Canada is emerging as a global leader in Artificial Intelligence (AI), with some of the world’s top minds and inventors of AI technology, and Waterloo is one of Canada’s main AI development hubs. Already there are more than 90 companies in Waterloo’s AI cluster and the list is rapidly growing.
Artificial Intelligence has already permeated the world around us, with such profound implications for the economy and society it can be classified as a second digital revolution, as powerful as the first.
Unlike programming computers with a detailed set of instructions, much of today’s artificial intelligence (AI) uses “machine learning” techniques to enable the software to “learn” from the data fed into it.
It is being employed in a wide array of applications, from Facebook and personal digital assistants such as Siri and Alexa to customer service call centre chatbots and advanced driver-assistance systems in cars. Increasingly, it is also making its way into industrial, medical and financial applications.
What is Artificial Intelligence about
Alexander Wong, an expert in artificial intelligence, speaks about the growth of AI and Waterloo’s role as a vital global player in the applied AI sector. Wong is an Associate Professor in at the University of Waterloo, and also a Canada Research Chair in the field of AI. He is a co-inventor of Generative Synthesis and Evolutionary Deep Intelligence, as well as co-founder of the start-up DarwinAI.
What is Artificial Intelligence and how does it work?
Everyone has a different definition for it depending on what their purpose is. But I would say a high-level definition is that it is a machine or a computer that mimics some aspect of human cognition — such a speech, vision, perception or decision making.
There are many different artificial intelligence approaches, but what we call the machine learning paradigm is being used to tackle problems that are very abstract and complex, or where there are just too many rules to possibly define using the traditional approach of coming up with the rules to create the AI system.
What you are doing is trying to get the machine to learn directly from the data itself. You would have a lot of data as well as hints as to what that data is, and you are using it to train what we call neural networks.
For example, let us say you want the system to learn the difference between dogs and cats. Using the traditional rules-based approach, the task is too difficult. If you can set up a rule that says a dog has this weight, and this height, but that won’t work because a Chihuahua is no bigger than a cat. With a deep neural network, you instead train it with the images of dogs and cats until it can correctly identify dogs and cats.
Through this repetitive process of showing it data and the information of what it should be, it is learns this particular task of identifying dogs and cats.
What are the applications of AI? Is it for robots or self-driving cars?
There are a huge number of applications. In the media, you often hear about AI for robots, self-driving cars or smartphones. Those are very important but there are also many other applications, especially in the industrial realm where AI can be extremely useful and extremely effective, but you don’t hear about that as much because it’s not as cool as running a robot.
In manufacturing, for example there are applications like defect inspections, quality inspections, logistics planning and supply management. In the realm of finance, there is fraud detection, risk analysis for loans and mortgages, stock prediction or company valuation predictions.
There are also applications like environmental monitoring. You can use it to monitor water quality contaminants and predict and identify contaminants.
In the auto sector, we hear about self-driving cars, but even before self-driving cars, there is the driver-assist technology we have now, which is not quite self-driving but it helps you drive better.
There are also applications in the medical realm. AI is already aiding clinicians to make better and more well informed decisions based on the wealth of data they have at hand.
Drug discovery, product design and marketing analysis are other applications.
So, AI can be leveraged in many different applications that may not sound as glamorous but are of huge importance.
What is causing this huge growth in the AI applications right now? Researchers have been working on artificial intelligence for a long time, but why are we suddenly hearing so much about AI now?
It is because of the massive wealth of data that we now have, as well as the amount of computing capacity that we have. It has really come about because of the data generated by the internet.
To train a neural network, you need data and you need to tell it what it means. Well, guess what? Now, people are posting a huge amount of information on YouTube, Instagram and Facebook, with descriptions. So now there is this wealth of information to learn from. With a well-balanced diet of data, the AI can come up with much better decisions.
Is the Internet-of-Things a part of that?
Oh, yes. The key thing is that in our modern world, pretty much everything has sensors in it, be it your camera, your car, your washing machine or your refrigerator. Turing all of that collected data into knowledge is where AI comes in.
What has made Waterloo such an important hub for AI?
There are many different factors, but one of the most important is the University of Waterloo. It’s not just focusing on the foundational and theoretical side of artificial intelligence but also, what makes us unique, is what I refer to as operational [or applied] artificial intelligence. We take fundamental concepts and use them to address real-world problems. We are figuring out how to make it useful and beneficial for the wide range of industrial and societal applications.
