A chat with Myriam Côté, a woman in machine learning
Myriam Côté is the Director of AI for Humanity at Mila — Quebec Artificial Intelligence Institute . Myriam has been part of the institute since 2009, and prior to that has had experience in small tech startups and as a professor teaching at École de Technologie Supérieure.
I sat down with her to hear about her experiences and insights as being a woman in the tech workforce.
How has your role evolved at Mila since 2009? In 2009 we were only 30 people, including professors. I was hired by Yoshua Bengio in order to help manage the institute since I came with industrial and academic experience. With my doctorate in artificial intelligence, he thought I could help bridge the research done at the institute with industrial partners and so my first introduction to the institute was as a project manager. My role then evolved to become the Executive Director of Mila which was referred to as Montreal Institute for Learning Algorithms. Within that same period of time, I was appointed as the Director of the R&D and Technological Transfer Team. It has only been since September 2018 that I have been the Director of AI for Humanity.
Can you describe how your academic path has been? How did you pick your field of study? I am going to start my answer with this simple fact that I was someone who always enjoyed the arts. Prior to university, in CEGEP I did a double major in music and science and then entered university thinking I wanted to do a bachelors in music. I started this bachelors thinking I would be able to express myself through music and become a musician for a living but soon realized that with this choice, I would find myself teaching music for a living and that was not what I wanted.
After acknowledging that I would not become a virtuous piano player, I explored my other interests. I had always been a feminist and with my strong beliefs that women had to play an important role in society, I opted for a less traditional path. I needed to pick something that still had a creative aspect to it and back then I had a very romanticized idea of what engineering was and so… I started a bachelor in Engineering Physics. Following my undergraduate studies I worked at the National Film Board, this was a good option for me since I still had a strong passion to mix arts and sciences.
I worked there for two years and then started a master’s degree in photonic optics in Paris. I left for 3 years and did a second master’s degree in artificial intelligence. It was during this master’s degree that I was offered a PhD position in the field of artificial intelligence and I thought this was a great opportunity for me. I obtained my diploma from École National Supérieure des Télécommunications ( currently referred to as Telecom ParisTech ). My PhD was a joint venture between three universities and enabled me to travel back and forth between Montreal and Paris.
Would you consider yourself among one of the first women in Artificial Intelligence? In my PhD there were women and men, more men than women though. However, in my bachelors, I was one of the 6 women among 30 men and when mixed with electrical engineering students I was 1 of 6 women out of 100 men.
As a woman, have you ever been subject to bias? Yes, but it can be subtle. I did not feel like I was set aside because I was a woman while studying. When I started working, I felt like I was treated differently and I think this is because the working scene is a little more complicated. You are faced with different generations, cultures, and hierarchy. What I can say is that I have noticed a change over time because people are speaking up about it.
Have you ever had a mentor or someone you looked up to? I wish I had a mentor! I did look up to someone, this person was a woman in my family that was 20 years older than me and that was very modern for her time. My mother was more of a traditional woman, however, education was very prioritized at home and never was I discouraged while pursuing my studies.
Have you ever seen bias in a dataset you played with? Datasets aside, have you seen biases in any other context? My dataset was MNIST so, there wasn’t really place for a bias in there, however, I’ve seen it in other visual datasets. Outside of datasets, yes I remember one specific conference I participated in while working at the NFB. Certain cameras were on display and you could look into them to see what was filmed and on display were women working out in leotards, it made me feel very uncomfortable. This is how they were displaying their material which I felt was not so appropriate as a young engineer.
Documentaries are meant to portray reality but truly, they are the representation of someone’s perception and could be defined as bias. Do you think robots will have their own perception? The robots will execute whatever they are taught. However, the question is about the data the robot is trained on, if the data is taken from our society where biases are present then yes, this will be reflected in the robot’s actions. This is why the work that the biaslyAI team is doing is important and not everybody wants to go through the procedure of building a dataset.
Do you think it is weird to try and quantify bias? Well, yes! It is a step to take however perception can be very different for every individual and who are we to say what is what. What will the baseline mean once you quantify bias? There is a whole cultural dimension which is already hard to understand. As mentioned by your team, it is important to ask people in the field of linguistics or gender studies what they think of gender bias (which will usually be described in qualitative form). If we take laws as an example, a law is supposed to reflect social consensus and morale; this is why laws change and differ by country, because opinions may not be shared across countries. This applies to bias and this is why we will need to apply standards. Developing and defining methods for an ethical artificial intelligence falls under the same train of thought and the Montreal Declaration of Responsible AI is a good example of this effort.
