“Self-driving technology needs AI as its focal point"
![© Natalia Dolan © Natalia Dolan](/image/111288/1108x622.jpg)
© Natalia Dolan
Raquel Urtasun, who completed her PhD at EPFL in 2006, is one of the highest-profile figures in the self-driving vehicle industry. After leading Uber’s self-driving car division, she just founded her own company, called Waabi. She spoke with us about her vision of the future of transportation.
You’re originally from Spain. Can you tell us about your background and how you ended up at EPFL for your PhD?
I grew up in Pamplona – a small town in northern Spain known mostly for its festival. I left to complete my Bachelor’s thesis at Eurecom in Nice, France – a school that, incidentally, was cofounded by EPFL. My initial plan was to go back to Spain after the nine-month project was over – but I ended up never going back. I joined Telecom Paris, and that’s where I heard about EPFL’s pre-doctoral program and decided to apply. I spent six years at EPFL and graduated with a PhD in computer science in 2006.
What memories do you have from your time at EPFL and in Switzerland?
Fantastic ones. First of all, EPFL is a great place for research and has an incredible international community with people from all over the world. I also really fell in love with Lausanne and its surroundings. I made great friends and enjoyed many outdoor activities.
One feature of your career is that you’ve always been involved in both industry and academia. How have you been able to maintain that balance?
I spent ten years in academia after leaving EPFL. In 2011, I switched my research topic to self-driving cars and quickly saw that, in this field, researchers need to have close ties with industry. And research is what I love to do! So, when I joined Uber ATG – Uber’s self-driving car unit – in 2017 as chief scientist and head of R&D, I made sure I wouldn’t have to give up research. Since then, I’ve been able to further my research efforts as a Professor at the University of Toronto’s Department of Computer Science, where I have around 20 students working with me. I spent four great years at Uber and learned a lot, especially from its CEO Dara Khosrowshahi. Now that I have my own company, it’s getting harder to maintain the balance between business and academia. But it should get easier as our company grows and more people join our team.
What are the benefits of self-driving vehicles over cars as we know them today?
There’s a whole list of benefits. The first one is improved safety. Nearly two million people die on the road every year, and self-driving technology can reduce this number significantly. Another important issue is that many people around the world don’t have good transportation options, for either cost or geographical reasons. Self-driving cars could help in that way, too. A lot of people see self-driving vehicles as the antithesis of public transportation – but in fact they could improve transit systems in many ways, such as reaching further distances.
Self-driving cars can also help preserve our natural resources. Today, cars just sit there parked around 96% of the time. Self-driving vehicles can enable shared transportation systems and reduce the total number of cars out there. Another benefit worth mentioning is the effect on real-estate prices. Right now, prices are directly related to location – that is, what you can reach conveniently from where you live. Better transportation options could mean less disparity in that regard, too.
We’ve been hearing about self-driving cars for a few years now. How soon will they be a part of our daily lives?
That’s the million-dollar question! We’re at a point where a lot of promises have been made, but many technological hurdles remain – stemming mainly from the complexity of driving and the sheer quantity of information a driver has to process. Another reason self-driving vehicles still aren’t commercially viable relates to what I would call the “traditional approach” used to develop the technology. In 2004, the US Department of Defense held a competition for self-driving vehicles called the DARPA Grand Challenge. That was 17 years ago, and a lot of the technology used back then is still central today. Self-driving technology is very engineering focused and artificial intelligence is seen as only a small piece of the puzzle – but AI should be the focal point. That’s what sets Waabi apart: we’re building a system that takes full advantage of the power of AI and that could be scaled up at a lower cost.
To get self-driving technology to market, you need regulation on your side as well. How do you think governments can be convinced to allow self-driving cars on the road?
It will be crucial to provide transparency in and accountability for what we’re developing and the challenges we’re facing, so that regulators and citizens can trust our technology. At the same time, self-driving technology is advancing at a rapid pace so we need to maintain constant dialogue with regulators to help them better understand what our technology is – and isn’t. Only through this kind of open dialogue can we get to a point where the technology can be deployed safely and society can reap the benefits I mentioned earlier.
Another obstacle is obviously the cost. What could make self-driving cars more affordable?
There are two types of costs. First, the development costs, and here companies could look at ways to develop the technology in a cost-efficient manner. Our industry is extremely capital-intensive. R&D units need expensive equipment, engineers, simulators and road-testing to check whether their self-driving vehicles are working like they’re supposed to. At Waabi, we plan to address this issue by doing more with less. Our software requires fewer engineers, and our advanced simulation technology eliminates the need for millions of miles of road testing. That means we can develop our technology much more safely and at a fraction of the cost.
The second type of cost is the selling price of self-driving vehicles. Who will be able to afford the technology once it’s available? These costs arise from self-driving cars’ unique features, such as sensor kits and embedded maps that need to be updated constantly. Scaling up production will help keep the technology affordable, but what will provide the biggest boost in my opinion is sharing these vehicles. I think in the end it won’t be about people owning self-driving vehicles but rather using a service.
