From the very beginning, Anthony Levandowski has been a key player in the push to develop autonomous control systems for commercial vehicles.
After earning degrees in Industrial Engineering and Operations Research from the University of California, Berkeley, Levandowski went to work for Google, where he worked on the company’s Street View technologies. In 2008, he took a major foray into autonomous vehicle development, founding Anthony’s Robots, a startup that focused on developing a self-driving passenger car, as well as other ventures.
After those businesses were bought by Google, Levandowski found himself working for the tech giant once more; this time on the company’s autonomous car program. In 2016, he left Google to found Otto, a company geared to developing retrofit autonomous control systems for Class 8 tractor-trailers. That company was acquired by Uber in late 2016, and Levandowski became head of that company’s autonomous vehicle division in addition to working on autonomous truck technology.
Levandowski’s relationship with Uber soured in 2017 when a lawsuit was filed against the company by Google, alleging that he had obtained autonomous technology trade secrets before resigning to found Otto and had used that data at his new company.
Now Levandowski is focused anew on his initial vision to develop retrofittable autonomous systems for heavy-duty trucks. Late last year, he announced the founding of his latest venture, Pronto AI, which aims to get autonomous technology to fleets in the near future.
HDT Senior Editor Jack Roberts caught up with Levandowski to learn more about Levandowski’s new startup and where he thinks autonomous technology for commercial vehicles is headed.
HDT: You’re back in trucking with Pronto AI, and once again working on retrofittable autonomous control systems for heavy trucks, correct?
Levandowski: We are back in trucking. But the biggest aspect to Pronto AI isn’t the ability to retrofit autonomous technology onto trucks. We’re looking at this more as a gradual, revolutionary, approach to introduce this technology to the commercial vehicle market. Previously, I was looking at this market and assuming we’d simply go straight to Level 4 autonomous systems. But after taking a step back and looking at the 10-year evolution of this technology, it looks like it’s going to take a lot longer to get to those levels of autonomous vehicle control.
HDT: So, this is a more measured approach to bringing this technology to market?
Levandowski: Yes. Long term, there’s no question this is incredibly powerful technology. And it will eventually be deployed in trucking. But that’s going to take some time. And in the meantime, we have all this great technology that is available now with incredible benefits for both fleets and drivers. We’re talking about safety, ease of use, and comfort.
HDT: Long-term angst about autonomous technology aside, I’ve felt for some time that in the early adoptive stages, these systems would prove to be very popular with drivers. Are you seeing any sign of that?
Levandowski: Yes. The overwhelming feedback we get from drivers who have experienced our product is that it is so much more comfortable for them to drive a truck. They can lean back and relax. They don’t have to work as hard in crosswind conditions because the system makes small adjustments to compensate for wind gusts and they can let it do all the hard work.
HDT: One of the more universal worries about autonomous technology is that drivers will lose focus when the systems are engaged. Is that a problem?
Levandowski: We’re still early in learning how to work with autonomous vehicle technology. Still, we haven’t seen that is the case. We know from various news stories that in passenger cars, there are a lot of drivers doing other activities behind the wheel – on their phones and other things they shouldn’t be doing. But professional drivers seem to view this as technology that makes their jobs a little easier to do. But they understand that they’re still on the job and responsible for the safe operation of that vehicle.
HDT: Some news reports suggest that you are “frustrated” with the slow pace of acceptance of autonomous technology in trucking. Is that an accurate reflection of your views?
Levandowski: Well, I still think it’s going to happen much faster than many people think. But it’s just not going to happen as fast as many of us early developers initially thought. And I am frustrated. Because this technology has so many benefits to offer the trucking industry. With technology, it’s easier to overestimate it’s acceptance and impact in the long term, and underestimate those factors in the short term. So now, I believe it’s going to take awhile before we see fully autonomous trucks making long-haul runs down our highways. Because that is an incredibly complex technical problem to solve. But the benefits this technology offer can happen early on. I see all these technologies becoming more commonplace in passenger cars. But they have not yet made their way to commercial vehicles. At Pronto AI, we’re trying to accelerate those trends so we can get those benefits to fleets and drivers and allow them to use technology that is available now.
