More than a decade into development of self-driving cars, widespread deployments remain elusive.
Raquel Urtasun hopes to soon accelerate the plodding pace of progress. The former Uber Advanced Technology Group executive and computer science expert has started her own self-driving technology company, Waabi, that's already working on artificial-intelligence advances.
The company made its first public pronouncements early Tuesday.
"Commercial deployments are limited to small operational domains, and in my opinion, that's due in part to particular approaches that have not realized the power of AI, and instead rely on complex and time-consuming methods," Urtasun told Automotive News. "So it's very costly to scale. I believe a radical new approach is necessary."
Founded three months ago in Toronto, the company announced a Series A funding round of $83.5 million. Waabi will first target the self-driving truck market. But the company's self-driving system will be versatile enough to be fitted on a variety of vehicle platforms.
If that sounds like a blueprint established by the likes of Aurora, it's notable that Aurora is among Waabi's first investors. Aurora recently completed its acquisition of Uber ATG, where Urtasun had worked since May 2017 first as head of the company's Toronto operations and later as its chief scientist and head of R&D.
"[Aurora CEO] Chris [Urmson] and his team really believe in the technology we are building," Urtasun said.
The funding round was led by Khosla Ventures, with further participation from Uber, Radical Ventures, 8VC, OMERS Ventures, BDC Capital's Women In Technology Venture Fund, and others. Terms were not disclosed.
Urtasun remains a University of Toronto computer science professor and co-founder of the Vector Institute, a nonprofit research corporation founded in 2017 to study AI and deep learning, that has received funding from the Canadian government and private sector.
Waabi, a name selected from the language of the Objibwe group of indigenous Canadians, translates to "she has vision." In terms of advancing AI progress in the company, Urtasun envisions borrowing from the best of a traditional rules-based approach to developing a self-driving system and from one that depends upon deep learning.
"We need to take the lessons of the last few decades and build a unique and differentiated approach, in particular with new algorithms that leverage deep learning and probable inference," she said. "It needs to be end-to-end trainable and we can trace back to why the system decides to do very specific maneuvers. It's a new generation of algorithms."
