While Artificial Intelligence (AI) is constantly in the news, the average Canadian’s interaction with the technology may be limited to short Google and chatbot summaries of those very news stories. This public-facing generative AI, while omnipresent, belies where the private sector’s next wave of growth could come from, and where much of its long-term value may lie once we all tire of producing images of politicians we don’t like in compromising positions.
There are an estimated 1,500 AI companies in Canada, compared to 660 in 2020; a 44 percent increase in four years. These run the gamut from tiny startups to the Toronto-based Cohere, a flagship provider of chatbots and other corporate services valued at approximately $7.6 billion (despite its ongoing legal woes). Many of these companies offer products meant to streamline office bureaucracy, but Canada’s manufacturers and resource extractors are also being transformed by algorithms built to spot problems in the field or on the production line at a speed and scale impossible for human employees.
“The AI landscape has evolved tremendously,” Iman Zareian, a product manager for Oakville, Ontario’s Saiwa, tells The Hub. “When we launched Saiwa [in 2021], most of our early efforts focused on building foundational technology. AI adoption was still relatively cautious in many traditional sectors, like agriculture or manufacturing. But over the past few years, we’ve seen a shift—companies are now much more open to integrating AI.”
Putting algorithms to work in the field
Saiwa offers image restoration and enhancement tools, as well as “AI-powered agriculture and environmental services,” which include estimating crop yields and spotting weeds at a digital glance. In a project for Ducks Unlimited, Saiwa’s AI crunched footage from aerial drones to spot the invasive European water chestnut in Ontario waterways, freeing biologists from dozens of hours of grunt work.
According to Statistics Canada, 6.1 percent of Canadian businesses have now integrated AI in some capacity, although industries like oil and gas (1.6 percent) and agriculture (0.7 percent) lag well behind “information and cultural industries” and “professional, scientific and technical services,” which make extensive use of customer service chatbots and image analysis.
But more industrial use cases are emerging. David Chan of the Alberta Machine Intelligence Institute rattled off several to The Hub.
“Cameras on production sites can monitor for fugitive emissions, to make sure things like methane aren’t leaking into the atmosphere without anyone’s knowledge,” Chan says. “There are also secondary benefits. You already have these cameras collecting data. You can use that footage to see who’s going in and out of your workplace. Are they wearing a hard hat? Is a vehicle authorised to be on site?”
Cameras linked to an artificial intelligence system monitor cattle in a pen on the Perepelkin family farm near Leslieville, Alberta, on Thursday, November 30, 2023. Jeff McIntosh/The Canadian Press.
As another example, Chan explained that AI could optimize maintenance by using data to predict when an assembly line machine might break down, then intervening before that expensive failure comes to pass and ensuring maintenance workers are using their time on the most important tasks.
“It opens the doors to doing things that aren’t possible when you’re relying on humans,” Chan says. “There are applications for going into hazardous environments. If you can use a drone with a camera [with AI technology], it reduces the risk a human would have to take, assuming you could even send a human in. You can start to get insights into environments where that previously wouldn’t have been possible if you’re dealing with chemicals or high temperatures.”
Chan, who helps businesses implement AI software, agrees that the field has grown rapidly. “Companies are collecting lots of data and want to leverage it. They think they’re sitting on a gold mine, but how do they actually make use of it?”
Nonetheless, there are concerns that all these uses could produce pink slips. New technologies naturally lead to fears of workers being made redundant. If you have AI monitoring your safety regulations or warehouse inventory, what do you do with the employees who were once responsible for that?
“It’s similar to other technologies in that whenever there’s advancement, there is concern in how it’s going to shift the workforce,” Chan describes. “AI is not really that different. It’s going to affect the labour force, but often in positive ways. It can help free up some of your capacity. You can focus on other things and step in only when there’s something ambiguous that needs the critical thinking of humans.”
So far, at least, the statistics appear to back Chan up. Self-reporting from employers isn’t an ideal source, but Statistics Canada found that only 12.8 percent of businesses flagged “automating tasks to replace employees” as AI’s greatest value, well behind “increasing automation without cutting jobs or hours” at 46.1 percent.
Only time will tell whether those intentions hold up, but Chan has noted a “shift” in how industries choose to implement AI tech.
“The use cases that we’re seeing are different than what I saw five years ago,” he says. “We get a lot of interest from businesses where everyone at the organization is talking about AI, whereas previously it might have been individual champions. There’s more top-down buy-in; people are hearing it from their leadership.”
