Toronto is Leading the Charge of the Global Robot Rising
Written by Shelley White
For many people, the term “artificial intelligence” conjures up visions of brainy humanoid robots who frequently have a bone to pick with their mortal creators.
We can blame pop culture for that. From 2001: A Space Odyssey’s deranged (and disembodied) HAL 9000 and the doomed replicants of Blade Runner to the murderous, nearly-human “hosts” of HBO’s Westworld, artificial intelligence in TV and movies is often about sentient beings getting too smart for humanity’s own good.
Despite the sci-fi connotations, artificial intelligence (AI) isn’t just a futuristic fever dream. It’s already being used in many ways in industries across the spectrum—from service and manufacturing to finance and law. And that AI revolution is particularly strong in Toronto, which is shaping up to be an epicentre of AI innovation and enterprise, leading the way for the rest of the world.
A legacy of AI expertise
On the academic front, Canadians have been preeminent in AI for decades. Home-grown luminaries like Geoffrey Hinton* and Yoshua Bengio were pioneers in the field of “deep learning” at the University of Toronto in the 80s and 90s, developing some of AI’s foundational algorithms. Now, the federal government, industry and academia are determined to build on the country’s strong AI foundation and translate it into business success.
As part of his new budget, Prime Minister Justin Trudeau pledged $125 million to support AI research centres in Toronto, Montreal and Edmonton. One of those centres will be the Vector Institute for Artificial Intelligence, a non-profit, independent facility that will be located at MaRS in Toronto. The institute will be helmed by AI “godfather” Hinton, who as part of a Google AI team, has spent the past few years producing hundreds of research papers on artificial intelligence.
The Vector Institute aims to build on the expertise of the renowned deep learning team at U of T by producing, attracting and retaining top AI talent.
“Over the last 10 years or so we’ve seen a drain of really excellent people going to other places, like London, England, and New York and to the Bay area,” said Brendan Frey, co-founder of the Vector Institute, on CBC Radio’s Metro Morning in March.
“The reason we need the institute is to bring people like that back and to bring other international, world-leading talent to Toronto and to support the AI ecosystem in Toronto and in Canada more generally,” said Frey, who is also the CEO of Deep Genomics, a company that combines expertise in machine learning with genome biology.
But perhaps the bigger news is that the Vector Institute will be a public-private collaboration. It received a total of $170 million in funding—$40 million from the federal government, $50 million from the Ontario government, and $80 million over 10 years from a group of 31 corporate donors—including Shopify Inc., Google, the Big Five banks, Magna International Inc. and Air Canada.
It’s a vote of confidence from the business world that AI is here to stay and recognition that Toronto is the place to be when it comes to studying it. It’s also an acknowledgement that AI isn’t just a theoretical exercise about what could be, it’s about applying those technologies in the world we live in now.
Making waves in the design space
But what is AI exactly? As Frey puts it: “A good way to think about artificial intelligence is it’s computer software that’s indistinguishable from human intelligence.”
In other words, it’s the ability of a computer to understand what you’re asking and infer the best answer from the available data. That could be Siri helping iPhone users or Google DeepMind’s artificially intelligent player AlphaGo beating a top-ranked human player at complex board game Go. As for deep learning, it’s about systems beginning to progress and learn on their own.
One of the biggest players harnessing AI in the design space is software giant Autodesk, which is based at MaRS in Toronto. Mike Haley, Autodesk’s senior director of machine intelligence, says AI is on its way to revolutionizing the design process, a transformation that has already begun. He says the first use of AI in design is to eliminate the more tedious parts of the job.
“Designers spend an enormous amount of time doing things that are not creative,” says Haley. “They are doing things like trying to find the right model for something or figuring out the right software. They are the tedious things you have to do surrounding that core creative part. In the next five to ten years, you’re going to see more and more AI helping with those things.”
Autodesk has two products in pre-market stage that harness AI—the Design Graph and BIM 360 IQ.
The Design Graph is used in the design stage of a building project, to create a “living catalogue” of every component and design that a team has ever created. Through machine learning, the program learns to identify and understand designs based on their characteristics—like shape or structure—rather than labelling. And it’s not just relying on logic that’s been programmed into it, it’s learning independently, identifying patterns and making predictions based on what it has learned and continues to learn. The Design Graph is currently available for free in beta form.
“You can think of it as the language of design,” says Haley. “The design graph has the ability to adapt and add to its knowledge, based on what you’re doing.”
The BIM 360 IQ works at the other end of the spectrum, allowing people to manage the construction of buildings more efficiently. BIM stands for Building Information Module, “the modern term for the full set of information for building a building,” says Haley. It’s currently piloting with large construction companies.
Like the Design Graph, BIM 360 IQ takes in large quantities of data, but in this case, it’s all related to the construction process. Based in the cloud, the program takes advantage of the natural aggregation of data there, learning as it goes and as the client adds data to the mix. After synthesizing all the information, BIM 360 uses AI to predict risk.
