those in the AI industry.

Improve immigration policyAs nationalistic sentiment continues to develop globally, Canada not only has the opportunity to continue positioning itself as a top tier destination for immi-grants and refugees, but also to attract additional best-in-class talent from parts of the world where individuals no longer feel welcome. Canada can offer these individuals a more inclusive home that retains strong access to the US market-place through the North American Free Trade Agreement (NAFTA).93

Canada and Ontario have the opportu-nity to utilize a series of policy levers to capture exceptionally skilled individuals who are shopping around for where they want to settle down and live their lives.

Ontario can continue negotiations with the federal government to increase the proportional size of its provincial nominee program (PNP) as recom-mended by the Premier’s Highly Skilled Workforce Expert Panel.94 In 2017, Ontario selected only 6.2 percent of its immigrants through its PNP, as opposed to 40.0 percent in the rest of the coun-try.95 Ontario is best equipped to judge the skill sets needed for its domestic

economy, so any improvement in skill matching for the AI industry will build Ontario’s competitive advantage.96

Canada can also work to improve the retention of international students who attend its universities. The federal retention rate (transition to permanent residency) of international students who graduated in 2005 to 2009 was only 16 percent after five years.97 By not retaining these individuals, Canada is missing out on a cohort with recognized credentials, strong language skills, domestic support networks, and valu-able skills in the development of AI or industries working alongside it.98

Create meaningful industrial policyGiven that Ontario did not receive funding for its AI ecosystem through the federal Innovation Supercluster Initiative, the opportunity to estab-lish a provincial level cluster remains. Provincial funding towards collabo-rative opportunities or an organizing cluster initiative could be impactful in enhancing Ontario’s position in the global AI market.99 By funneling investment through a cluster initiative, the government would not need to pick winners, but could still implement productivity-enhancing industrial policy in this relatively young industry.

Incorporate AI into business strategy

Firms will face important business decisions as the capabilities of AI evolve. First will be the choice to make AI the company’s priority, or continue business operations as before, while prioritizing traditional profit strategies. Firms must realize that pursuing AI is not a costless improvement. By making AI the priority, a firm is sacrificing other aspects of its core business in the short term in order to unlock long-term potential.100

Before pursuing AI as a core component of their business, a firm should identify the problem it aims to solve and quan-tify the additional value it can expect to capture. The firm needs to determine what “strategic dilemma” it faces, if it can be solved by reducing uncertainty, and if it has access to AI that can reduce the corresponding uncertainty to the required level.101

A firm’s strategic dilemma is an alterna-tive way of doing business that imposes a cost-benefit trade-off.102 Currently, the new way of doing business would produce a lower return on investment on an existing product or service or a negative return on investment for a new venture. The return on investment of the alternative strategy needs to have the property of being enhanced by increased predictive capacity.103 Finally, the firm has to have the capacity to create AI that makes accurate enough predictions, so that the return on investment of the new strategy exceeds the alternative.104 The identification of these three factors should be a starting point for firms in Ontario looking to enhance their operations by investing in AI.

Canada and Ontario have the opportunity to participate in the global AI ecosystem. By making the necessary investments now, Canadian companies can compete in the global market, jobs will be created domestically, and those who are displaced can find fulfilling opportunities in the new economic landscape.

1 Encyclopedia Britannica. “Artificial intelligence.” May 3, 2018, https://www.britannica.com/technology/artificial-intelligence.

2 Turing, Alan. “Computing Machinery and Intelligence.” Mind 49, no. 236 (1950): 433-60.

3 Agrawal, Ajay, Joshua Gans, and Avi Goldfarb. Forthcoming. “Introduction.” In The Economics of Artificial Intelligence: An Agenda: Introduction. Agrawal, Ajay, Joshua Gans, and Avi Goldfarb, eds. Chicago: University of Chicago Press.

4 Advisory Panel on Federal Support for Fundamental Science. “Investing in Canada’s Future: Strengthening the Foundations of Canadian Research.” 2017.

5 Note: PWC reports US$15.7 trillion, which was converted by OECD 2017 PPP to C$19.8 trillion. Source: PwC. “Sizing the Prize: What’s the Real Value of AI for Your Business and How Can You Capitalise?” 2017.

