The Fourth Industrial Revolution is fundamentally changing the way we live, work and relate to one another (World Economic Forum, 2020). A key driver in this revolution is Artificial Intelligence (AI), which is intelligence demonstrated by machines. Machine learning (ML), a sub-discipline of AI, is contributing to an increased scope and scale of the deployment of AI across various aspects of life, and is expected to transform numerous occupations and industries.
In 2020, MaRS Data Catalyst undertook an independent analysis to understand how machine learning could impact Canadian occupations and industries. More specifically, we sought to understand the potential for Canadian occupations, as made up of their composite tasks, to be substituted by machine learning. This analysis was based on research by Brynjolfsson and Mitchell (2017), who developed a rubric to evaluate the potential for machine learning to replace specific occupational tasks in the U.S. Brynjolfsson et al. (2018) applied this rubric to build measures of “suitability for machine learning” (SML) for occupational tasks found in O*NET, an American occupation information database.
To understand the potential impact of ML on Canadian occupations, MaRS Data Catalyst mapped SML scores from American occupations in O*NET to Canadian occupations in the National Occupation Classification (NOC) using an O*NET-NOC crosswalk. Our main findings suggest that:
The above indicates that (1) ML will impact occupations relatively equally, regardless of income and that those in smaller metropolitan areas are likely to experience greater impact, however this relationship bears further exploration, and that (2) certain industries are at risk of greater exposure based on their composition of occupations and the nature of the tasks required by those occupations. A deeper look at the sectoral impacts illustrates the potential for machine learning to affect occupations regardless of the level of education attained.
Recommendations for further research are to apply Brynjolfsson and Mitchell’s rubric to the “main duties” found in the NOC. Doing so would address limitations inherent in the crosswalk and would allow for a direct measure of the SML of Canadian occupations. Further, the expansion of this rubric to assess the legal, social and organizational aspects of ML applicability and adoption in addition to other forms of automation, such as robotics, will allow for a more comprehensive understanding of impact of the Fourth Industrial Revolution on Canadian occupations, industries and the economy.
The dynamic visualization below displays normalized SML measures for Canadian occupations. Users can interact with the visualization to understand the relative impact of machine learning across occupations.
This interactive data visualization was created by FeiFei Han, who is a previous Data Analyst Intern with MaRS Data Catalyst.
 World Economic Forum. (2020). Fourth Industrial Revolution. Retrieved from https://www.weforum.org/focus/fourth-industrial-revolution#:~:text=Fourth%20Industrial%20Revolution,second%20and%20third%20industrial%20revolutions.
 Brynjolfsson, E., & Mitchell, T. (2017). What can machine learning do? Workforce implications. Science Magazine, Vol. 358, Issue 6370, pp. 1530-1534. Retrieved from https://www.cs.cmu.edu/~tom/pubs/Science_WorkforceDec2017.pdf
 Brynjolfsson, E., Mitchell, T., & Rock, D. (2018). What can machines learn, and what does it mean for occupations and the economy? American Economic Association Papers and Proceedings, Vol. 108, pp. 43-47. Retrieved from https://www.aeaweb.org/articles?id=10.1257/pandp.20181019
 Their main findings suggest that (i) ML will affect different occupations than those affected in prior automation waves, (ii) most occupations involve some tasks that are suitable for ML, (iii) few occupations are fully automatable using ML and (iv) realizing the potential for ML requires a redesign of the job task content.
 This crosswalk is a table that maps the relationship between the U.S. occupations in the O*NET taxonomy and Canadian occupations in the National Occupation Classification.
 We note that these findings differ from Brynjolfsson et al. (2019) whose research suggests that lower wage occupations will be disproportionately affected as well as those employed in the retailing and transportation industries.
 SML measures were produced by Brynjolfsson, Mitchell and Rock (2018) for U.S. occupations in O*NET and were mapped to NOCs using a crosswalk.
 SML measures were min-max normalized to rescale the range of measures to [0-1], inclusive.
 Employment population is from a custom Statistics Canada table of employment statistics from the 2016 Canadian Census.