Edmonton is a global leader in the field of artificial intelligence. I recently sat down with Dr. Jonathan Schaeffer, former Dean of Science and current professor at the University of Alberta’s Department of Computing Science, to hear his thoughts on Edmonton’s role as an international center for artificial intelligence research and development.
YEGBiz: Jonathan, tell us how you came to your current role and what you’ve been doing in the last couple of years.
Jonathan Schaeffer: I’ve been at the University of Alberta almost 35 years. I came in January 1984 as a Lecturer and the following year become an Assistant Professor. My research area is Artificial Intelligence. I’ve had a fantastic research and teaching career at the University of Alberta.
In 2005, I got tapped on the shoulder to have a career change. I got into administration and stayed in administration for 12 to 13 years because I discovered that I could get a lot of gratification by helping other people be successful.
I was very successful as a researcher. Many people were involved in helping me achieve significant accomplishments as a Chair of Computing Science, as an Associate Vice President at the University of Alberta, and then for the past six years as Dean of the Faculty of Science. I could make decisions that positively impacted people’s careers, making them successful. My long-term legacy isn’t my career, but it’s the people who I helped to make successful.
I did over six years as Dean. As of 10 days ago, my deanship has ended, and I am now back to being an ordinary professor, which means after one-year leave, I’ll be going back to doing research and teaching just like I was before 2005.
YEGBiz: Jonathan, in your mind, what is the definition of Artificial Intelligence?
Jonathan Schaeffer: There are many different definitions out there, and people will argue as to what the real definition is. People think that Artificial Intelligence is all about creating human intelligence in a computer. I don’t even know what that means, but I like to think that I’m not creating intelligence. What I’m doing is creating the illusion of intelligence.
Computers are fundamentally different than the human brain. They work in very different ways. They have different strengths and weaknesses than the human mind. What we’re doing on computers is developing solutions that have a result or output that looks like the kind of decision-making capabilities a human would have. How the computer arrives at those decisions and the reasons for those decisions are completely foreign to what a human would usually do. We’re not creating intelligence. We are creating systems that mimic intelligence. I like to think that I’m creating the illusion of intelligence.
The definition, to me, is that it doesn’t matter how I achieve the result. The only thing that matters is the end result. When you look at how autonomous vehicles drive on their own it has no resemblance to what a human does. On the other hand, if they can drive on the road and they can drive safely, do we really care what’s inside the black box? Absolutely not. It’s the result that matters.
YEGBiz: What are some of the major applications of Artificial Intelligence today?
Jonathan Schaeffer: The world has changed. AI has been around for decades and has been in widespread use for over 20 years. People use AI on a daily basis, and they don’t even know it. A simple example is a credit card transaction. Every time you make a credit card purchase, there’s AI behind the scenes.
What is the AI doing? It has observed every single credit card purchase you’ve ever made – where you buy, what you buy, the value of what you buy – and it builds a model of you. Then when you make a credit card purchase that is inconsistent with your past buying behaviour, it will flag it as possible fraud.
For example, once I was in Edmonton, and I received a call from my credit card company asking about a large purchase about to be made in New York City. The system knew that I made a purchase the day before in Edmonton. The transaction to be made in New York was over $20,000. I never purchase $20,000 items on my credit card. I’m a professor, I do not buy luxury items, and this was utterly inconsistent with my purchasing patterns. They flagged it. It was not a real purchase on my part, and it got denied.
Sometimes, I get called, and it is just safety. It’s an unusual purchase, and the credit card company approves it because I can’t be predictable all the time. It’s a game of poker in some sense that the more predictable you are, the better the AI is. Credit cards are a good example.
Another example that everybody uses: email spam detection. AI monitors what email messages you read, what ones you delete, which ones you ignore, and it builds a model of spam. It’s not 100% perfect, but it gets it right most of the time.
What has changed more recently, it’s about a decade old now—people will say it’s a new technology, but in my mind, it’s not a new technology. It’s an incremental improvement on old technology, and we call it Deep Learning. What Deep Learning has done is take some rather old AI tools and make them much more powerful. With that power has come some amazing results.
For example, Deep Learning is incredibly powerful at image recognition. Facial recognition or visual recognition of any kind of object, scenery, faces, tools. Choose whatever you want. Humans were always better at image recognition, but Deep Learning has allowed computers to become superhuman.