6 Frey, Carl Benedikt and Michael A Osborne. “The Future of Employment: How Susceptible Are Jobs to Computerisation?” Technological Forecasting and Social Change 114 (2017): 254-280.

7 ibid.

8 McDonald’s Corporation. “McDonald’s Unveils New Global Growth Plan.” Cision PR Newswire. March 1, 2017. https://www.prnewswire.com/news-releases/mcdonalds-unveils-new-global-growth-plan-300415634.html.

9 IBM. “10 Reasons Why AI-Powered Automated Customer Service is the Future.” https://www.ibm.com/blogs/watson/2017/10/10-reasons-ai-powered-automated-customer-service-future.

10 ibid.

11 Frey and Osborne, “The Future of Employment: How Susceptible Are Jobs to Computerisation?”

12 Encyclopedia Britannica. “John Kay.” December 12, 2017, https://www.britannica.com/biography/John-Kay; Encyclopedia Britannica. “Edmund Cartwright.” April 17, 2018. https://www.britannica.com/biography/ Edmund-Cartwright.

13 Timmins, Geoffrey. The Last Shift: The Decline of Handloom Weaving in Nineteenth-Century Lancashire. Manchester: Manchester University Press, 1993, p. 19.

14 Cowen, Tyler. Forthcoming. “Neglected Open Questions in the Economics of Artificial Intelligence.” In The Economics of Artificial Intelligence: An Agenda: Introduction. Agrawal, Ajay, Joshua Gans, and Avi Goldfarb, eds. Chicago: University of Chicago Press.

15 Pretty, Jules. “The Rapid Emergence of Genetic Modification in World Agriculture: Contested Risks and Benefits,” Environmental Conservation 28, no. 3 (2001): 248-262.

16 Varian, Hal. Forthcoming. “Artificial Intelligence, Economics, and Industrial Organization.” In The Economics of Artificial Intelligence: An Agenda: Introduction. Agrawal, Ajay, Joshua Gans, and Avi Goldfarb, eds. Chicago: University of Chicago Press.

17 ibid.

18 Somanchi, Sri Harsha. “The Mail You Want, Not the Spam You Don’t.” Official Gmail Blog. July 9, 2015. https://gmail.googleblog.com/2015/07/the-mail-you-want-not-spam-you-dont.html.

19 Kannan, Anjuli, Karol Kurach, Sujith Ravi, Tobias Kaufman, et al., “Smart Reply: Automated Response Suggestion for Email.” KDD ‘16 Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (2016): 955-64.

20 Melitz, Marc J. “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity.” Econometrica 71, no. 6 (2003): 1695-1725.

21 Angele A. Gilroy. “The Net Neutrality Debate: Access to Broadband Networks.” Congressional Research Service, 2018.

22 ibid.

23 Varian, “Artificial Intelligence, Economics, and Industrial Organization.”

24 ibid.

25 Note: Graphical Processing Units (GPUs) and Tensor Processing Units (TPUs) are the preferred hardware for developing and running artificial intelligence. AI performs its computations better on this type of hardware as opposed to the traditional central processing unit (CPU). Source: Sato, Kaz, Cliff Young, David Patterson. “An In-Depth Look at Google’s First Tensor Processing Unit (TPU).” Google Cloud Big Data and Machine Learning Blog. May 12 2017. https://cloud.google.com/blog/big-data/2017/05/an-in-depth-look-at-googles-first- tensor-processing-unit-tpu.

26 Goldfarb, Avi, and Daniel Trefler. Forthcoming. “AI and International Trade.” In The Economics of Artificial Intelligence: An Agenda: Introduction. Agrawal, Ajay, Joshua Gans, and Avi Goldfarb, eds. Chicago: University of Chicago Press.

27 ibid.