We work very heavily with industries to actually make AI happen for them, and that is one of our big differentiators. It is also why we created the Waterloo Artificial Intelligence Institute.
What is the Waterloo Artificial Intelligence Institute and what is its role in the Waterloo AI ecosystem?
Within the Waterloo AI Institute we now have over 100 different researchers. We have one of the largest such research clusters in Canada and our focus is to work with industry and with academic partners to really develop and push useable AI, be it in the clinical space, or for climate analysis, manufacturing, automotive, aerospace, or other sectors. We try to help our industry partners understand AI and how to improve it and leverage it for their own scenarios.
This has really helped to turn Waterloo into a hub for getting things done in AI.
There are a lot of businesses right now that are interested in AI and it could be a game-changer for them, but those companies don’t necessarily have the expertise in-house, not just in terms of building the AI but also in making sure it tackles the right problems. Not only do we help them come up with creative solutions or innovations, but we also help them make sure that they are leveraging the AI in a way makes the most impact.
One of the difficulties with AI initiatives is that if the people developing the AI solution do not understand the domain, such as medicine, and what is needed in that domain, then the solution won’t be as good. We have people who specialize in AI for different domains and who do understand what is needed.
Can you tell us a little bit about the growth of the AI ecosystem along the Toronto-Waterloo Corridor?
We have a big mix of different companies with different levels of maturity and understanding of AI all along the technology corridor.
There is such synergy in an area where there is so much of innovation going on, and so people are just very motivated to push forward from a technology perspective. It really helps.
From our perspective, we see companies in traditional areas, like agriculture, that don’t know as much about AI, but they are very interested in learning how it could help their businesses. We can help them with that. There are also the technology giants that have more understanding of AI, but even they don’t have expertise in some areas. We can help them through consultation and collaborative research.
What needs to happen for Waterloo and Canada to continue to be a leader in the AI field?
One of the key things that I am thinking about with a company that I co-founded, DarwinAI, is how to turn a start-up into a scale-up, especially in the AI realm.
We have good leadership in Canada in regards to artificial intelligence research, but in terms of taking an idea and commercializing it within Canada and really building it within the Canadian technology ecosystem, that’s where we have some challenges. This is something that governments need to look at.
Also, we have to promote the kind of connectivity that brings people together, to satisfy not just the needs of companies that are in the AI realm but also the more traditional industries that can really benefit from the power of AI.
What challenges remain in the field of AI and in building out the AI ecosystem?
There is what I call the triple threat that I am trying to address in my own research and also in my company.
The first problem is that machine learning is still very hard to build. It takes an excruciating amount of trial and error to build AI that is useful.
Secondly, even if you build it, you can’t just run it on portable mobile, edge, and IoT devices, which is where you need it. It needs a lot of computer resources, and you can’t really put giant server racks on the back of a tractor. You can run the AI on computer servers in the cloud, but in areas like medical, that presents privacy issues.
Thirdly, even if you build it and run it, there is a ‘black box’ problem where engineers can’t really explain how the AI arrived at a solution. Being able to do that is important for things like regulatory compliance, safety and bias mitigation. Those are the problems we are addressing with our solutions at DarwinAI.
The AI we have now is very good at very specific tasks (such as facial recognition) but we also hear about the term ‘artificial general intelligence,’ the concept of having multi-tasking artificial intelligence that can think more like a human. Is that technology on the horizon?
A lot of researchers are exploring artificial general intelligence, and it is very exciting research that needs to continue, but I think it will be a long time before that arrives. We are not really even close yet; we’re just at the infancy. The AI we have now is very task-specific, but there is nothing wrong with that. It does a job and it helps with automation and it helps people do a better job and make better decisions.
How revolutionary is AI in terms of the impact it will have on society. Is this the new digital revolution?
I would say yes. Artificial Intelligence is already a ubiquitous part of society, and I think the uses will continue to grow. If you use Google Maps, you are using AI. If you use Google Translate, you are using AI. Pretty much everything we do online now uses AI and those uses will continue to grow because AI is helping us make better decisions.
But we do need to be aware of the ethical issues as AI becomes a more integral part of our lives. That is something at we are also looking at very carefully at the University of Waterloo.
A lot of people have this idea that just because it is based on data, that it is unbiased. But if the data itself is biased, then the results will be biased. In that way, it is much like training a child with biased information and the child picking up biases.
Article published on Waterloo EDC