Another cost-related area where Waabi can help is long-haul trucking. Trucks are very expensive – each one typically costs $150,000 – and transport companies often struggle to find truck drivers. Self-driving technology could lower these costs while making deliveries more efficient.
Do you plan to draw on the self-driving technology that already exists for subways, trams and planes?
The level of difficulty is very different, since those systems operate in a relatively unoccupied environment. That’s nothing like driving in downtown New York, for example! That said, we’ll definitely use some of the lessons learned in those fields, most notably in terms of safety.
Do you think humans will still play a role in self-driving vehicles?
No, it needs to be a complete switch. You can’t have a human taking control of a vehicle after paying no attention to the road previously – that would be too dangerous. Humans will be just passengers.
What kind of information will self-driving cars share with each other? Would that give rise to data protection issues?
Self-driving vehicles need to be able to operate without transmitting and receiving data from other vehicles or road infrastructure. For safety reasons, they can’t be dependent on external devices. However they could share data for mapping or construction sites, for example. That doesn’t have to be vehicle-to-vehicle data transfer; it could simply involve updating a central system. We’ve also been working on systems for vehicles to communicate with each other and share their sensor data so that each vehicle has a better understanding of its surroundings. These are aspects that we’ll keep in mind for future development work at Waabi.
Every time you introduce a new feature, you also create new vulnerabilities. So we’re developing technology to prevent systems from being hacked and to shield against cyberattacks. Data protection is obviously a top priority, and AI can help us there too.
How will our cities and roads need to be adapted to accommodate self-driving vehicles?
Today’s technology is being developed for cities and roads as they currently exist. Changing this infrastructure would be very expensive and time-consuming, and few places would be able to adapt. Beyond self-driving vehicles, we envision a future where all vehicles are electric. Therefore the most important change we will need to make to existing infrastructure is to build many more charging stations.
You completed your first funding round, for an impressive US$83.5 million. Does that put more pressure on you to get your technology to market faster?
Our Series A funding round was one of the largest in Canadian history – and it happened really quickly, in just two months. But I don’t feel that puts us under more pressure. It simply shows that investors understand the importance of our technology and the need for a new approach. We will not compromise on safety to get our technology to market faster, and our investors agree with this approach.
How did the pandemic affect your company in its early days?
Toronto has been in some sort of lockdown since the start of the pandemic, so we haven’t seen our offices in over a year. Starting a new company without actually seeing people and talking to them face to face certainly comes with its share of challenges. But thanks to technology we can work very efficiently, even remotely.
As far as self-driving vehicles are concerned, the pandemic further underscored the need for self-driving technology. Demand for certain goods soared during the pandemic and, as I mentioned earlier, truck drivers were hard to find. So self-driving vehicles could absolutely be a solution for these kinds of situations.
Transportation is an issue that affects all of us on a daily basis. As an engineer working in this field, what kind of social responsibility do you feel you have?
My duty is to leverage the knowledge and skills of everyone who works in my lab or at my company for the common good. That’s the main reason I got involved in self-driving vehicles in the first place. I know too many people who’ve had accidents – I want to improve road safety and have a genuine positive impact.
In previous interviews, you’ve mentioned that it can be hard for women in scientific and engineering fields. Has this improved over the past few years?
When I studied electrical engineering in Spain, half of my class was women. That seemed very natural at the time, but I haven’t seen that since. Over the course of my career, I have known way too many episodes of discrimination. That’s something female engineers still struggle with. Women still have to accomplish ten times as much as men to be recognized. So I’m not sure we can say that things have really changed. Personally, I’m the kind of person who never gives up – the more people tell me I can’t do something, the more I want to do it.
Do you still have ties to EPFL?
Definitely. I’m still in touch with many friends from EPFL and I still speak with Prof. Pascal Fua, who was my thesis supervisor, on a regular basis. These contacts keep me up to date with what’s going on at the School. If I ever decide to move back to Europe, EPFL would definitely be one of my top choices!
What advice would you give EPFL students?
Aim high – your generation can really change the world. The world needs talented individuals to lead it, and EPFL has some amazing ones. And don’t just follow the trend, trust your uniqueness.
What was your favorite place on EPFL campus?
The BC building, which is amazing. The School of Computer Science moved into that building while I was at EPFL, and I have fond memories of its terrace and view.
What’s the best thing about living in Toronto?
The people. Toronto is extremely multicultural, with over half of its residents coming from abroad. Diversity is embraced here, there’s no pressure to change.
What do you prefer, Spanish, Swiss or Canadian food?
I have to say Spanish! I grew up with Spanish cuisine and my parents were great cooks. Now I cook a lot too, and Spanish food is still part of it.
Short bio
2006: Graduates from EPFL with a PhD in computer science
2014: Joins the University of Toronto as an assistant professor
2017: Appointed chief scientist and head of R&D at Uber ATG
2020: Appointed full professor at the University of Toronto
2021: Founds Waabi and raises US$83.5 million in Series A funding