END NOTES

https://www.britannica.com/technology/artificial-intelligence

https://www.prnewswire.com/news-releases/mcdonalds-unveils-new-global-growth-plan-300415634.html

https://www.ibm.com/blogs/

https://www.britannica.com/biography/John-Kay

https://www.britannica.com/biography/

https://gmail.googleblog.com/2015/07/the-mail-you-want-not-spam-you-dont.html

https://cloud.google.com/blog/big-data/2017/05/an-in-depth-look-at-googles-first-tensor-processing-unit-tpu



28 ibid.

29 Agrawal, Ajay, Joshua Gans, and Avi Goldfarb. “How AI Will Change Strategy: A Thought Experiment.” Harvard Business Review. October 3, 2017. https://hbr.org/2017/10/how-ai-will-change-strategy-a- thought-experiment.

30 ibid.

31 Lauren Thomas. “Amazon Grabbed 4 Percent of All US Retail Sales in 2017, New Study Says.” CNBC. Jan 3, 2018. https://www.cnbc.com/2018/01/03/amazon-grabbed-4-percent-of-all-us-retail-sales-in- 2017-new-study.html.

32 US Congress, House, An Act to Protect Trade and Commerce Against Unlawful Restraints and Monopolies. HR 647, 55th Cong., 1st session, introduced in House July 2, 1890, http://legisworks.org/sal/26/stats/ STATUTE-26-Pg209a.pdf.

33 “2012 Founders’ Letter.” Alphabet Investor Relations. 2012. https://abc.xyz/investor/founders-letters/2012/.

34 Duhigg, Charles. “The Case Against Google.” New York Times. February 20, 2018. https://www.nytimes.com/2018/02/20/magazine/the-case-against-google.html.

35 Note: Bloomberg reports €2.4 billion, which is converted by OECD 2017 exchange rates to C$3.5 billion; White, Aoife. “Google’s Record Fine of $2.8 Billion Was a ‘Deterrent,’ EU Says.” Bloomberg. December 18, 2017. https://www.bloomberg.com/news/articles/2017-12-18/google-s-record-fine-of-2-8-billion-was-a-deterrent-eu-says.

36 Bériault, Yves, and Oliver Borgers. “Overview of Canadian Antitrust Law.” McCarthy Tétrault LLP, 2005.

37 Goldfarb and Trefler, “AI and International Trade.”

38 ibid.

39 ibid.

40 ibid.

41 Athey, Susan, Christian Catalini, and Catherine Tucker. “The Digital Privacy Paradox: Small Money, Small Costs, Small Talk.” MIT Sloan Research Paper no. 5196-17, 2018.

42 Tucker, Catherine. Forthcoming. “Privacy, Algorithms and Artificial Intelligence.” In The Economics of Artificial Intelligence: An Agenda: Introduction. Agrawal, Ajay, Joshua Gans, and Avi Goldfarb, eds. Chicago: University of Chicago Press.

43 ibid.

44 ibid.

45 ibid.

46 ibid.

47 ibid.

48 Jin, Ginger Zhe. “Artificial Intelligence and Consumer Privacy.” NBER Working Paper no. 24253, 2018.

49 Goldfarb, Avi, and Catherine Tucker. “Privacy and Innovation.” NBER Working Paper no. 17124, 2011.

50 “GDPR Portal: Site Overview.” EU GDPR. 2017. https://www.eugdpr.org/.

51 Note: GDPR reports €20 million, which is converted by OECD 2017 exchange rates to C$29 million. “GDPR Key Changes.” EU GDPR. 2017.

52 Note: Tertiary education consists of certificates or diplomas from college, as well as certificates below a Bachelor’s degree, and Bachelor’s, masters and doctoral degrees.

53 CIFAR. “Pan-Canadian Artificial Intelligence Strategy Overview.” March 30, 2017. https://www.cifar.ca/assets/pan-canadian-artificial-intelligence-strategy-overview/.

54 ibid.

55 Government of Canada. “Innovation Superclusters Initiative.” February 21, 2018. https://www.canada.ca/ en/innovation-science-economic-development/programs/small-business-financing-growth/ innovation-superclusters.html.