For example, I can look at you and I know who you are. We’ve met before. It’s easy for me to recognize you. But, maybe it’s not. The room you are in is not well lit, but I can still handle it. Your face could be at an angle, so it’s partially occluded. You might have been wearing a hat. You’re wearing glasses, but if you’re wearing sunglasses, it might have been a lot more difficult. I haven’t seen you in a year or so. Maybe you would have grown a mustache or a beard.
Under ideal conditions, like me looking head-on to you, it’s easy for me to do facial recognition. But, if you’re a face in a crowd, it becomes much harder. Computers are now superhuman at facial recognition, which means there can be cameras anywhere that not only record people passing by, but they can identify who they are. That’s a bit creepy, and some people will say it’s scary and maybe it’s an invasion of privacy, but it’s an example of what this Deep Learning technology can do.
Another good example is tumour analysis. An image of a tumour is a picture. Let’s get the computers to analyze it. It turns out that humans are pretty good at identifying tumours but now computers are also pretty good at analyzing tumours. When you get a human and a computer working together, they’re almost perfect at detecting these growths.
One of the exciting things we see now is a lot of technology that I’ll call symbiotic AI. It’s computers and humans working together because humans do it one way and have strengths and weaknesses that way. Computers do it a different way and have strengths and weaknesses. When you bring them together, the combination is even better.
Perhaps the poster child for AI these days is the autonomous vehicle. If you’re travelling in a car at a 100 kilometres an hour, you’ve got to be watching the road in real time as to what’s going on. Identifying what’s happening. Making immediate decisions. You need very fast image processing and analysis. You need software that’s going to drive the car, manipulate it, plan the actions.
A highway in Arizona is pretty easy. The roads are straight, in good shape, little rain, and there isn’t a lot of traffic. Try driving in a city’s downtown core with pedestrians weaving in and amongst the cars. Add in some rain, snow, or sleet, and it becomes a very challenging problem.
At least in good weather conditions, computers are now able to drive cars at a much, much safer rate than humans can. That technology is mature, and I for one welcome the opportunity to have a car drive me. I hate that drive from Edmonton to Calgary. The first time I did it, it was interesting. The second time, it was okay. But the next 100, 200 times, it is boring.
Let the car do it. Let the car drive itself. That’s another example of AI. We are seeing it pop up everywhere. Absolutely everywhere. The world’s biggest AI research labs are places like Amazon, Google, IBM, and Microsoft because they’re developing products with AI inside. In most cases, you have no idea that it’s even there.
Sorry for the long answer.
YEGBiz: No, it’s a great answer. Keep going.
Jonathan Schaeffer: I’ll give you one of my favourite examples because nobody would ever think of it. It’s a familiar example. It’s an elevator. Who would think that you put AI in an elevator? What the AI does is, it monitors what floors the buttons get pushed and when you get on what floors you get off.
That sounds pretty easy, but people’s patterns change. In the morning, you’re probably going from your room down to the main floor. At certain times of the day, like at dinnertime or the end of the day, you’re probably going from the main floor up to your room. The patterns are probably different on weekends.
What the AI does is it monitors that and builds models of the elevator’s usage as a function of the day of the week and the time of the day. When an elevator is idle, rather than just sitting there, the AI says, “With high probability, the next button is going to be pushed around floor #20. So, the elevator, rather than sitting idle, moves up to floor 20. If you’re on 22 and you push the button, it’s a very short wait for you to get on the elevator, and down you go. An example like this is a little thing, but they are all behind-the-scenes improving our quality of life.
YEGBiz: Is Deep Learning the same thing as Machine Learning?
Jonathan Schaeffer: There’s a lot of confusion. First of all, the field is called Artificial Intelligence. It’s a terrible name. It’s not a very descriptive term. I think a more proper name is Computational Intelligence. Within the field of Artificial Intelligence and Computational Intelligence, there are many subfields. Natural language processing, robotics, heuristic search. One of the subfields is called Machine Learning.
Artificial Intelligence is much more encompassing. Within Machine Learning there are many different fields, subfields, or areas. The interesting thing about Machine Learning is it’s fragmented. In our brains, we have the ability to learn. You tackle a problem that you’ve never seen before. You’ll figure it out. You’ll learn new things. When you read a book, you’ll learn new things. But you are completely oblivious to the process of learning. But clearly, there’s an algorithm inside our brains that allows us to learn.
The problem we have in Computational Intelligence when it comes to Machine Learning is there is no one algorithm. We’ve got dozens and dozens of algorithms. They all have strengths. They all have weaknesses. One of those algorithms is called Deep Learning.
