Another global first for Edmonton
By Michael O'Toole

Photos By: Ellis Brother Photography

Artificial intelligence may still be widely viewed as the yolky domain of eggheads, spindly academics bent on establishing that angels can indeed dance on pins.

Nevertheless, as with so many other latent conveniences in life, most of us rely on this often invisible task facilitator everyday, but seldom have the grace to say thank you. This is particularly regrettable when we consider that AI is perhaps the one algorithm that might actually have the wits to appreciate the gesture.

“The exciting thing is that artificial intelligence is real,” declares Dr. Russ Greiner from the University of Alberta’s AI research unit—a man whose mild resentment of the egghead label does not prevent him from fraternizing forgivingly with representatives of his own species. “People are making money from it. It’s not just an academic exercise or a frivolity. Part of our mandate is to get business interested in what we’re doing.”

Greiner’s speciality of machine learning is itself a vast, seemingly limitless sub-category of AI predicated on the effort to detect patterns and make decisions based on those patterns. “Organizations that employ artificial intelligentsia—machine learning (ML) people—to solve problems do so because there’s no other technology that works for these problems,” Greiner asserts.

It seems there are three main situations where this typically holds good: 1) when nobody knows the answer; 2) when people know the answer but can’t articulate it; and 3) when the answer may be known and easy to articulate, but it’s just too cumbersome to keep redoing the task. Spam filtering is a choice candidate in the last category, while healthcare, a key focus of Greiner’s research efforts, is a major potential beneficiary in the first scenario.

“What genes are associated with this disease? The way ML approaches this is to build a classifier. You give me lots of data. Give me 100 patients who’ve been at the same hospital… 50 patients have the disease and 50 don’t. Tell me their name, age, gender, smoking history, maybe their genotype. Then I’ll find patterns, or my algorithms will find patterns.”

Greiner’s algorithms, with the help of the Alberta Ingenuity Centre for Machine Learning are currently partnering with the Cross Cancer Centre in Edmonton, divining patterns that will identify patients who should respond well to a certain form of treatment. The research is wide ranging and, in some respects, groundbreaking. “My team has built the world’s best analyzer of brain tumors—the world’s best, right here. This is a project that people at MIT and McGill have tried to solve, and we solved it by using machine learning techniques. And that’s real data, real concerns, real interests—a project which makes a difference.”

Beyond medical science, the pattern-finding capability of ML—and its AI cousins —makes it a key component of credit card security, e-commerce, delivery routing, load level prediction in the energy industry and a whole array of business applications that fall mostly under the umbrella of business intelligence. Internet search applications alone are legion.

So who should be talking to Greiner and his colleagues?

“Anybody who has data. Try to find me a business sector that doesn’t have a gigabyte or two sitting around and problems to solve. We’re trying to get people excited about this. In fact, we are already an international player. We have collaborations around the world. Edmonton’s been a harder sell. It’s frustrating for me because we’ve got such a great answer and there are so many applications that can use it.”

An Edmontonian long since sold on the proposition is Celcorp’s founder and chief technology officer, Bruce Matichuk, one of the region’s most vocal and prolific champions of artificial intelligence and its power to leverage business operations. His own company, which specializes in systems integration, is already a veteran exploiter of the technology.

Donning his industry cheerleader hat, though, Matichuk turns to the more mall-friendly realm of video games. “If you look at revenue worldwide, currently the pc games industry alone makes more than Hollywood, so it’s absolutely gargantuan and doubling in size regularly. At the core of many of these games are things called AI engines.” These, Matichuk expands, are all of the code and algorithms that make you, the player, think that you’re interacting with another person—a character that is actually using reasoning to decide what its next best move is.

Encouragingly, Edmonton is by no means a passive observer in this massive playpen. Sponsored by the Alberta Ingenuity Fund, the University of Alberta proudly sports an electronic entertainment and games research centre described by Greiner as the strongest in the world. “We’ve contributed ideas to Electronic Arts to help them debug their programs by using our technology,” says Greiner. “That’s been very effectively used.” 

Also a key partner is the high profile Edmonton games producer BioWare, which is currently harnessing the power of AI in world class products like Baldur’s Gate.

So much for the software. What about technology you can really roll around in the mud with? Michael Bowling, another jovial star of the U of A’s AI research hub, is proud to represent robots of all shapes, personality types and political leanings. As if they can’t speak for themselves.

“There certainly is a fairly heavy industry associated with robotics,” Bowling points out, “but artificial intelligence in robotics is still fairly undeveloped, so it’s unclear what the market looks like yet. One of the things I’m interested in is to push the intelligence and understanding of our world enough that we can start deploying robots in more situations where it wouldn’t be cost effective to have a human monitor it.”

With this in mind, Bowling is currently involved in geo caching research that requires a mobile robot to find a hidden object by using GPS coordinates and its own precocious talent for acquiring street savvy. “How do we build a map of our environment, especially an unstructured environment?” Bowling posits. “There has been a lot of research on how to do this inside of buildings. But now, try to take that same technology and move it outdoors. It doesn’t quite work so well. With its sensors, it has to figure out that this is drivable terrain, this is not, or maybe this is risky and if I can’t figure out any other way to go then maybe I should drive over this.”

Among the future applications of AI robotic technology, Bowling sees huge potential within the oil and gas industries, where it currently makes no financial sense to send humans to carry out inspections of facilities on a regular basis.

In common with other areas of artificial intelligence, though, entertainment appears to be the motor zone. The latest incarnation of Sony’s celebrated AIBO Dog can recognize a 1700-word vocabulary and incorporates learning techniques to mimic canine behaviour so you can watch your robotic dog ‘grow up’. The dogs are among several robot types that regularly compete in the four separate AI soccer leagues at the U of A—all, despite the chuckles of onlookers, as a testbed for various sober, practical applications of the underlying technology. While Sony is turning its attentions elsewhere, Bowling will continue improving AIBO hardware to solve general problems.

“The reason AI in robotics is not as big as I think it will be in future is that robots are still a little bit expensive,” Bowling concludes. “As the price of robots goes down, that will change.”

Within the artificial intelligence industry as a whole, image problems linger. “AI as a field has made some huge blunders over the last several decades,” Greiner reflects ruefully. “We’ve had, unfortunately, some charlatans, which makes everyone call into question anything we say.”

Matichuk shares the concerns about public perceptions, also finding irony in the fact that as soon as the technology actually becomes useful, people don’t think of it as artificial intelligence any more. Though this societal mindset can be a hindrance from a marketing perspective, Matichuk takes heart in observing, just a little quirkily, that “these algorithms are finally getting out of the research labs and into products that people are using every day.”