One foot at Facebook, the other at McGill University in Montreal, the Canadian artificial intelligence expert is changing codes with a pioneering idea: collaboration and sharing in research.
Juggling two worlds, Joelle Pineau is used to it. Originally from English-speaking Ottawa, Canada’s capital, she has worked for more than fifteen years in the French-speaking province of Quebec, in Montreal. She grew up between music – she played the violin – and mathematics. Now,at the age of 45, the scientist co-directs the eight artificial intelligence (AI) research laboratories of one of the world’s largest companies, Facebook, while overseeing the work of about 20 students at McGill University.
A double private cap – public that does not bother her. Quite the opposite. “You always have to pay attention to issues of conflict of interest and the sharing of intellectual, property,” she acknowledges without taboo.. We have to make sure that the partnership is seen in a positive way on both sides.. But there are benefits for everyone.»
Private-public collaboration around artificial intelligence
This collaborative model was established since the creation of Facebook’s Artificial Intelligence Research Laboratory (FAIR) in 2013. On the academic side, however, the process of integrating the industrial sector was far from natural. At McGill University, Joelle Pineau shook up habits en by agreeing three years ago to work for Mark Zuckerberg’s social network. “The university just had to have a little imagination to see what it could do in a computer department, as was already done in those of medicine, engineering or business,” explains enthusiastically the director of the learning and reasoning laboratory at the prestigious Montreal university. .
His area of expertise: reinforcement learning, reinforcement learning, a branch of AI dedicated to training agents through a system of rewards and consequences. A system “that anyone with children or a pet knows well. It helps to adapt the behavior. It’s applied to machines,” sums up the algorithm expert and mother of four. Machines learn to make the right decisions from experience, not instruction. Joëlle Pineau’s own research applies to robotics and health.
Ensure the quality of Facebook’s content
At Facebook, the researcher works on vast projects that extend , beyond Montreal’s FAIR to laboratories in Menlo Park, California, facebook headquarters, New York and Paris, the three largest, to Seattle, Pittsburg, London and Tel Aviv. She is involved in team management and research strategies focused on data integrity issues and hate content tracking. Work is being done more and more by learning algorithms.
“We are now able to translate between fifty and sixty languages very quickly. We aim for the hundred languages, those expressed by Facebook users, in order to be able to quickly detect problematic content. The social network’s priority today? Ensuring the quality of content,at the heart of the criticisms of recent years,especially since the 2016 U.S. presidential campaign.
“We had an election in Canada in the fall, it was a kind of preparation for the big elections next November in the United States.” Lessons to be learned to improve the visibility of Facebook users, help them better understand what they read, see and hear,and teach them how to filter better. Criticism of Facebook is justified, pineau acknowledges, but the issues raised are not unique to Mark Zuckerberg’s group. “If Facebook stopped tomorrow morning, these problems would arise on other platforms.. However, Facebook has the staff and the means to find solutions that can be shared with others.»
She admits that she would not have agreed to work in a private laboratory other than Facebook’s. “The work is done in an open science mode, so we publish,we talk about our work in a very open way. In the long run, this benefits all companies that do not necessarily have the resources to improve things. This is close to my heart as a researcher. A sense of sharing that does good.