Let me give you a good example. Pretend that you’re interested in working with wood. When you’re working with wood, you need a saw to cut wood. There are hundreds of different saws out there. The saw that you’re going to use to cut down a tree is not the same saw as you’re going to use to cut lumber. It’s not the same saw as you’re going to use for fine woodworking. There are chainsaws, rotor saws, handheld saws, and many different saws. Even though there’s a tool called a saw, we have many different saws. Based on the problem that you have, that you want to solve, you choose the appropriate saw.
In Machine Learning, it’s the same thing. There are many Machine Learning algorithms out there. Given your application, the problem that you’re trying to address, you need to understand what its characteristics are, and then choose the appropriate algorithm to help solve it.
Today, Deep Learning is the buzzword. It’s a popular learning algorithm because it has had some spectacular successes. But by no means is it the only algorithm out there, or that it is the right choice for your application. Just because you have a problem that needs Machine Learning does not mean you necessarily need Deep Learning.
YEGBiz: Artificial Intelligence as a science, what were its origins?
Jonathan Schaeffer: They go back a long, long way. If we talk about Computational Science you can go back to Charles Babbage, who was the father of the modern computer. In the 1840s he designed the first physical computer. Part of his interest was applications applied to very simple games because he recognized that many types of problems could be represented as numerical problems to solve.
If you fast-forward, people were talking about AI as a possibility in the 1940s. Some of the first applications that were developed besides numerical ones like military applications were games in the 1940s and 1950s.
The first real Artificial Intelligence event was held at Dartmouth University in the summer of 1956 when a number of pioneers in the field got together and exchanged ideas. That’s the watershed moment where Artificial Intelligence came into its own. That’s really only 60 years ago. Since then it’s been growing rapidly in popularity in terms of an important research area.
YEGBiz: You’ve been involved personally with more than one AI project. Give us the background of what you’ve done?
Jonathan Schaeffer: I was interested in AI going way back—I hate to date myself. As an undergraduate student at the University of Toronto in the late 1970s, I was a very strong chess player, a master. I had dreams of becoming the World Chess Champion, which was rather naïve. But I went into computing science, and I realized that if I couldn’t be World Chess Champion, then maybe I could write a computer program that would be World Chess Champion.
In the late 70s, I started reading about computer chess. In 1979, I decided that I was going to go to graduate school and do a Master’s in computer chess. Literally, for almost 40 years, I’ve been doing Artificial Intelligence research and using games as the way of demonstrating my Artificial Intelligence work.
Very early on, people said, “Well, you should be using AI for an important application. There’s a medical diagnosis system. You should try and improve that.” Or, somebody had an idea for using AI to understand legal cases and be judge and jury. I said, “You know what? Medicine and law are important applications, but I have a lot of fun with games. I can explain games to people.” Since then I’ve been doing AI research using games as my experimental test bed.
I do not research games; I do research into Artificial Intelligence, and I use games to demonstrate my research. My research, for example, has been used in Bioinformatics, GPS systems, the commercial computer games industry, and a variety of other places. Games are at the heart of how I demonstrate the ideas and then let other people work on generalizing them into other real-world applications.
YEGBiz: Jonathan, were you able to patent any of your work?
Jonathan Schaeffer: I looked at it very early on. Three or four times I was encouraged to do patent applications. I started down that path, and then people told me it was a waste of time. The reason is very simple: unless you’re willing to protect the patent, there is no reason to do it. The university in the 80s and 90s, when I was generating a lot of IP, they weren’t willing to pay for the patenting process. I didn’t see the value of doing it.
Now, in AI, if you have something good, nobody patents it anymore because the field is moving so quickly, the length of time it takes to get a patent is unacceptable. The cost, especially if you’re going to protect it around the world, is unacceptable. The answer is no – I have no patents – and I have no regrets.
YEGBiz: How does Edmonton rank globally for its AI prowess?
Jonathan Schaeffer: That’s a multifaceted question. Some people would say it’s hard to answer. Everybody will spin their story in whatever way that makes them look advantageous. We have built over the last 40 years a suburb group of Artificial Intelligence researchers here in Edmonton.
YEGBiz: Did you say 40 years?
Jonathan Schaeffer: The last 40 years. Some people have come and gone, but over the last 40 years we’ve been building our expertise up. When you go to the academic community, people in the academic world know we were very good. But we were invisible. As one of my friends once told me, “How did you guys build such a strong AI Research Group in the subarctic?” Edmonton has a bit of a credibility problem.
