Beyond the LLM: Raquel Urtasun and Waabi are Redefining Physical AI with a Generalizable, Consequence-Aware Platform
The narrative surrounding artificial intelligence often gets lost in the dazzling buzz of Large Language Models (LLMs). While LLMs are revolutionary, they primarily deal with symbolic data—the language we use....
The narrative surrounding artificial intelligence often gets lost in the dazzling buzz of Large Language Models (LLMs). While LLMs are revolutionary, they primarily deal with symbolic data—the language we use. Raquel Urtasun, the founder of Waabi, and her team have wisely focused on the next, much harder frontier: Physical AI. This is not just about prediction; it’s about action, consequence, and real-time physical interaction.
Waabi’s core innovation is their sophisticated, generalizable AI platform designed for autonomous systems, initially proving its capability in autonomous trucking and passenger mobility. This system moves far beyond simply training on vast amounts of real-world data. Instead, it centers on an architecture that forces the system to 'build abstractions' of what it sees and, critically, to 'reason about the potential consequences' of every possible action in real-time. This ability to model and predict outcomes—a process Urtasun describes as considering 'thousands and thousands of possible things'—is the engineering genius at the core of their offering.
From a technical standpoint, the platform is built around two pillars: the end-to-end AI system (the 'brain') and the simulator. The true ingenuity lies in its architectural design for generalization. By allowing users to plug in any available sensor and adapt the brain for various form factors and sensor distributions, Waabi has created a modular, scalable system. This contrasts sharply with older methodologies that required building bespoke solutions for every vehicle or environment.
Waabi’s breakthrough isn't just in AI capability, but in platform generalization. By building a simulator and a consequence-reasoning 'brain' that can adapt to any sensor array and countless operational variables, they are creating the foundational infrastructure for mass-scale, capital-efficient autonomous systems, moving the industry past data reliance and into universal applicability.
This approach is proving commercially transformative. By emphasizing a capital-efficient, vertically integrated strategy—building sensors and software from the factory floor—Waabi circumvents the enormous cost and time sink associated with collecting massive amounts of human-driven data in the real world. This vertical integration is a huge advantage, making the platform highly adaptable across multiple verticals, from logistics to passenger robotaxis, ensuring maximum scalability and speed of deployment.
Given the increasing global focus on tech sovereignty, particularly in defense and critical infrastructure, Waabi's ability to deliver a robust, generalizable physical AI platform is perfectly timed. It represents a tangible, high-value Canadian deep-tech asset that directly addresses the need for self-reliance in critical automation systems. As Canada positions itself as a hub for advanced, sovereign technologies, Waabi’s work is not merely an advancement; it is a critical component of the nation’s industrial autonomy strategy.