After weeks of speculation following Sanja Fidler’s departure from Nvidia, the leading AI researcher has announced plans for her new project. Fidler, who is a co-founding member of Toronto’s Vector Institute and an associate professor at the University of Toronto, has launched Veeda, a startup that aims “to unlock the scaling moment for physical AI,” per its website. In essence, the company, which has bases in Toronto, Zurich, San Francisco and Singapore, is creating virtual models of the world around us so that robots can learn how to successfully navigate and interact with people, places and things IRL. Think of it like the “Matrix,” but for robots.
Compared to chatbots and much of the AI that has become embedded in the fabric of our daily lives, developing AI for robotics is tricky because of data scarcity, explains Florian Shkurti, an assistant professor of computer science at UTM and the director of U of T’s Robot Vision and Learning lab. “For LLMs, there’s readily available data on the internet. And these companies have scraped hundreds of millions of books,” he says. “There’s no internet-scale data for robotics. Even with self-driving cars, there’s only a limited number of companies that have access to a continuous data stream because their cars are constantly being deployed on roads.”
