University of Toronto computer science professor Raquel Urtasun is launching a self-driving car startup called Waabi, as my Forbes editor Alan Ohnsman reports. Urtasun created the KITTI dataset, which remains a standard benchmark for robotic perception. She also joined Uber ATG as their chief scientist.
My friends at Uber ATG always had great things to say about her, so I’m excited she’s going out on her own. It’s also a natural result of Uber’s sale of ATG to Aurora, which is a minority investor in Waabi.
Waabi is launching with an $83.5 million funding round. For context, that’s more money than Voyage raised in total, across four years and deployment with real passengers (and safety operators). Waabi should be able to do a lot with $83.5 million dollars, presumably all the more so in Toronto, a lower cost region than Silicon Valley. According to TechCrunch, Waabi already employs 40 people.
Waabi seems likely to pursue a machine-learning first approach to autonomous vehicle development, based both on Urtasun’s statements and her academic background. Even the name, “Waabi”, hints at the goal.
Waabi means “she has vision” in Ojibwe and “simple” in Japanese.
Kirsten Korosec, TechCrunch
Ohnsman reports in Forbes that Waabi plans to focus “heavily on cutting-edge AI tools and less of what Urtasun calls a traditional ‘robotics’ mindset.” Urtasun is a deep learning expert, so I would expect to see a deep-learning-first approach at Waabi, or maybe even a deep-learning-only approach.
“You end up with an approach that requires much less to actually develop. It’s much less capital-intensive and doesn’t require this driving and driving and driving on the road. You get much more automated, fast-paced solutions, and with the ability to come up with much more complex systems.”
Raquel Urtasun in Forbes.com
That focus is reminiscent of Drive.ai, which applied a similar ML-first philosophy to self-driving cars, and also had an academic foundation. Drive.ai eventually ran out of funds and was acquihired by Apple.
Deep learning continues to advance, however, and with Urtasun at the helm, a deep-learning-first approach to self-driving may finally be poised to succeed.
David Silver 