Roundtable host: Ryan Grant

AI is fundamentally changing the way we work and the future of the way we interact with technology. A huge part of this change is physical AI, the intersection of AI with robotics is at the forefront of revolutionary technology that will impact lives around the world. With physical AI comes the imperative for compute to match the needs of technology and ensure our society continues to remain under our control.

This working session will develop a blueprint for determining how much computing power is needed for physical AI in the future. The focus of this roundtable will be on supercomputing resources needed for physical AI. Our goal is to connect the robotics community and the supercomputing community for them to work together on future sovereign computing major projects and ensure that Canada remains in control of its technological future. 

This session's mandate is to establish the first estimation system for physical AI needs in Canada and develop an approach to meet those needs supported by the private and public sector.

Agenda

• The Challenge (15 minutes): We will first discuss the challenges of supplying the supercomputing resources to train and improve physical AI systems. We will discuss the challenges of estimating the required compute and set the stage for subsequent discussion. 
• How Much Compute and How do we Build it? (60 minutes): This part of the roundtable will be a discussion of how we can go about estimating the needs for high-performance compute and estimate the datacentre footprint required for a Canadian Sovereign supercomputing platform. We will discuss how we can create a system for estimation and then discuss the strategies that can be used to fund and deploy the needed compute. We will develop two deliverables from this portion of the roundtable and estimate of the compute needed and two or more strategies for making supercomputing for physical AI a reality in Canada. These strategies should represent Canada taking a leadership position in physical AI compute on the world stage.
• Action Plan & Partnership Development (15 minutes): This segment will focus on the partnerships needed and the action plan for concrete further work in making the issues discussed at the roundtable a reality.

Outcomes

• A Method for Physical AI Compute Requirements Estimates: We will produce an estimation method for calculating how much compute we need to make Canada a Physical AI compute global leader.
   

Collaborations and Strategy: As this area is new, we will forge new collaborative partners and develop a strategy to work together to make physical AI possible in Canada.