Montreal Researchers Accelerating AI in Health

How well-positioned is Montreal in the field of artificial intelligence for health? 

There is a wealth of data that is constantly created, not just in Canada, but in many other countries as well. Bringing healthcare up to speed with technology is one of the largest problems among the practical applications of artificial intelligence. In Montreal, we have a unique combination of talent, know-how, and established organizations around universities like McGill, Université de Montréal and so on. There is expertise here in healthcare, medicine, and neuroscience. Neuroscience is a field that I am quite interested in and I am working on the intersection of AI and neuroscience. There is also plenty of expertise in biology, neuroscience, healthcare, and a variety of other aspects of medicine. At the same time, and especially recently, there has been plenty of expertise in artificial intelligence and computing. The synergy of those two is quite unique and that makes Montreal a very attractive place for companies who would like to work at the intersection of AI and healthcare.  

“Bringing healthcare up to speed with technology is one of the largest problems among the practical applications of artificial intelligence.” 

I would point to talent, know-how, and the fact that Montreal is a top university city. It was ranked above Vancouver, Toronto, and New York, so it is the best university city in the world. It has the largest number of students, including post-secondary. This is really unique. It is almost like a Silicon Valley of Canada. Just like Stanford and UC Berkeley produce talent to fuel the Silicon Valley economy, that is what is happening in Montreal now. 

What are some applications of artificial intelligence in healthcare that you are most excited about? 

We had multiple discussions with my colleagues and started several projects under the umbrella of a so-called COVID-19 task force. It started from various conversations with my colleagues in math and biology like Guy Wolf, for example. We were wondering if they can do something about data analysis to determine the most critical factors, whether genetic or related to a particular person’s health state. It was a totally open question of what determines a disease’s progression trajectory. There is so much variance there.  

“We started working on different projects where we are trying to see how AI can help with different aspects of pandemics.” 

Many people at Mila, including Yoshua Bengio, immediately joined the task force that we initiated with Guy Wolf and Joelle Pineau, as well as with folks from Valence Discovery, the startup that is collocated with Mila. We started working on different projects where we are trying to see how AI can help with different aspects of pandemics. With Yoshua Bengio and his team, we worked on AI-based contact tracing.  

The idea with Yoshua Bengio and his team was to make much more efficient notifications by taking clues from the data in terms of who the person met but also who those contacts met. It is basically propagating the information and using probabilistic inference to assess the risk of you being exposed and being likely to develop a disease. Doing that probabilistically using inference techniques from machine learning was supposed to provide extra speed in terms of notifications and give you some way of judging the confidence of this kind of inference.  

What is being done to train, attract and retain top talent in artificial intelligence in Montreal? 

The CIFAR program is amazing in terms of providing incentives for people to join universities in Montreal and Mila. The CIFAR program was announced in 2017 and there was $175 million invested. It is a huge investment by the Canadian government and it creates a positive feedback loop or virtuous cycle. There are significant investments from the government towards fundamental research, helping to attract both professors and students.  

“There are significant investments from the government towards fundamental research, helping to attract both professors and students.” 

There are many other grants available that professors can apply for. This is really unique to Canada. Investment from government and also companies has been key. Montreal is now home to many AI research labs: Microsoft, DeepMind, Facebook AI Research, and even virtual labs between IBM Research and Mila. I came from IBM Research and part of my mission was to make sure that we establish tighter collaborations. Companies are both supporting fundamental research with grants and providing resources.  

What recent foreign investments in Montreal’s AI for health space would you highlight?  

Back in October, Novartis opened a lab in collaboration with Mila. There were multiple other companies before but this is the most recent example. The rationale behind that was similar to what was previously mentioned by other companies: companies want to be collaborating with the top minds in AI, healthcare, biology, and medicine. In this sense, Montreal is a very attractive place. I understand there is a pandemic right now but there are many startups that are medical or biology-related that are located next to Mila in the same building, and this is an advantage. Novartis’ new lab is following a hot trend and it is nice to see that more and more healthcare companies are choosing Montreal and Mila in particular as close collaborators. 

How embedded is Montreal in the global artificial intelligence community? 

Montreal, especially recently with everything made virtual, is very connected not just with universities and research labs around the world but also with companies.  

“There is collaboration not just among labs located in Montreal but also with their counterparts in different countries.” 

There are very close collaborations between Mila and European AI organizations. One example is the Max Planck Institute for Intelligent Systems, Bernard Scholkopf’s group, doing very hot topic work on discovering causal variables, which is a very important topic for healthcare applications in trying to build models that are robust to changes in data, to out-of-distribution generalization. This is a hot topic, which seems to be core to pushing AI to the next level. It is also very important for healthcare and medical applications. Collaborations along those lines go on in Europe and with multiple universities and research groups in the US such as Berkeley, Stanford, and the Massachusetts Institute of Technology (MIT). We are also collaborating with IBM Research. There is also Facebook AI Research and DeepMind. There is collaboration not just among labs located in Montreal but also with their counterparts in different countries. Mila is really virtually connected to the rest of the academic world but also to industry.  

Thanks to the reputation created over the past decade with the advancement of deep learning in Montreal, the city is truly an international leader in AI. We have collaborations not just with European or North American companies but we have quite a bit of support from the Korean research institute, Electronics and Telecommunications Research Institute (ETRI). I just had a conversation last evening with Yoshua Bengio and with a colleague from ETRI—it is an ongoing collaboration with Mila. We have connections with labs, companies, and universities across the world. It is truly international. 

How do you see the Montreal artificial intelligence ecosystem evolving in the coming years? 

We are adapting to the pandemic and this online existence, and it is well known that adaptation is pretty much the core property of any intelligence system and the AI ecosystem in Montreal is quite intelligent at adapting. In some sense, it is even better now in terms of connectivity with other labs and people across the world, because all the conferences are online now and you do not have to travel there. In terms of predicting the future for Montreal’s AI ecosystem, I only see a steep upward trend. I might be overly optimistic, but something tells me this optimism is well-founded.