About two years ago, we accidentally got discovered. There’s a site called csrankings.org. What it tried to do is rank Computer Science departments based on an objective measure. There are a lot of other measures out there. Some of them are qualitative. Some are quantitative. They’re all fuzzy. What csrankings.org said is “We’re going to identify what everybody acknowledges as the premier publication venues in the world. We’re going to figure out who’s publishing the most in the best places because that’s a measure of quality.” Some people will argue with that point of view. It doesn’t matter. It’s completely objective. There are no qualitative fuzzy factors. It’s all hard numbers.
When the rankings came out two years ago, much to people’s shock in the area of Artificial Intelligence and Machine Learning – remember, Machine Learning is a subset of Artificial Intelligence, but in this ranking, they separated the two – we came out as #2 in the world, and people were stunned.
Nobody was surprised at #1, Carnegie Mellon University. But #2 shocked a lot of people. You’d have to go further down the list to find Harvard, Stanford, and MIT. The University of Toronto was at 12 or 15, somewhere there, but we came out as #2. Even if you go back into the mid-1990s, we were always near the top. It wasn’t a flash in the pan.
Right now, if you look at the recent rankings, we’re #3 in the world. A major Chinese University has overtaken us. We’re never going to catch up to them because the rankings are quality and quantity. All the places that are above us, and even some that are below us, have the quantity. They have a lot more people doing a lot more research than us. So, we’re very proud to even be in the top 10. It’s incredible, but to be ping-ponging#2 and #3 for 25 years is absolutely incredible.
YEGBiz: This is the result of a 40-year investment. Then for the last 25 years, we’ve been in the top 2 or 3 spots as a result of that focused investment by the university. Would that be correct in saying?
Jonathan Schaeffer: I would disagree in that in the sense that if you say it’s an investment, it implies that there was a strategy. There was no strategy. In the late 80s and 90s, we made some key hires. The success of the few people allowed us to attract some really fantastic applicants, which we were able to convince the Chair and the Dean to hire. Around 2004, we made some big hires that really cemented our reputation.
Success breeds success. Once you get the ball rolling downhill, you gather more momentum, and we’ve gathered a lot of momentum. Which is all the more incredible because again, and I really hate saying this, we’re at a huge disadvantage. We’re the University of Alberta. We are not a major U.S. university. We don’t have access to the resources of a major university in the United States. We don’t have that big international cachet. We’ve been a n unassuming group that attracts quality, works together in a collegiate atmosphere, and produces good publications. Quietly and without any deliberate strategy, we’ve been building excellence.
I will say that one seminal result out of that was in 2002. Four of us got together, including myself. We wrote an application to the Provincial Government to create a research centre that we called AICML – Alberta Innovates Centre for Machine Learning. In 2002 the government funded it at about 2 million dollars a year, which in 2002 was a lot of money. We’ve been continually funded by the Alberta Government since 2002.
A couple of years back, the name was changed from AICML to AMII – Alberta Machine Intelligence Institute. That’s helped us because it has provided research money to bring in more students, more post-doctoral fellows, and to help us attract great professors.
Right now, you have to take the good with the bad. When we’re quiet and not on people’s radars, we were just doing great research. Now that we’re on people’s radar, everybody wants into the AI game. All of the sudden, AI is a hot commodity and AMII’s resources are in high demand.
YEGBiz: There have been some great investments in the Edmonton region recently by initiatives like Google’s DeepMind. Can you outline some other investments that have been made in Edmonton’s AI infrastructure.
Jonathan Schaeffer: I mentioned earlier that in 2004 we made a couple of strategic hires. One of them was Rich Sutton. We brought Rich Sutton here, who is basically the father of a learning technology called Reinforcement Learning.
Next to Deep Learning, it’s the learning technology that is attracting attention worldwide. It is slowly coming up in second place and continuing to gather momentum as being a very powerful tool. To have Rich Sutton here is absolutely fantastic because we are now the world centre for Reinforcement Learning. Because of that, DeepMind, which was purchased by Google about five years ago, decided to make big investments into Reinforcement Learning. When looking around for who was the big player in this area, they realized it was the University of Alberta. It helped that a lot of the key people in DeepMind were University of Alberta graduates because they were hiring these people, and the best in Reinforcement Learning were from the University of Alberta.
Last year, they announced that they were setting up shop here in Edmonton, which was a huge news story. It caused enormous repercussions in the academic community because it caught people by surprise. Since then, many other companies have set up shop in Edmonton or will be setting up shop.
