Raquel Urtasun is Taking Self-Driving Tech in a New Direction

After heading up Uber's self-driving division, Raquel Urtasun is bringing a new simulation-first model to autonomous driving.

If you ask Waabi founder Raquel Urtasun, the road to commercial self-driving vehicles is paved with AI.

“As a scientist, I’m always attracted to the most difficult problems that require innovative solutions,” Urtasun told BrainStation Magazine. “Self-driving is a challenging field that has captivated me for years because once solved at scale, it will change the world as we know it.” 

With Waabi, Urtasun may just be overseeing that global change herself. The Toronto-based self-driving car startup launched to the public early last month, catching many in the industry by surprise after earning more than $83.5 million in Series A funding led by Khosla Ventures, with additional funding from high-profile backers including Uber. The investment brings the commercialization of autonomous vehicles a giant step closer to reality, not least of all due to Waabi’s proprietary AI tech, which is backed by Urtasun’s two decades of experience in the field.

“The more time I’ve spent in this industry, the more certain I’ve become about what it will take to actually bring this technology to life,” Urtasun says. “Our AI-first approach is unique in the industry and, I believe, key to commercializing this technology.”

A history of self-driving expertise

Urtasun founded Waabi after four years behind the wheel at Uber’s Advanced Technologies Group, where she oversaw the company’s self-driving car lab. Urtasun is also a co-founder of the Vector Institute for AI, the high-profile Toronto-based AI research hub that also names AI pioneer Geoffrey Hinton as its chief scientific advisor. 

During her tenure leading all things self-driving with Uber, the company invested hundreds of millions of dollars in autonomous tech, with nearly $500 million being invested in self-driving and flying car R&D in 2019 alone.

While at Uber, Urtasun realized the need for commercially viable self-driving vehicles — and the fact that, currently, the autonomous vehicle industry is not leveraging AI tech nearly as much as it should be. Rather than apply her idea for an AI-first approach to driverless vehicles to Uber, Urtasun told the University of Toronto Magazine that “if you really want to change technology, the best way to do it is to start a new company.”

Urtasun standing in front of a self-driving Uber ATG car she helped develop.

A new approach to autonomous driving

The “AI-first approach” is, indeed, what sets Waabi apart: “If you look at the more traditional approaches to building self-driving technology, AI is simply one piece of the puzzle,” Urtasun says. “At Waabi, AI is at the center of the solution.” 

The company’s AI system’s advantages are threefold: it is interpretable, end-to-end trainable, and capable of complex reasoning. Interpretability means Waabi’s AI produces representations that are explainable, which allows the company’s engineers to trace back the reasons why the AI system made a particular decision. End-to-end trainability means that the entire self-driving system can be learned from data, which speeds up development, and its complex reasoning capabilities mean Waabi’s AI is able to closely mimic the human decision-making process.

“Our AI-first approach is unique in the industry and, I believe, key to commercializing this technology.”

Raquel Urtasun, founder and CEO of Waabi

These facets potentially put Waabi’s AI tech leagues above competitors. When combined with the fact that Waabi’s technology is being developed entirely in simulation, the model potentially circumvents the massive safety concerns that have plagued the self-driving industry so far. With Waabi’s technology and approach, there is less of a need to log millions of miles driving with a physical car on the open road—an approach her previous employer Uber focused on. Instead, those test miles can be simulated to the same degree of safety and scaled exponentially.

The AI tech she’s pioneering with her team at Waabi will first focus on trucking. Highway driving, while challenging, is an easier problem for Waabi to solve than city driving, and by focusing on the trucking industry, Waabi will be able to fine-tune its AI tech while also solving the issues of safety and chronic driver shortages that plague the trucking industry. Urtasun points to Waabi’s closed-loop simulator as another example of how the company is leading its competitors. 

“Other companies are using simulation technology, but at scale, they test only motion planning,” she says. “[Our] closed loop simulator … enables us to test at scale both common driving scenarios as well as safety-critical edge cases. This is a game-changer, as we significantly reduce the need to drive in the real world in order to understand how our system performs.” 

“Furthermore, our autonomy system can learn to act in situations we might have never seen before by training our AI-first approach in simulation.”

Think of it like this: if a self-driving AI company’s first approach to training their system is through driving on real roads and logging that data, it may only run into certain situations (for example, a low-visibility blizzard at night) a handful of times. With Waabi’s system, that scenario can be replicated and trained thousands, if not millions of times, without wheels ever touching the asphalt.

Commercializing self-driving

Waabi will be entering a highly competitive field here, as leading AV companies including Google’s Waymo, Aurora, and car manufacturers Daimler and Volvo are all competing to be the first to introduce autonomous tractor-trailers to the highway. 

Urtasun says she believes Waabi’s AI-first approach will quickly move the company towards a commercial product but acknowledges that there are a number of external factors — including regulatory confidence in Waabi’s tech, in addition to consumer buy-in — that will determine just how quickly. 

“I intend to lead the industry in transparency and accountability.”

Raquel Urtasun

“Commercialization requires regulators and communities buying in, as you mentioned. People need to believe the technology works and is safe,” she says. “While we’re building, I intend to lead the industry in transparency and accountability, so society is ready to move forward with us by the time we’ve reached our milestones.”

Waabi currently has over 40 employees in Toronto, and a few in the Bay Area, with plans to grow the team in the coming months, led by Urtasun’s conviction that self-driving tech is “one of the most exciting and important technologies of our generation.” And, really, it’s all in the name: among Canada’s First Nations peoples, “waabi” means “she has vision” — a translation that perfectly encapsulates Urtasun’s mission for the future of transportation.