56 Institute for Competitiveness & Prosperity. “Clusters in Ontario: Creating an Ecosystem for Prosperity.” 2016.

57 Porter, Michael E. Competitive Advantage of Nations. New York: Free Press, 1990.

58 Institute for Competitiveness & Prosperity, “Clusters in Ontario: Creating an Ecosystem for Prosperity.”

59 Government of Canada. “Canada’s New Superclusters.” February 14, 2018. https://www.ic.gc.ca/eic/site/093.nsf/eng/00008.html.


https://hbr.org/2017/10/how-ai-will-change-strategy-a-thought-experiment



https://www.cnbc.com/2018/01/03/amazon-grabbed-4-percent-of-all-us-retail-sales-in-2017-new-study.html

https://www.cnbc.com/2018/01/03/amazon-grabbed-4-percent-of-all-us-retail-sales-in-2017-new-study.html

http://legisworks.org/sal/26/stats/

https://abc.xyz/investor/founders-letters/2012/

https://www.nytimes.com/2018/02/20/magazine/the-case-against-google.html

https://www.bloomberg.com/news/articles/2017-12-18/google-s-record-fine-of-2-8-billion-was-a-deterrent-eu-says



https://www.eugdpr.org/

https://www.cifar.ca/assets/

https://www.canada.ca/

https://www.ic.gc.ca/eic/site/093.nsf/eng/00008.html

60 idem. “AI-Powered Supply Chains Supercluster (SCALE.AI).” Government of Canada. February 15, 2018. https://www.ic.gc.ca/eic/site/093.nsf/eng/00009.html.

61 ibid.

62 Vector Institute. “Vector Institute for Artificial Intelligence.” https://vectorinstitute.ai/.

63 idem. “Opportunities.” https://vectorinstitute.ai/#opportunities.

64 idem. “Driving Excellence in Machine Learning and Deep Learning.” 2018. https://vectorinstitute.ai/ wp-content/uploads/2018/04/vector-institute-information-deck-20180410.pdf.

65 Onstad, Katrina. “Mr. Robot.” Toronto Life. January 29, 2018. https://torontolife.com/tech/ai-superstars- google-facebook-apple-studied-guy/.

66 Financial Transactions and Reports Analysis Centre of Canada. “Canada’s Anti-Money Laundering and Anti-Terrorist Financing Regime.” May 30, 2015. http://www.fintrac-canafe.gc.ca/fintrac-canafe/ antimltf-eng.asp.

67 Institute for Competitiveness & Prosperity analysis based on data from Council on Ontario Universities, Table A3, Degrees Conferred by Program.

68 Ministry of Economic Development and Growth. “Ontario Boosting the Number of Graduates in Science, Tech, Engineering, Mathematics and Artificial Intelligence.” Ontario Newsroom. October 18, 2017. https://news.ontario.ca/medg/en/2017/10/ontario-boosting-the-number-of-graduates-in-science-tech- engineering-mathematics-and-artificial-inte.html.

69 Sorensen, Chris. “Ontario to boost number of STEM grads, with a focus on AI, to attract Amazon and others.” University of Toronto. October 18, 2017. https://www.utoronto.ca/news/ontario-boost-number-stem-grads- focus-ai-attract-amazon-and-others.

70 Ministry of Economic Development and Growth. “Ontario Boosting the Number of Graduates in Science, Tech, Engineering, Mathematics and Artificial Intelligence.”

71 Creative Destruction Lab. “Creative Destruction Lab Toronto.” https://www.creativedestructionlab.com/ locations/toronto/.

72 idem. “Build Something Massive.” https://www.creativedestructionlab.com/.

73 “Canadian Company Wins Competition at RSNA.” Canadian Healthcare Technology. February 21, 2018. http://www.canhealth.com/blog/canadian-company-wins-competition-at-rsna/.

74 TD Bank Group. “TD Bank Group Acquires Artificial Intelligence Innovator Layer 6.” TD Bank Group Newsroom. January 9, 2018. https://td.mediaroom.com/2018-01-09-TD-Bank-Group-acquires-artificial-intelligence- innovator-Layer-6.

75 Acerta. “Detect and Identify Hidden Issues, Faster.” http://acerta.ca/.

76 Ontario Investment Office. “ACERTA: Using AI to pave the way for self-driving cars.” March 5, 2018. https://www.investinontario.com/success-stories/acerta-using-ai-pave-way-self-driving-cars.

