TORONTO — As the AI boom sends tech firms scrambling for more data to improve their models and puts a premium on the companies that own it, data scientists say businesses and governments need to understand that the technology won’t be useful or accurate without their work. 

“Data-driven discovery is becoming so impactful,” said Lisa Strug, a professor in the University of Toronto’s statistical and computer science departments and a senior scientist at SickKids. As businesses in many sectors and research fields like microbiology, astronomy and the social sciences all adopt AI, she says, they’ll need new data science methods and more practitioners.

Take genomics, for example. Researchers must first secure enough compute to handle all the information they’ll be processing. Then they must ensure they’ve scoped out a representative sample in the population whose genes they’re studying. 

To capture the human genome, researchers use a technique that’s similar to shooting a movie. Once the whole sequence is recorded, the resulting frames must be turned into data, which must be cleaned of errors and converted to the right format. Then the information must be further weighted to ensure it’s representative. The researchers must also define the disease or trait for which they’re looking. Miss or mess up any of the steps, and “anything that comes out at the end is useless,” said Strug, who is also the director of U of T’s Data Sciences Institute. 

While AI tools can help with some of the work, data scientists still need to know how to spot when things go wrong, and root out any biases. “These skills are so fundamental,” Strug said.

As the AI race heats up, businesses are also seeking data science know-how. Toronto-based Birch Hill Equity Partners has a 30-person team that works with the firms it acquires to ensure they’re getting and analyzing the right information. “For AI programs to be used successfully, you need to have great data,” said Gary Glasser, a managing director at the firm and head of data science and strategy.  

Birch Hill is a major private equity player in Canada. Its portfolio includes pharmacy chain Rexall, carrier Harbour Air Seaplanes, and mortgage provider First National Financial. Birch Hill previously owned hockey gear firm CCM and IT company Softchoice. 

Glasser’s team starts with the data its companies already have, then adds in industry information from third-party sources. Birch Hill may also find new ways for its firms to generate data, like applications to track drivers or sensors on assembly lines. It then looks for trends and patterns to identify opportunities to improve efficiency or drive new business. To ensure its industries remain competitive, Canada needs “better-run companies that are using their data and AI properly,” Glasser argued. 

Birch Hill’s data science team uses AI tools for analysis, and to speed up coding. Still, Glasser expects staff to develop the skills to ensure they’re getting good answers from the machines. “You can’t come to me and say, ‘Oh, it’s wrong because Claude screwed up,’” he said.  

That data matters is hardly news to the worlds of business or those working in AI. Before generative tools made their debut, headlines and thought leaders touted data as the new oil and encouraged businesses to drill for as much of it as possible.

AI developers also understand the value of data science, Strug said. “At Google or OpenAI, they know they need these people.”

In recent years, AI firms have employed legions of annotators and subject-matter experts to check and fix the data that goes into their models and the answers that come out. Those jobs are often temporary, as developers move from one topic they want their AI tools to master to the next. In the rush to grab and understand data, artists and media publications also claim tech firms have violated copyright, a charge they deny.

As Canada tries to ensure it has a place in the AI race, Strug claims data science is getting left behind. Graduates in the field are “snapped up very quickly,” she said. “There are not enough students going into these areas, because there are not enough training funds.” There’s also no federally backed data science centre akin to Canada’s three national AI institutes, she said.

Ottawa’s forthcoming AI strategy should include funding and programming for data science, according to Strug. “We can’t forget about the part where we make the data useful, so that the predictions and the discoveries are reliable.”