The self-driving software firm Embark Trucks said Wednesday it plans to go public in a deal that would raise more than $600 million and value the company at more than $5 billion. Then there’s Alphabet’s Waymo, which just raised $2.5 billion in fresh funding as it tries to expand its self-driving taxi fleet outside of Phoenix. And General Motors has increased its planned investment in autonomous vehicles.
Chris Gerdes is a professor of mechanical engineering and co-director of the Center for Automotive Research at Stanford. He’s also a safety adviser with Ford’s autonomous-vehicle division. I asked him where all this money is going. The following is an edited transcript of our conversation.
Chris Gerdes: If you think about everything that you do when you drive a car, everything from just making sure you can maintain your position in the lane, to thinking about how you predict the actions of other road users around you, thinking how you slow down automatically if you can’t see well because of rain on the windshield, all of these things are technical challenges that require different skill sets and require different levels of engineering development. So, you’re not necessarily going to see a single advance from a bunch of smart people in a garage lead the field forward. It’s going to take a large team of engineers working together, and they’re going to require a lot of funding.
Amy Scott: It should be said that a ton of money has already been invested in this. How far has that gotten us?
Gerdes: It’s a good question, and I think that’s one thing that the industry is asking themselves is, how close are we to the finish line? And the reason that’s a hard question to answer is that there really is no fixed list of requirements of what are the things that you need to do to develop a self-driving car. Normally, engineering works in terms of you set a list of requirements, and then you develop a product to meet those requirements. But in fact, the requirements are still being discovered as companies field their automated vehicles. So really, the goal posts have been continuing to move.
Scott: We’ve talked with other experts in this space, who say no amount of money is going to teach neural networks to drive like people, and we should stop throwing money at this issue and take a whole different approach. Do you think this is a problem money can solve?
Gerdes: I do think these are solvable problems. Now, they’re challenging problems, and they have to be solved individually. And then you have to ensure the safety of everything when it’s integrated into a whole. And when you look at what affordable transportation that is safe and accessible could do for society, I think this is actually money well spent.
Scott: Last month, our regular host, Molly Wood, spoke with professor Missy Cummings about how neural networks learn to drive. She gave the example of showing it millions of images of a stop sign, but that if the car sees one covered in kudzu, it might not recognize that. How do you solve that problem?
Gerdes: I think there’s a couple of ways of going about solving recognition problems. One is simply to give the system more data. So as cars are out there, cruising through the world, they’re seeing a lot of images and cameras, and so you get ever more data that you can use to train neural networks to recognize. But I think the other thing is to realize that automated vehicles are not humans, and they don’t have to do everything exactly the way a human does. So simply, you could program the car to know that they are at an intersection and there’s a stop sign there. So they should come to a stop whether or not they see a sign.
Scott: I read a piece from Bloomberg, back in 2015, you ended this piece with a line about how with any new technology, there’s sort of a peak-hype phase and then the trough of disillusionment. And at the time, you thought we were still in peak hype. I wonder where you think we are today.
Gerdes: I think we’ve certainly moved past peak hype. People are getting a little bit longer in their time horizon predictions about when these vehicles are going to become commonplace. And so, I think we may be seeing some disillusionment set in such as your question about whether this is still a worthwhile area to spend money in. So, I think we may be going through that disillusionment at the moment, but there’s tremendous progress being made.
Scott: You don’t sound disillusioned.
Gerdes: I’m not disillusioned, actually. I mean, these are hard problems. But in fact, the more that I work with people and see how far the industry has come, actually the more confident that I am that we’re going to get to a situation one day where people will be able to hail an automated vehicle and have it take them about their daily activities.
Related links: More insight from Amy Scott
If you missed it last month, here’s that interview Molly did with Missy Cummings, who is the director of the Humans and Autonomy Laboratory at Duke. She’s pretty skeptical that truly self-driving cars are coming anytime soon. Though Tesla sells a $10,000 upgrade option it calls “full self-driving,” its cars are not actually fully autonomous. The California Department of Motor Vehicles is looking into whether the claim violates state regulations. According to Bloomberg, highway safety regulators have opened 30 special crash investigations since 2016 that potentially involved Tesla’s autopilot. Several of those accidents were deadly.
Cummings said all this money flowing into trying to get to fully autonomous vehicles is going into a pit that so far has no bottom. And the driving public doesn’t seem that much more optimistic. AAA’s latest automated vehicle survey found that only 22% of people think carmakers should focus on the technology. Eighty percent said, how ‘bout improving existing safety systems like automatic emergency breaking and lane detection first?
Embark Trucks is just the latest self-driving truck company to announce plans to go public. Its CEO, Alex Rodrigues, by the way, is 25 years old. TuSimple, backed by Volkswagen and Goodyear, among others, debuted on the Nasdaq in April to the tune of almost $1.4 billion. And rival Plus.ai plans to begin selling shares later this year. Like Embark, Plus is going public by merging with one of these special-purpose acquisition companies, or SPACS, that are all the rage. According to The Wall Street Journal, the SPACs have raised nearly $110 billion this year. Last year’s total of $80 billion was more than the total raised in the almost 30 years SPACs have been around.