Making AI approachable
As our understanding of AI technology evolves, there’s a growing recognition of the distinction between generative AI—such as the language models powering chatbots and cheating students—and less flashy machine learning models. The former is designed to churn out new data, useful or otherwise, while the latter analyses existing data to improve at a specific task. If you feed a machine algorithm on a farm thousands of images of an invasive insect, for example, it can “learn” how to spot the troublesome bug in camera footage.
“Both are at play,” Chan says. “Generative AI is really what’s captured our imaginations. It’s so ubiquitous that it’s entered every single conversation. I talk to my kids about ChatGPT. It has use cases in industrial settings, companies are using it to make internal policies and documentation more accessible to their staff. But that represents one bucket of use cases. There’s a whole other bucket where non-generative can have a huge impact.”
To Zareian, it’s important to move beyond buzzwords and explain AI using language that describes its outcomes. “AI can often seem like a black box to people who don’t work with it directly,” he explains. “Our platform helps farmers automatically detect weeds or count crops. Machine learning is simply training a system with many examples so it can start making accurate predictions with new data. By focusing on real-world outcomes, we make the tech approachable and useful.”
Ideally, these platforms can scale up to a significant degree. The B.C.-based Semios—now rebranded to Almanac—which uses machine learning to help farms monitor weather, crop yields, insect populations, and other important information, employs 500,000 sensors to cover over 150,000 orchard acres. Collectively, these sensors gather roughly hundreds of millions of data points daily, which are then synthesized and crunched down into useful conclusions at a rate that would be impossible for human analysts.
Pistachio trees, for example, need to experience a certain degree of “chill” during winter to prepare themselves for spring. Semios sensors can produce a “chill map” that shows which of a farm’s acres are getting cold enough to be sufficiently chilled, and which might require chemical assistance to counteract the side effects of poor chill. Similarly, during heatwaves, the AI can crunch dendrometer data to determine which of a farm’s trees are stressed by a lack of water, and which are holding up. Such use cases can help keep crops healthy while limiting water and chemical waste, a benefit to both farms and the environment.
Better logging practices, for instance, may not make many headlines. But Zareian notes that, while “many companies focus on generic models or platforms,” AI companies that want to prove useful and survive will have to go “deep into solving very specific, data-heavy problems in high-impact, underserved sectors.” And those data-heavy problems won’t be solved overnight.
Staying competitive in the long term
Once you set aside ridiculous, news-driving claims that the AI singularity is nigh, it becomes evident that we’re investing for the long haul.
“We’re just scratching the surface,” Chan says. “The reality is that the transition in how you do your work and implement these tools can take a really long time. AI is going to become more pervasive, but that goes with education. Then we’re not adding in AI as an afterthought, but really thinking about how we can use it.”
Is Canada ready to take advantage of this potential? While nowhere near the scale of the United States and China, Canada claims 9.6 percent of the global AI market. By some metrics, such as researcher numbers, we appear to be in a strong position, but we’re slipping in other statistics, such as AI infrastructure (like data centres and computer hardware). The 2024 budget set aside $2 billion for AI, but Zareian notes that more can still be done.
“Canada,” Zareian says, “has been a fantastic place to build an AI company thanks to its strong research ecosystem, diverse talent pool, and early government investment.” He adds, however, that many startups struggle to jump from laboratory theoreticals to commercial application.
“More targeted funding for pilot programs and better incentives for Canadian companies to adopt domestic AI would go a long way in supporting sustainable growth.” The nearly-trillion-dollar manufacturing sector is one of Canada’s biggest, and resource extraction, like mining and oil and gas, isn’t far behind. Every new efficiency helps. Assembly line analysis and data-enhanced crop rotation strategies will continue to slide under the flashier headlines, but long after the short-term hype dies down, these uses are exactly where Canadian industries can make smart, sober tweaks to their workflows in a way that, ideally, free workers to perform more valuable tasks, not send them to the unemployment office.
“I don’t know if there’s another industry that’s seen as much change and development as we’ve seen in AI,” Chan says. “There’s lots of expertise we have here in Canada, and we’d really like to continue to figure out where the opportunities are, and how to pursue them in a safe and responsible way.”
This article was made possible by Google and the generosity of readers like you. Donate today.
