By: Kathleen Sandusky
15 Dec, 2023
AI Canada! Premier global AI conference includes strong Canadian representation
December 15, 2023
Researchers affiliated with Canada’s three National AI Institutes account for nearly 5% of all accepted papers at NeurIPS 2023.
The world may have only recently awakened to the power of AI, but the field of neuroscience-informed computation has been evolving over decades. Many of these advances have their origins in Canada, largely driven by members of the world-leading community of research convened by CIFAR through the Learning in Machines & Brains program and the Pan-Canadian AI Strategy.
Globally, the preeminent conference in machine learning is NeurIPS, hosted internationally every winter since 1986. Presenting a paper at NeurIPS is considered a rite of passage for early career researchers in artificial intelligence. The conference sets the standard of excellence for the field, featuring the biggest breakthroughs and advances from both academia and industry, making it a hallmark event for researchers, companies, and policy-makers interested in charting the course for the societal and economic impact of AI. For 15 of its 37 years the meeting has been hosted in Canada, demonstrating the central leadership of Canadian scientists in the field.
This year hosted in New Orleans, NeurIPS 2023 once again featured strong Canadian representation, with more than 170 accepted papers from researchers affiliated with our three National AI Institutes. This represents nearly 5% of all papers, an outstanding research footprint considering the size of Canada’s population.
An overview of papers accepted to NeurIPS 2023 by researchers from the Pan-Canadian AI Strategy ecosystem:
Amii
Amii Fellows, Canada CIFAR AI Chairs and early-stage researchers contributed 16 papers at this year’s NeurIPS. These range in theme from methods to optimize the efficiency and predictive power of reinforcement learning algorithms, to better performance of AI agents in uses that include robotics, health care and industry. This year Amii challenged their University of Alberta graduate student researchers presenting at NeurIPS to summarize their papers via one-minute plain-language talks, available to the public. These include Christopher Solinas on better algorithms for inferring predictions from imperfect game data, Zichen Zhang on more effective applications of reinforcement learning during a continuous timescale, and Revan MacQueen on the reasons some self-learning strategies work better than others in different game environments, and how this might apply to real-world problems.
Mila
Mila boasts a whopping 96 papers at NeurIPS 2023 from Mila-affiliated authors. The institute also participated in NeurIPS tie-in events that included a talk by Mila-affiliated Canada CIFAR AI Chair Dhanya Sridhar at the Women in Machine Learning Workshop on strategies at Mila to drive better representation by female researchers in the field of AI. Another talk by Mila scientists Anna Richter and Brooklyn Sheppard detailed a project to detect subtle signals of misogyny in language training models. The work is the outcome of a prototype conceived four years ago by trainees at the AI4Good Lab, an annual summer training program for female-identified ML researchers that is supported by CIFAR through the Pan-Canadian AI Strategy. And Climate Change AI, an international non-profit group co-founded by Mila Canada CIFAR AI Chair David Rolnick, hosted a NeurIPS workshop and poster session looking at machine learning solutions for climate change, with Mila scientific director Yoshua Bengio (CIFAR Co-Director, Learning in Machines & Brains and Canada CIFAR AI Chair) co-organizing the event.
Vector Institute
This year, Vector’s researchers secured 65 accepted papers at NeurIPS, including 18 “Spotlight” papers ranked among the top 3% of all NeurIPS submissions. Among these is a paper led by Canada CIFAR AI Chair Richard Zemel that presents a novel framework to assess how equitable an AI system’s decisions will be across an entire population, an important measure in advancing responsible AI. Another Spotlight paper from Canada CIFAR AI Chair Jeff Clune proposes a new method, “Thought Cloning,” that requires AI agents to explain the rationale behind their actions as the “thinking” takes place, improving reliability and explainability. And a paper from a team led by Vector co-founder Raquel Urtasun, a Fellow of CIFAR’s Learning in Machines & Brains program and founder of the company Waabi, proposes a system for better simulating different light conditions to train autonomous vehicles in urban environments, a longstanding performance challenge in the industry.
Read more about how CIFAR and Canada’s three National AI Institutes are advancing the science of responsible AI on our new Pan-Canadian AI Strategy impact page.
Banner image: Image generated using Adobe Firefly. CIFAR has adopted principles for responsible use of generative AI. The Firefly model is trained on Adobe Stock images, openly licensed content, and public domain content where copyright has expired.
Related Articles
-
Canada CIFAR AI Chairs gather in Banff for annual AICan meeting
June 20, 2024
-
Indigenous perspectives in AI
June 18, 2024
-
How does the brain give rise to the mind?
June 13, 2024
-
Racing for cause
June 13, 2024