Solving the Self-Driving Puzzle

"Raquel Urtasun left her position as head of research and development and chief scientist at Uber Technologies Inc.'s self-driving unit earlier this year. Within three months, she had raised $83.5 million in seed funding -- one of the largest first funding rounds by a Canadian startup -- and launched Waabi, a Toronto-based autonomous-vehicle startup that is building self-driving systems for long-haul trucks.

Ms. Urtasun, 45 years old, has specialized in artificial intelligence for 23 years. After working on autonomous-vehicle technology for a decade, she sees an industry that is mired in the frustrations of teaching cars to navigate busy intersections, step-by-step. Uber, which sold its autonomous-vehicle division to Aurora last year, is an investor in Waabi.

Most self-driving companies program onboard computers and then drive their cars millions of miles to test them using real-world conditions -- an expensive and time-consuming approach that doesn't necessarily test responses to rare scenarios, such as a mattress falling off a car. Ms. Urtasun's solution is to use AI and an advanced simulation she has developed. Other self-driving companies often use AI to solve smaller problems, such as teaching cars to parallel park or avoid pedestrians, and then try to build a system from the separate elements, says Ms Urtasun. Waabi's AI tries to solve the whole problem of self-driving at once.

Ms. Urtasun spoke with The Future of Everything about Waabi's focus on autonomous trucks and how driverless technology could reshape cities.

Why are you initially focused on driverless trucking?

Long-haul trucking is where we will first see this technology really making an impact. There is a current shortage of drivers. At the same time, to be a truck driver is one of the most dangerous professions in North America. From a self-driving technology perspective, driving on highways, although very difficult, is much easier than driving in cities. Last-mile delivery, food delivery, it's not clear what that product should look like. Robotaxis are an extremely hard problem, and if you want to really have a market, you have to go to the downtown core, and it's much more difficult. For long-haul trucking, you can automate certain corridors.

How do you want to see autonomous technology progress?

I would love to see this thing get to scale everywhere. Meaning providing the mobility in our cities, particularly to people that don't have the ability to move now. Potentially this could bring more equity to the neighborhoods. For example, oftentimes the price of housing depends on proximity to public transportation. With this technology, maybe we can have a solution for that.

How does that transform the city?

Right now, every vehicle is parked 96% of the time. It means that a very large portion of our cities are essentially parking spots. Imagine what you can do with that [if self-driving vehicles reduce the need for parking]. The future will also be three-dimensional. Meaning it's not just about driving on the surface of the earth, it will also be flying self-driving vehicles.

Could this crowd out public transportation?

Think of redefining public transportation. We've seen in certain towns or certain places, where actually the government has partnered with ride-sharing companies to use ride-sharing as the public transportation.

---

Interview has been edited and condensed." [1]

1. The Future of Everything: The Transportation Issue --- Q&A: Solving the Self-Driving Puzzle
Monga, Vipal.  Wall Street Journal, Eastern edition; New York, N.Y. [New York, N.Y]. 10 Nov 2021: R.8.