77 Penn, David. “FinovateFall 2017 Best of Show Winners Announced.” Finovate. September 12, 2017. http://finovate.com/finovatefall-2017-best-show-winners-announced/.

78 Finn AI. “Deepen Your Client Relationships with Personal Banking Software from Finn AI.” http://finn.ai/

79 “Finn AI Launches Personal Banking Chatbot with BMO.” Accesswire. March 15, 2017. https://www.accesswire.com/viewarticle.aspx?id=493170.

80 ibid.

81 Canadian Securities Administrators. “CSA Regulatory Sandbox.” https://www.securities-administrators.ca/industry_resources.aspx?id=1588.

82 ibid.

83 Ricardo, David. Principles of Political Economy and Taxation. London: John Murray, Albermarle-Street, 1817.

84 Government of Canada. “Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP).” March 3, 2018. http://international.gc.ca/trade-commerce/trade-agreements-accords-commerciaux/agr-acc/cptpp-ptpgp/text-texte/index.aspx?lang=eng.

85 Schwartz, Paul. “Legal Access to the Global Cloud.” UC Berkeley Public Law Research Paper no. 118, 2017.

86 ibid.

87 “Alphabet Inc (GOOG.O).” Reuters. 2018. https://www.reuters.com/finance/stocks/company-profile/GOOG.O.



https://www.ic.gc.ca/eic/site/093.nsf/eng/00009.html

https://vectorinstitute.ai/

https://vectorinstitute.ai/#opportunities

https://vectorinstitute.ai/

https://torontolife.com/tech/ai-superstars-google-facebook-apple-studied-guy/



http://www.fintrac-canafe.gc.ca/fintrac-canafe/

https://news.ontario.ca/medg/en/2017/10/ontario-boosting-the-number-of-graduates-in-science-tech-engineering-mathematics-and-artificial-inte.html



https://www.utoronto.ca/news/ontario-boost-number-stem-grads-focus-ai-attract-amazon-and-others



https://www.creativedestructionlab.com/

https://www.creativedestructionlab.com/

http://www.canhealth.com/blog/canadian-company-wins-competition-at-rsna/

https://td.mediaroom.com/2018-01-09-TD-Bank-Group-acquires-artificial-intelligence-innovator-Layer-6



http://acerta.ca/

https://www.investinontario.com/success-stories/acerta-using-ai-pave-way-self-driving-cars

http://finovate.com/finovatefall-2017-best-show-winners-announced/

http://finn.ai/

https://www.accesswire.com/viewarticle.aspx?id=493170

https://www.securities-administrators.ca/

http://international.gc.ca/trade-commerce/trade-agreements-accords-commerciaux/agr-acc/

https://www.reuters.com/finance/stocks/company-profile/GOOG.O

89 Note: STEM definition from Statistics Canada where matched to the following programs: Agriculture & Biological Science, Architecture, Computer Science, Engineering, Food Science, Math, and Physical Science; “Variant of CIP 2011- STEM groupings.” Statistics Canada. April 30, 2013; Council of Ontario Universities. “A3 Degrees Conferred by Program.” 2018.

90 Council of Ontario Universities, “A3 Degrees Conferred by Program.”


92 Public Service Commission Singapore. “Public Service Commission (PSC) Scholarship.” July 17, 2017. https://www.psc.gov.sg/Scholarships/public-sector-scholarships/browse-by-scholarship/public-service-commission-psc-scholarship-PSC.

93 Government of Canada. “North American Free Trade Agreement (NAFTA).” November 18, 2017. http://www.international.gc.ca/trade-commerce/trade-agreements-accords-commerciaux/agr-acc/ nafta-alena/fta-ale/background-contexte.aspx?lang=eng.

94 The Premier’s Highly Skilled Workforce Expert Panel. “Building the Workforce of Tomorrow: A Shared Responsibility.” Government of Ontario, 2016.