Huawei from China is here in town. They probably have 12 to 15 people, all researchers working on a variety of topics, in particular, autonomous vehicles. There are several other major companies that have set up shop in town. The Royal Bank of Canada has set up their first Borealis Research Institute. It was their first non-Toronto research institute. We put it on the University of Alberta campus, and it’s grown.
The AI interest has sparked a lot of other big investments. Service Credit Union has moved into the AI space working with the Department of Computer Science. Alberta Treasury Branch is another example. Now, we’re seeing spin-off companies. AltaML started last year and is now doing a lot of machine learning applications for local Edmonton companies. A lot of Edmonton-based companies are moving into the space.
What I like to think is the following. As a researcher here for 35 years, I’ve been very frustrated because I’ve trained over 75 graduate students. My view has been that I’ve been a net exporter of this talent to the rest of the world. The world has come to Edmonton to be trained. Then we export them back to the world because there were no jobs in Edmonton or in Canada.
With Google and RBC and AltaML, and all the others, I like to think that we’re now doing less exporting and more importing. The world comes to Edmonton to get a graduate degree. Because these companies are setting up shop and growing, we’re keeping many of these people. We’re creating an AI ecosystem. That means that people like it here, and the quality of life is great. Compared to many of the places these people are coming from, it is often superior and they want to stay here. That means that we are slowly trying to help diversify the Edmonton economy.
YEGBiz: It is clear that Edmonton is a global hub for Artificial Intelligence and research. What do you think that means for its impact on the Edmonton and Alberta economy?
Jonathan Schaeffer: It’s hard to say right now. I guess I’ll say a politically incorrect statement, but we have an incredible opportunity here, a once-in-a-lifetime opportunity. Through luck or happenstance, we’ve happened to build up one of the very best research groups in the world in the hottest area of technology. Who would have believed that it’s here in Edmonton, Alberta, Canada? If we don’t take advantage of it, then we are stupid.
The long-term impact, particularly in a province like Alberta that is heavily dependent on the resource sector, a sector that is depressed, a sector that has a large unemployment rate, a sector that a lot of people believe is going to rebound. Quite frankly, I think they are delusional.
To have the hottest technology area in the world, and to have an incredible opportunity here means that we should be doing whatever we can to invest in it, grow it, and diversify. Will we ever—will AI exceed the GDP that comes out of the natural resource sector? No, but if we can create thousands of high-paying, highly-skilled jobs in Edmonton – including growing new start-up companies, enhancing existing Edmonton and Alberta-based companies, attracting global companies here to Edmonton – then we will have moved a long way away from being so dependent on the oil and gas sector. Therefore, I think that as a community, as a city government, and as a provincial government, we should be investing heavily in here to try and strike gold.
Whenever you make investments, it’s a risk. Any investment is a risk, but if you risk nothing you’ll never get anything. Right now, we’ve put all of our bets on one sector, and it’s not doing well, and it’s unlikely to do well in the long-term. AI is a hot sector, and there are other sectors out there that are hot I’m sure: the cannabis sector, and other things.
YEGBiz: It’s smoking hot. The cannabis sector is smoking hot.
Jonathan Schaeffer: Okay, good.
The government and private sector, philanthropic, and wherever there’s money, needs to realize that we must make strategic investments. We need to get off the oil and gas resource sector bandwagon. AI is an incredible opportunity. It could go belly-up. Maybe everybody will get carried away by Google or Microsoft, and we’ll be left with nobody here. I think that’s unlikely. I think we should be doing whatever we can to ensure that this technology sector has every opportunity to succeed.
YEGBiz: Jonathan Schaeffer, thank you very much.
Jonathan Schaeffer: Thank you.
Kurian Mathew Tharakan is the founder of sales and marketing strategy firm StrategyPeak Sales & Marketing Advisors, and a 27 year veteran of the sales and marketing industry. He has consulted for companies in numerous sectors, including Manufacturing, Distribution, High Technology, Software, Non-Profit, and the Life Sciences. In addition to his consulting practice, he is also an Executive in Residence at two business accelerators, NABI and TEC Edmonton, where he assists clients with their go to market strategies. Prior to StrategyPeak, Mr. Tharakan was vice-president sales & marketing for an enterprise class software firm where his team achieved notable wins with several members of the US Fortune 500. Previous to his software experience, Mr. Tharakan directed the sales and marketing programs for the Alberta practice of an international professional services firm.
Kurian Mathew Tharakan
Direct: 780 237-1572