By: Krista Davidson
31 May, 2019
A new partnership between CIFAR and the OSMO Foundation is tackling diversity in AI with the next generation of women leaders.
The AI For Good Summer Lab program, an initiative that provides women in STEM fields with exposure to training and networking opportunities in AI has kicked off its third year in Montréal. As part of the CIFAR Pan-Canadian AI Strategy, the AI For Good Summer Labs will expand training opportunities for women to other parts of Canada over the next three years.
The AI For Good Summer Lab is the brainchild of long-time friends and colleagues, Angelique Mannella (AM Consulting) and Doina Precup (Canada CIFAR AI Chair, Mila, McGill University, DeepMind Montréal). The two realized there was a gap in AI and machine learning training programs that targeted women. In 2017, they launched the AI For Good Summer Lab.
Lack of diversity in STEM and AI
According to the 2019 AI Global Talent Report, just 14 percent of AI researchers in Canada are women. This isn’t a surprise to Precup and Mannella. “I really wanted to encourage the participation of underrepresented groups in AI and machine learning, and the obvious place to start was with women,” says Precup.
“The idea was to build a Lab that would engage women in solving problems common to society today; problems that would make a difference to the world and encourage AI to be used for good.”
“I was inspired by the Lab model because it provided opportunities that were not there for women who want to pursue careers in research and industry in a male-dominated field,” explains Mannella. “We came up with the concept of the Lab as a way to help women turn their passion for technology and social impact into careers.”
“AI for good means a lot of things to different people,” explains Precup. “To me it means using my work to make the world a better place.”
Precup grew up in communist Romania and comes from a family deeply involved in computer science, engineering and math fields. “There was no gender gap for me,” she explains. “I developed a strong passion for AI and reinforcement learning early on.”
Precup came to Canada from the US in 2000 to join McGill University’s School of Computer Science and is one of the 46 Canada CIFAR AI Chairs named as part of the CIFAR Pan-Canadian Artificial Intelligence Strategy. In 2017, the Government of Canada appointed CIFAR to develop and lead the $125-million strategy, the world’s first national AI Strategy. It includes the goal of increasing the number of outstanding researchers and skilled graduates in Canada and training the next generation of AI leaders.
“We have this powerful technology with AI, and it’s going to change the way we live and work as a society, so we really need everyone represented. Everyone should have a voice at the table when we talk about algorithms and how they will be applied.”
The Lab, which takes place this year over seven-weeks from May 8 to June 21, 2019, is engaging 30 women from across Canada in a mashup of lectures, workshops and projects that tackle real world problems.
This year, the OSMO team received close to 200 exceptional applications from across Canada.
“These are women who are smart, motivated and talented, and they want entry into the AI ecosystem. They’re passionate about AI for social good and that is a huge asset for Canada and the world,” explains Mannella.
“In terms of Canadian values and culture, nurturing diversity in talent will 100 per cent contribute to better technology and positive impacts for Canadians. Women in AI are a tremendous resource to Canada.”
Past projects from the Labs have included using machine learning to detect pathology in MRI images, using reinforcement learning to improve the delivery of energy and making cities safer for cyclists.
A tool for detecting bias in gender
Carolyne Pelletier, an AI For Good alum from the 2018 cohort and Université de Montréal graduate student, is one of the co-founders of Biasly, a machine learning tool that uses natural language processing to detect gender biases in text.
Biases make dangerous assumptions that are harmful to gender equity and inclusion in society.
“Typically in machine learning, you build a model based on a dataset. We couldn’t find any datasets to train our models on so we built a dataset entirely from scratch to create the tool, which is really novel and different.”
“We have biases that are so deeply ingrained in us that we might not realize we’re not using inclusive language to the people in our network. A tool like this could highlight online bias in advance, and hopefully make individuals and companies more aware and inclusive. There’s definitely an educational component to the tool,” says Pelletier.
Mahsa Khosravi is one of the talented women enrolled in the program. She came to Canada from Iran and developed an interest in applying machine learning applications to medical signals, and has just completed her Masters in Electrical and Computer Engineering from Western University.“I see the Labs as an important next step for my career,” she explains. “I love the idea of building and applying machine learning algorithms in healthcare, specifically in neuroscience and brain surgery.”
A network of support
For Precup and Mannella, the Lab is foundational to supporting long, successful and diverse careers in AI for women.
“After this year’s Lab, we’ll have close to 100 alumni who we can engage with and encourage them to use their knowledge and networks to build their careers and serve as mentors for future generations of women,” says Mannella.
“I would like them to have the sense that AI is a very creative discipline where there’s a lot of potential and work to do on the basics of the science of algorithm development, as well as application,” says Precup. “I’d like them to leave here with a really solid group of friends, mentors and collaborators that will be part of their network for the next 30-40 years.”
CIFAR is a registered charitable organization supported by the governments of Canada, Alberta and Quebec, as well as foundations, individuals, corporations and Canadian and international partner organizations.