95 Institute for Competitiveness & Prosperity based on data from Immigration, Refugees and Citizenship Canada, Open Government Portal, Canada, Permanent Residents – Monthly IRCC Updates, 2017.

96 Institute for Competitiveness & Prosperity. “Immigration in Ontario: Achieving Best Outcomes for Newcomers and the Economy.” 2017.

97 Lu, Yuqian, and Feng Hou. “International Students who Become Permanent Residents in Canada.” Statistics Canada. 2015.

98 Institute for Competitiveness & Prosperity. “Immigration in Ontario: Achieving Best Outcomes for Newcomers and the Economy.”

99 idem, “Clusters in Ontario: Creating an Ecosystem for Prosperity.”

100 Agrawal, Ajay, Avi Goldfarb, and Joshua Gans. “Prediction Machines: The Simple Economics of Artificial Intelligence,” Harvard Business Review (2018).

101 ibid.

102 ibid.

103 ibid.

104 ibid.

ENDNOTES TO SIDEBARS

a Acemoglu, Daron and Pascual Restrepo. “Artificial Intelligence, Automation and Work.” NBER Working Paper no. 24196, 2018.

b Cobb, Charles W. and Paul H. Douglas. “A Theory of Production.” The American Economic Review 18, no. 1 (1928): 139-165.

c Statistics Canada. “Gross operating surplus; Canada; Seasonally adjusted at annual rates (v62295552).” February 13, 2018.

d World Bank. “Population, total.” 2017; World Bank. “Individuals using the Internet (% of population).” 2016.

e State Council. “New generation artificial intelligence development planning notice.” July 8, 2017. http://www.gov.cn/zhengce/content/2017-07/20/content_5211996.htm.

f Goldfarb and Trefler, “AI and International Trade.”

g Greatfire.org. “Greatfire.org: http://google.com.” https://en.greatfire.org/google.com.

h Note: PWCCN reports US $7 billion, which is converted by OECD 2017 exchange rates to C$9 billion; PwC “China Entertainment and Media Outlook 2016-2020.” 2016.

i Goldfarb and Trefler, “AI and International Trade.”


https://www.psc.gov.sg/Scholarships/public-sector-scholarships/browse-by-scholarship/public-service-commission-psc-scholarship-PSC



http://www.international.gc.ca/trade-commerce/trade-agreements-accords-commerciaux/agr-acc/

http://www.gov.cn/zhengce/content/2017-07/20/content_5211996.htm

http://google.com.%E2%80%9D

https://en.greatfire.org/google.com


Annual ReportsFIRST ANNUAL REPORT – Closing the prosperity gap, November 2002SECOND ANNUAL REPORT – Investing for prosperity, November 2003THIRD ANNUAL REPORT – Realizing our prosperity potential, November 2004FOURTH ANNUAL REPORT – Rebalancing priorities for prosperity, November 2005FIFTH ANNUAL REPORT – Agenda for our prosperity, November 2006SIXTH ANNUAL REPORT – Path to the 2020 prosperity agenda, November 2007SEVENTH ANNUAL REPORT – Leaning into the wind, November 2008EIGHTH ANNUAL REPORT – Navigating through the recovery, November 2009NINTH ANNUAL REPORT – Today’s innovation, tomorrow’s prosperity, November 2010TENTH ANNUAL REPORT – Prospects for Ontario’s prosperity, November 2011ELEVENTH ANNUAL REPORT – A push for growth: The time is now, November 2012TWELFTH ANNUAL REPORT – Course correction: Charting a new road map for Ontario, November 2013THIRTEENTH ANNUAL REPORT – Finding its own way: Ontario needs to take a new tack, November 2014FOURTEENTH ANNUAL REPORT – Disruptions ahead: The making of a dynamic and resilient Ontario economy, November 2015FIFTEENTH ANNUAL REPORT – Collaborating for growth: Opportunities for Ontario, December 2016SIXTEENTH ANNUAL REPORT – Strength in numbers: Targeting labour force participation to improve prosperity in Ontario, Dec. 2017

