Deep Generative Models with the best - Interview with Deep Learning pioneer Yoshua Bengio

Yoshua Bengio’s contributions to Deep Learning would span a book in itself. As an early pioneer in the fast-changing field of machine learning he’s authored three books, over 200 publications, and has been associate editor of the top machine learning and neural networks journals. A true academic, Yoshua co-directs the Canadian Institute for Advanced Research program focused on deep learning, heads the Montreal Institute for Learning Algorithms (MILA) — currently the largest academic research group on deep learning, and continues his research into generative models, biologically inspired learning algorithms and natural language understanding.

Taking his knowledge further into the field, Yoshua co-founded Element AIto help companies develop an AI-first strategy. He champions Canada as the leader in the AI revolution for startups and researchers, encouraging investment in this sector.

Yoshua is undisputedly a key leader of Machine Learning thinking today and we can’t wait for him to tell us more about Deep Generative Networks at AI With the Best online developer conference 29–30th April! For now we got to ask him a few questions.

QYou’ve played a key role in shaping the deep learning landscape and machine intelligence today. How has research process evolved since you started in the field?

That was 30 years ago, so much has changed ;-) There was no internet…

Q. What do you personally find most exciting about deep learning today?

Making progress towards machines which understand how our world works, and feeling these advances coming from a wider and enthusiastic community.

Q. How can less well-known researchers get better recognition for their contributions — often overlooked by journalists?

There is indeed a lot of good research that does not get as much attention, but the issue is not that they are not getting lots of public attention, the issue is that there is so much attention on very few people and results. My concern is thus about how some researchers (like myself) are getting more credit than they deserve in the public’s eye because the focus of journalists on a few individuals and on the romantic idea of breakthrough. One should always remind the public that most of the work done (including what is branded as breakthrough) is one little step at a time, building on lots of previous work from a whole community.

Q. Your students are exploding into the world of research and industry — do you maintain contact with your prodigés and what would you tell some of the first to have graduated under your tuition?

That they should realize how privileged they have been to be part of and contribute to a historic moment in science!

Q. How do you ensure collaboration and smooth organisation in your team?

By encouraging freedom and minimizing institutional rigidity and barriers, empowering students to have a strong word in the organization of the life of the lab and manage some of the lab funding which concerns them directly, by using forums like Discourse and Slack to facilitate exchange and tracking of ideas and debates within the lab, by teaming new students with more senior ones, etc.

Q. How can corporations support research by academic institutions and independent labs?

With sponsorship, data, and collaborations!

Q. What advice would you give to budding AI developers?

There is a lot to learn and to read, but it’s worth it! Understanding what you are doing gives you an immense edge, so take the time to study and try to understand the basics.

Thank you Yoshua Bengio!

You can ask Yoshua your own questions, and explore Deep Generative Networks during his talk at our upcoming AI With The Best, Online Developer Conference 29–30th April.