“It simply ranks things in terms of risk,” says Haley. “Is the concrete going to arrive on time? Did we go over price? Are we on schedule?”
Construction can be a messy process, he says, and this kind of product can reduce the mess and increase efficiencies.
“AI isn’t the panacea that solves every problem, but it can learn in complicated environments,” says Haley. “Whenever anything is messy, that’s a good hint that there might be a value for AI.”
Setting precedents in the legal field
Another industry where AI is making headway is law. Blue J Legal is a Toronto-based company that creates AI-driven tools that allow lawyers to predict the outcomes of cases.
The company’s lead product, Tax Foresight, harnesses deep learning to analyze case law, discovering “hidden patterns” and providing relevant cases in seconds. It generates tailored explanations of its analysis, and it’s 90 percent accurate, says Benjamin Alarie, CEO of Blue J Legal.
“It's not going to replace professional judgment, but it's a really great starting point for professionals to say, ‘Where are we on this one?’” says Alarie. “In five minutes, our tax accountants and tax lawyers can get a really good sense of how they're doing.”
Similar to Autodesk’s AI tools, Blue J Legal’s system is fed lots of data. “There are hundreds of cases that we can train the system on and new ones are getting decided all the time,” says Alarie. As new cases get decided, they enter them into the system – the “raw materials” to help the tool make predictions.
They’ve started with tax law, but Alarie foresees their technology expanding to include employment law, immigration law, contract law, and even criminal law. He notes that legal research is going through a paradigm shift. First, research was done in an analog way (with books), then it moved to digital. The next paradigm shift is computational research, says Alarie, which is where AI gets involved.
“That’s happening now. Lawyers are making decisions influenced, in part, by computational legal research that harnesses the insights from hundreds of cases that they didn’t have time to read otherwise,” he says.
The possibilities for the future are intriguing—imagine the average citizen being able to interact with some kind of AI system to get real-time legal guidance. Or someday, algorithmically-influenced decision-making could even become the law.
“I don’t think we want to rush into any of these changes, but I don’t think we can resist them forever either,” says Alarie. “I think it'll happen slowly as the systems show their capabilities and it's going to improve over decades. We can already see it's starting to happen now.”
Scheduling meetings, and a whole lot more
Toronto’s Zoom.ai wants to give everyone their own automated assistant, says company founder Roy Pereira. It’s a way to eliminate the myriad administrative tasks, or “busywork,” that can distract people from their real work.
“Within companies, especially in North America, we've become very lean,” says Pereira. “A lot of the supporting positions that we had in the past are gone. And we’re really just fending for ourselves.”
A Zoom.ai automated assistant can schedule meetings, set reminders, book Uber rides, search for and book flights, provide meeting briefings and even facilitate “warm introductions” to someone you want to meet.
To ensure this automated assistant would be as efficient and accurate as possible, Pereira says they created their own machine learning system from scratch.
“A good [executive assistant] understands the person that they work for and anticipates what they really mean when they say, ‘I need a coffee with Joe,’” says Pereira. “Generally speaking, computer software is pretty dumb because it doesn’t know which ‘Joe’ you mean.”
Machine learning helps their system behave “more like a human, in some ways,” says Pereira—to deduce and predict which “Joe” the user is eager to see and where they want to have that coffee. As the system takes in more and more information and gets to “know” the user, the better it gets at predicting what they want.
Another important component of the tool is natural language processing, says Pereira. Users communicate with their automated assistant through chat, and they can understand common speech and even emojis.
The company’s main target is large enterprises, but they have a version of the product that anyone can use, from real estate agents to venture capitalists. Pereira says they recently did a successful pilot with one of the top four consulting companies in the world (which will remain nameless—for now) and they are starting the deployment phase.
Although Pereira agrees that AI is in use now, he notes that it depends on how you define it. He prefers the term “augmented intelligence.”
“A purist [might] look at our machine learning and go, ‘Well, that's not exactly what I would consider [AI].’ But there's different levels of gray, I would say.”
As someone who lived in Silicon Valley and came back to Canada, Pereira says he’s seen tremendous growth in Toronto in the past decade, both in AI expertise and excitement from the rest of the world.
“We’ve gotten a lot of engineers from around the world and the U.S who are very interested in coming to work here with us because we're in Toronto and we're AI-based,” he says. “We're getting interest from people who are working at Twitter, Google and the Valley, and that normally doesn’t happen. Normally it's the opposite – someone in Toronto is leaving to go to the Valley.”
The Trudeau government’s hefty investment in AI and the launch of the Vector Institute are positive steps because they put Canada, and Toronto, into the spotlight, he says.
“It sends a signal across the world that says we're serious about this, because this is the future.”