HDT: Is one of the stumbling blocks holding back more advanced autonomous vehicle systems the lack of vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) capabilities nationally?
Levandowski: There are a couple of different ways to look at that. Yes – it would definitely be better if every vehicle on the road was broadcasting where it was and where it was going, and every structure near a road was broadcasting information on traffic flow and things like that. It would be fantastic. But we have to look at how we can spend tax dollars now. And if we would simply repair and maintain our roads better, make them smoother and install clear lane markers, that would be a huge step forward as well. In fact, right now, I’d say that’s a higher priority than developing V2V and V2I. Because if you improve road conditions, existing autonomous technology already works very well on its own.
HDT: Do you see the tipping point for serious investment in V2V and V2I as dependent on the number of autonomous vehicles on the road? That is to say, why do it now if there aren’t that many vehicles that can take advantage of the technology?
Levandowski: I totally agree with that. There is just not yet an overriding need for those systems yet. And we’re not yet clear on what the future need for those systems will be. Once we have a significant number of self-driving vehicles on the roads, it will be a lot easier for us to predict those needs and build accordingly.
HDT: So, in the meantime, your focus will on safety?
Levandowski: Yes. And, as I said, it’s hard to calculate what the rate of acceptance for this technology is going to be. But as autonomous technology matures, I think it will be really interesting to look back in five or ten years to see how far we’ve come. And I think it will become increasingly commonplace -- people will say, 'Wow. I can’t believe we used to make and drive trucks without this standard of safety equipment on them!' Sort of like people looking at a truck without ABS on it today. These systems are in the pioneering stage now. But they will be standard equipment in the future.
HDT: Can you give us any specifics on what aspects of the technology Pronto AI is working on at the moment or what news we can expect in the near future?
Levandowski: We’re doing a lot of testing in different locations and in different weather conditions. And we’ll have some announcements on that work soon. Also keep an eye out for some news on who our first customers will be and how we are working with them.
HDT: You recently undertook a cross-country drive in a car equipped with your autonomous technology engaged 95% of the time. What was that experience like? Did you learn anything that surprised you?
Levandowski: Let me correct you: That cross-country drive was 100% under autonomous control. That’s why it was so significant. From the time I left San Francisco to the time I got to New York, I did not make any control inputs on the brakes, accelerator, or steering – not for fuel stops or overnight stays, or anything. There was no human intervention at all during the drive. And we think that was an important milestone in developing this technology. We wanted confirmation that the technology stack we’re building was capable of that feat. We knew our software would update itself as we made our way along the drive. And we know we need to repeat that process many thousands of times to get the functionality to where we want it to be. But we wanted to make sure we could make such a long trip using only camera systems and radar. And we did.
HDT: On a different front, there's been an uptick in news stories, magazine articles, and television shows voicing concern about the rise of robots and autonomous technology and the impact these will have on jobs. Do you have any thoughts on those concerns?
Levandowski: These changes are actually happening slowly and we will see additional changes coming. We can see them now, and we’re talking about this – and we don’t have the first autonomous vehicle that can operate commercially on the road yet. There will be a much slower rollout of these technologies than others that have radically altered our world. This technology is going to come into common use much slower than the internet, or any entirely new business that sprung from it did. Right now, we are focused on safety. And it will only be when autonomous vehicles have been proven to be safer than humans when we might start to see who this technology will replace people behind the wheel. But we think that even then, autonomous technology will create more jobs than it takes, because trucking is only going to grow in the future and the demand for the services it offers is going to expand greatly. In the future, anything you buy online will be no more than two hours away from showing up on your doorstep. And that level of efficiency can’t happen with people alone. Autonomous technology will be vital for achieving those goals. And there will still be great demand for human drivers. So, I think it will be a very long time before we see more trucks driven by robots going down the road than trucks driven by humans.