AI can be Remarkable.

Vector Institute’s second annual Remarkable 2025 conference showcased the strength and potential of Canada’s concentrated AI ecosystem, where world-class research capabilities, industry innovation, and a growing talent pool converge.

The two-day conference demonstrated this convergence through its structure: beginning with presentations on pioneering research and specialized technical sessions on day one, followed by dynamic industry panels that shed light on how these research breakthroughs lead to innovative AI implementation on day two.

As a central hub connecting academia, industry, and public institutions, Remarkable 2025 demonstrated how physical proximity and close collaboration create unique advantages for innovation and implementation. While speakers acknowledged Canada’s current adoption challenges, they emphasized the country’s strong foundation and readiness for leadership in AI, pointing to Vector’s successful industry partnerships and breakthrough research achievements.

Throughout both days, the community’s energy and commitment to greater ambition highlighted how Vector’s network of researchers, industry partners, and government stakeholders is positioned to accelerate Canada’s AI advancement.

1800+

attendees

41

countries represented

800K+

impressions

6

media outlets

32

large enterprise companies

19

thought-leading speakers

Day 1 – March 4

Research Symposium

At the Frontier of AI Innovation

Remarkable 2025’s first day showcased significant advances at the frontier of AI innovation through three keynote presentations, spanning the spectrum from theoretical breakthroughs to real-world applications.

Nitish Srivastava of Vayu Robotics demonstrated their camera-based autonomous delivery vehicle designed for small business accessibility, while Vector Faculty Member Kelsey Allen shared crucial insights into physical reasoning AI and the challenges of error handling in predictive models. Ruslan Salakhutdinov rounded out the keynotes with his presentation on the VisualWebArena benchmark for multimodal AI agents, offering valuable perspectives on the future of human-AI collaboration and the current state of multimodal AI capabilities.

These presentations collectively highlighted a crucial moment in AI development, where researchers are actively addressing both the tremendous potential and current limitations of AI systems. The research presented demonstrated clear progress in making AI more practical, accessible, and capable of handling real-world complexities.

“With agentic environments, the task is not precise. The future will involve humans working together with agents.”

Ruslan Salakhutdinov

Professor, Carnegie Mellon University and VP of Research, GenAI, Meta

Breakout Sessions

These keynote sessions were followed by a series of breakout sessions, showcasing the depth and breadth of current AI research while highlighting crucial concerns about AI’s future impact.

• AI Safety focused on catastrophic risk assessment and introduced Vector’s State of Evaluations report and its 16-benchmark evaluation system for AI models.
• Computer Vision explored advances in generative models, particularly image generation capabilities, and their health care applications through multimodal learning.
• NLP addressed critical limitations of large language models, training practices, and increasing concerns about AI bias in newer models.
• ML Security & Privacy explored PETs and cybersecurity solutions, with emphasis on federated learning applications and methods for secure ML model auditing.
• ML Theory presented research through lightning talks covering comparative learning and neural network architectures.

Day 2 – March 5

AI in Action

From Lab Breakthroughs to Market Impact

Remarkable 2025’s day two headline events brought together leaders from public and private sectors to address the practical challenges and opportunities in AI implementation, beginning with a crucial discussion on public sector innovation. Tony Gaffney and Dr. Laura Gilbert emphasized the urgency of AI adoption in government services, with Gilbert asserting that “one of the greatest risks is not doing anything at all.” Their session highlighted the potential for AI to enhance government services while addressing important considerations around regulation and implementation.

The day’s discussions turned to Canada’s AI adoption landscape, where a striking statistic emerged: 73% of Canadian businesses have not yet considered using generative AI. Industry leaders from major organizations including Google, RBC, and Unilever shared practical insights on overcoming implementation challenges, while representatives from CIBC, KPMG, Bell, and EllisDon explored the evolution of Canada’s AI talent landscape. In health care, leaders from Roche, Tali AI, and Hero AI emphasized the critical importance of trust and privacy in AI adoption.

The day concluded with strong calls for increased ambition in Canadian AI adoption, with TD Bank’s Maks Volkov highlighting the unique advantage of Canada’s concentrated AI ecosystem, and Vector’s Deval Pandya emphasizing that “Canada has everything necessary to lead in AI. We need to be more ambitious, and find ways of working together so we’re not just replicating each other.”

“One of the greatest risks is not doing anything at all.”

Laura Gilbert

Head of AI for Government, Ellison Institute of Technology