Working PapersWORKING PAPER 1 – A View of Ontario: Ontario’s Clusters of Innovation, April 2002WORKING PAPER 2 – Measuring Ontario’s Prosperity: Developing an Economic Indicator System, August 2002WORKING PAPER 3 – Missing opportunities: Ontario’s urban prosperity gap, June 2003WORKING PAPER 4 – Striking similarities: Attitudes and Ontario’s prosperity gap, September 2003WORKING PAPER 5 – Strengthening structures: Upgrading specialized support and competitive pressure, July 2004WORKING PAPER 6 – Reinventing innovation and commercialization policy in Ontario, October 2004WORKING PAPER 7 – Taxing smarter for prosperity, March 2005WORKING PAPER 8 – Fixing fiscal federalism, October 2005WORKING PAPER 9 – Time on the job: Intensity and Ontario’s prosperity gap, September 2006WORKING PAPER 10 – Prosperity, inequality and poverty, September 2007WORKING PAPER 11 – Flourishing in the global competitiveness game, September 2008WORKING PAPER 12 – Management matters, March 2009WORKING PAPER 13 – Management matters in retail, March 2010WORKING PAPER 14 – Trade, innovation, and prosperity, September 2010WORKING PAPER 15 – Small business, entrepreneurship, and innovation, February 2012WORKING PAPER 16 – Making sense of public dollars: Ontario government revenue, spending, and debt, May 2013WORKING PAPER 17 – Untapped potential: Creating a better future for service workers, October 2013WORKING PAPER 18 – Taxing for growth: A close look at tax policy in Ontario, October 2013WORKING PAPER 19 – The realities of Ontario’s public sector compensation, February 2014WORKING PAPER 20 – Building better health care: Policy opportunities for Ontario, April 2014WORKING PAPER 21 – Open for business: Strategies for improving Ontario’s business attractiveness, May 2015WORKING PAPER 22 – Better foundations: The returns on infrastructure investment in Ontario, September 2015WORKING PAPER 23 – A place to grow: Scaling up Ontario’s firms, January 2016WORKING PAPER 24 – Licence to innovate: How government can reward risk, February 2016WORKING PAPER 24.5 – Licence to innovate revisited: How government can reward risk, June 2016WORKING PAPER 25 – Toward a low-carbon economy: The costs and benefits of cap-and-trade, April 2016WORKING PAPER 26 – Clusters in Ontario: Creating an ecosystem for prosperity, June 2016WORKING PAPER 27 – Looking beyond GDP: Measuring prosperity in Ontario, October 2016WORKING PAPER 28 – Immigration in Ontario: Achieving best outcomes for newcomers and the economy, June 2017WORKING PAPER 29 – The labour market shift: Training a highly skilled and resilient workforce in Ontario, September 2017WORKING PAPER 30 – The future is not destiny: CEO perspectives on realizing Ontario’s potential, September 2017WORKING PAPER 31 – The final leg: How Ontario can win the innovation race, April 2018

PREVIOUS PUBLICATIONS

The Institute for Competitiveness & Prosperity105 St. George StreetSuite 9000Toronto, Ontario M5S 3E6Telephone 416 946 7300Fax 416 946 7606

Should you wish to obtain a copy of one of the previous publications, please visit www.competeprosper.ca for an electronic version.

To learn more about the Institute please visit us at:www.competeprosper.ca

Should you have any questions or comments, you may reach us through the website or at the following address:

EXECUTIVE DIRECTORJamison Steeve416 946 [email protected]

RESEARCH DIRECTORDorinda So416 946 [email protected]

RESEARCH SUPERVISORChristopher Mack (Project Lead)416 978 [email protected]

POLICY ANALYSTSJacob Greenspon416 946 [email protected] Weseem Ahmed 416 978 [email protected]

Margaret Campbell416 978 [email protected]

DESIGN Hambly & Woolley Inc. www.hamblywoolley.com Illustration: ©2018 Mike Ellis

HOW TO CONTACT US

http://www.competeprosper.ca

http://www.competeprosper.ca

mailto:[email protected]






http://www.hamblywoolley.com

ISBN: 978-1-927065-29-7

wp 32 from prediction to reality · 2 institute for competitiveness & prosperity exhibits...

Documents