The Mark Carney government is making AI a cornerstone of its plan to make Canada’s government more cost-efficient and productive.
By 2029, the government plans to cut 28,000 federal service positions, aiming to save $60 billion over the next several years, supplementing these positions with AI.
Replacing jobs with AI sounds like a way for the Carney government to save money; less money would be spent on salaries for work that the government could get for free. Unfortunately, AI is unable to do the complex work required, making tasks more expensive, not less.
According to Policy Options, cost-cutting across multiple sectors is not the way to go. Instead, professors Natasha Tusikov from York University and Blayne Haggart from Brock University suggest that it be used “after careful, case-by-case deliberation that pays close attention to how this technology interacts with the people using the tech in question.”
The main reason for this stems from how AI works, specifically “generative AI” and “discriminative AI.”
Most people know generative AI. It’s in almost every piece of hardware and software put into the smart tech ecosystem in the last three years. The issue is, what it learns from and what it spits out is from what it is fed, not new, in-the-moment information.
“Discriminative AI” is used for decision-making through pattern recognition in data. Once again, it is only as useful and reliable as the data that people put into it.
However, Kémy Adé, the research fellow and guest teacher in question, never listed those responsibilities. Those responsibilities were generated entirely by the AI program the IRCC was using to vet applicants.
One other reason that over-reliance on AI could be detrimental is deskilling, which could lead to fewer experts who could find and catch these errors in the AI processes.
While Carney’s government wants to use AI, it should consider the risks before the cost to fix issues is too high to ignore.
Sources: Policy Options, Toronto Star
