AI Will Benefit Workers. Government Must Speed Up Adoption

Despite all the hype and fear-mongering surrounding the deployment of AI, there is little evidence it is having a perceptible influence on Canada’s economy at the macro level. This is not because AI is too recent an innovation for its effects to be felt. For years, proponents of AI predicted that by 2023, it would have a discernible impact on our labour market, with measurable job losses, surging productivity, a widening of income inequality, and a shrinking middle class. 

The State of Canada’s Labour Market Today

The reality of Canada’s economy contradicts these forecasts of how and when AI will impact our society. Today’s labour market is characterized by near-record low unemployment, rising incomes for white-collar, middle-class workers, and slumping productivity. In particular, Canada’s track record on labour productivity coming out of the pandemic has been dreadful, with eight declines in nine quarters for a cumulative loss of nearly 6%. Such a downward spiral threatens our standard of living, our ability to contain inflation, and our capacity to service the enormous debts all sectors of the economy assumed during the pandemic.

Fears about job loss due to automation have always been overstated in the public’s mind. Back in 1989, Statistics Canada reported a majority of Canadians felt computers and automation would result in job losses. However, another study found that in reality, “firms that invested in robots… employ more, not fewer workers,” because robots were used more to improve product and service quality than to lower labour costs.

These findings by Statistics Canada are a reminder that technology’s impact is not as predictable or as negative for jobs as many presume. Recent employment trends in publishing and banking are good examples. While social media and the Internet led to huge job losses at newspapers, this was offset by gains elsewhere in the publishing industry, notably for online news sites such as BuzzFeed and The Huffington Post. Similarly, the widespread adoption of Automated Teller Machines (ATMs) appeared to threaten the need for human tellers. In fact, their employment expanded as bank branches proliferated and tellers were retrained to focus on providing financial advice rather than just executing transactions for clients. 

Debunking the Luddite Fallacy

One reason for the consistently erroneous predictions of technology’s impact is the so-called “Luddite fallacy” that technological innovation costs jobs. The truth is that productivity gains from innovation always have generated more incomes and jobs over the last two centuries. This reflects how jobs destroyed on farms and factories were outweighed by workers upgrading their skills and moving to new occupations where demand was expanding as a result of the increased incomes generated by higher productivity.

Another source of the poor forecasting record of AI’s impact is that it is not well understood, even by people working in the field. According to leading AI researcher Michael Jordan, “People are getting confused about the meaning of AI in discussions of technology trends — that there is some kind of intelligent thought in computers that is responsible for the progress and which is competing with humans. We don’t have that, but people are talking like we do.” 

The stalled roll-out of driverless vehicles is a good example of AI’s limitations. Human drivers were supposed to be redundant long before 2023, a significant threat to truck drivers who are the largest male occupational group in North America. Instead, the deployment of driverless vehicles has fizzled in the face of what seem to be insurmountable problems. These challenges include the necessity of human truck drivers to provide security, legal problems about who is responsible in the event of road accidents, the technical problem of driving in difficult conditions such as rain and snow, and ethical challenges about how to program vehicle instructions when forced to choose between actions resulting in serious harm to one person or another.

The Challenge of Predicting AI’s Impact

The notion that AI will replace a wide range of workers is reminiscent of the hype and exaggeration surrounding other technologies. These include driverless vehicles, cryptocurrencies, massive open online courses (MOOCs), and Google glasses. Conversely, some technologies have surprised pundits with their widespread adoption, such as the Internet, social media, online shopping, and search engines. 

Besides the difficulty of predicting technology, another reason for the slow take-up of AI is the extended time organizations need to adapt to new technologies. The delay between the arrival of a technology and its full exploitation reflects the time needed to discover how best to use it and to rearrange the world accordingly. It took decades to develop the full potential of innovations such as electricity. A prolonged period of adaption seems to be inevitable for AI as employees have to be trained, business models overhauled, and technologies scaled up. A recent survey of firms in North America found one-third of small businesses had no current plans to use AI because replacing outdated systems can be “costly, complicated, and painful.”  

While AI’s threats are exaggerated, its benefits are considerable for improving worker productivity and maximizing human capabilities. Research has shown conclusively that AI boosts productivity when deployed to replace routine tasks or used in collaboration with human skills. The Economist summarized how academic studies show an increase of three percentage points in the labour productivity of firms adopting AI. 

AI: A Tool for Humans, Not a Substitute for Humans

AI is best viewed as a tool that complements human capabilities rather than as a substitute, because of the inherent limitations of the technology. Scientist and policy analyst Vaclav Smil included a case study of AI in his recent book Invention and Innovation: A Brief History of Hype and Failure. His conclusion was that AI helps people become more productive but is “nowhere near advanced enough to start replacing our brains in reasoning, complex understanding of the real world, and social interactions.” 

Besides exaggerating AI’s capabilities, technophiles underestimate the capabilities of humans, as observed by Geoff Colvin in his book Humans Are Underrated. One reason human capacities are undervalued is that for most of history, human labour has not been put to its best use. For eons, work has meant back-breaking labour on farms, then dangerous and dirty conditions in factories, followed by mindless and repetitive paper-pushing in offices. None of these jobs exploited the fact that “our big brains are there primarily to deal with social matters, not to… cogitate about the second law of thermodynamics,” in the words of neuroscientist Michael Gazzaniga.

AI maximizes human potential by shifting work to areas we thrive on, notably professional and technical services that require judgement, empathy, and the ability to communicate, as well as nonroutine tasks ranging from therapy to the arts. Employment is growing rapidly in both these sectors, allowing humans to evolve from supplying animal muscle power or functioning as mindless drones in factory and office jobs to fully exploiting our capabilities for knowledge, judgement, and empathy.

Why Society Should Embrace AI

Governments and Canadian society should not resist the adoption of AI. It is quite the contrary. Rampant labour shortages, an ageing society, chronic weak productivity growth, and our flagging competitiveness in the global marketplace all point to the need to speed up the productivity-enhancing benefits of AI. To encourage its adoption, Canadian governments should not put unnecessary restraints on the evolution and the adoption of AI but instead work to allay the inevitable unease surrounding any new technology. 

However, governments in Canada should resist calls to adopt expensive new universal social programs, such as a Guaranteed Annual Income, in anticipation of mass unemployment or severe increases in inequality that likely will never materialize. As noted, past experiences clearly show the near-impossibility of forecasting the impact and evolution of new technologies. The federal government’s recent experiment with the Canada Emergency Response Benefit (CERB) during the COVID-19 pandemic demonstrated its impact on the labour market was far more disruptive than AI is likely to be. This is because the CERB severed the connection between employers and millions of workers in many service industries, leading to widespread labour shortages and helping to fuel higher inflation as the economy recovered. New social programs should only be introduced when and if AI leads to growing unemployment on a societal level. 

Canada should carefully coordinate the regulation of AI with other governments, learning from its mistakes of not coordinating policies on climate change and the taxation of incomes. Overly restrictive regulations will push AI development and adoption to migrate to other countries, increasing Canada’s already large competitive disadvantage.