Artificial intelligence is at the forefront of our collective mind. From Ex-Machina, to Westworld to Blade Runner, it permeates every aspect of popular culture. It fascinates and terrifies us. As scientific advances are made, sociological questions are raised. What holds the future of AI and how can we better understand what we are creating? Jackie Chi Kit Cheung, Assistant Professor in the Reasoning and Learning Lab - School of Computer Science at McGill University answers our questions about the future of AI and Montreal as an international hub for artificial intelligence.
Jackie Chi Kit Cheung’s expertise lies in natural language processing (NLP), an area of artificial intelligence in which computational models of human languages are built to better understand text and speech in order to generate language that is fluent and appropriate to the context. Since 2015, he has held a teaching position at McGill while assisting AI pioneer Joëlle Pineau who will lead Facebook’s long anticipated AI laboratory in Montreal. On December 7th, Mr. Cheung will be one of the professors invited to speak on the subject at Mat’Inno : Ethics and Artificial Intelligence, organized by the Quartier de l’Innovation in collaboration with the Jeune Chambre du Commerce de Montréal.
Can you explain what artificial intelligence is and what are some concrete applications of it in the health department?
From my perspective, artificial intelligence is the study of how we can use computation in order to produce intelligent behaviours. Artificial intelligence is about behaviour rather than how people think and is also about producing something that is rational and useful in some way. In terms of application in health, there are many ways that artificial intelligence can help enhance the ability of physicians and also help with patient care. There’s one project that some of my colleagues here at McGill are working on, which is to have an insulin delivery system: an artificial pancreas. You can image an automated system that can monitor the blood glucose level of the patient and automatically send signals to release an appropriate amount of insulin without the patient having to worry about it. So here, it is rational because it optimizes something – which is the amount of insulin needed for that patient – in a way that is effective and useful. There are many other examples – like diagnosis. You might have an AI system that can help a doctor diagnose patients. A third example would be in administration. You can automate a lot of the administrative work that needs to be done in a hospital in terms of finding cases that are similar to other cases or to schedule patients’ surgeries, etc. So it’s about efficiency and complementing the expertise of humans.
Photo: Ryan Reisert
It seems to be a critical time for AI. Can you tell us more about that?
Recently, there’s been a lot of excitement around new AI techniques and I think there are many people who are trying to take advantage of all this excitement, momentum and opportunity to create new businesses and even new sectors of business. This seems to be a moment in the progression of the field where there are a lot of commercialization opportunities. In Montreal and in various other centers in Canada, there are a lot of well-known people in the field who have been doing great work for a long time. There’s also a lot of existing businesses looking for good investment opportunities so I think all these factors contribute to the excitement, specifically in Montreal.
What role does Montreal play in the advancement of artificial intelligence research?
It plays a big role – much bigger than you would expect from a city of its size – its punching way above its weight. Some of the reasons are because some of the key people in the field are here, for example Yoshua Bengio at the Montreal Institute for Learning Algorithms. Also here at McGill we have leaders like Joëlle Pineau who have been working on reinforcement learning, which is one big part in the reason for all this excitement in AI.
Tell us more about your research in natural language processing?
I am working in the field of natural language processing- artificial intelligence applied to human language. I work on various issues to do with extracting meaning from text and processing in order to generate new text. The other big part of my research is to actually change that information and generate some new language out of it after you have the information in a format that the computer can process. So for example one application that I’ve targeted for a while now is automatic summarization. This is where you need to take some document, you have to pick out the important information in it, and then you have to reformulate the document to keep the important information while shortening it and condensing it. My concrete application of this, in healthcare, is that we’re looking at how we can read some texts that were originally written for clinicians and reformulate them and simplify them so that they can be useful and understandable for patients as well.
In your wildest dreams what would you like to accomplish with this research?
What would be really nice is to have some automatic agents who can communicate in a human-like way. I’m not aiming for somebody who’s going to be like another human being – I’m aiming for a system that is predictable and useful and that can interact with you in a very natural, human-like way so there’s a slight distinction here. So it’s having an intelligent agent that’s sensible and useful rather than having one that necessarily humans like.