Uber Exec: ‘We Did Not Steal Any Google IP’ – Forbes
On Monday, Anthony Levandowski, co-founder of self-driving-truck startup Otto and now head of Uber’s autonomous vehicle program, was interviewed on stage at the Mobile World Congress in Barcelona. Surprisingly, the interviewer didn’t mention the controversy that’s rocking the autonomous vehicle world: the accusation by Google that Otto, which Uber acquired last year, was built, in part, on patented trade secrets stolen by Levandowski. Before starting Otto, Levandowski was a senior engineer in Google’s self-driving vehicle team.
In October, I interviewed Levandowski for a Forbes cover story about Uber. Part of the interview focused on how self-driving technology has morphed from a seemingly crazy idea spun up by Google into something scores of car companies, and even startups like Otto, are capable of implementing. Unprompted, Levandowski tackled the issue of Google’s IP before we went on to talk about the origins the technology from the DARPA Grand Challenge to teams at Carnegie Mellon University, Stanford and other universities. Here are excerpts of the interview, edited for space and clarity.
Q: How many of the Google team (people) do you have here.
A: We have a bunch. A couple dozen.
Q: Five years ago, when Google first announced (it was working on self-driving cars), it seemed this crazy thing, like a research project. It seemed almost like PR. Today you have teams doing this all over the place. What is different between the various teams?
A: Teams are super different.
Q: Is there something fundamentally different between your approach and Google’s approach?
A: I think every team is totally different so the answer would be “yes.” I don’t want to talk about Google because obviously I came from there. But let’s talk in the hypothetical.
Every team is building the technology from the point of their existing business and the future businesses they want to have with the technology. So the view of those businesses are very different in terms of where you are coming from. If you are an automotive manufacturer and you’re producing all these vehicles out every year and you want to make sure you are still in business when the technology is available and that means your customers have better tech. When you are Uber, what we care about is the customer experience of getting somebody safely, cheaply, efficiently and reliably from where they are to where they want to go. So for us it is really not about the car or the sensor, it’s about the experience. How do you get a nice pickup? How do you drive safely to the destination and get a nice drop-off, and after that have the vehicle disappear?
The types of problems you want to solve depend on where you want to go. We want to go in cities, which is unfortunately the toughest kind of driving there is. For us, we focus on pedestrian detection and everything else versus most people that do highway.
Q: You want to go fully autonomous?
A: Yeah. For us there is no real advantage if we can provide more supply to the network.
Q: At TED in February of 2016, Travis Kalanick talked, before he acquired you guys, and he seemed to indicate that (self-driving cars) are further out than most people think…. Since then, he bought you guys, you launched Pittsburgh, all of a sudden it seems imminent.
A: At the end of the day, it’s going to take a really long time. Even if we make 1,000 self-driving vehicles, that’s a small amount of driver on-boarding for Uber.
Q: There are two questions there. One is how long is the transition to most cars on the road to being self-driving from most being drivers, and when is the first fully autonomous car.
A: The transition is super long tail. Short term, when is it happening? The pressure is mounting. The technology just isn’t ready yet. What we are doing is really trying to focus on that and we’re marching toward that as fast as possible, and I thnk we’ll have something exciting in a near timeframe. The first ones are within five years, for sure.
Q: And that is the science and engineering part of it, not regulatory societal issues.
A: It’s both. There are a lot of places that are going to be welcoming to new technology. And those places are where vehicles are going to deploy first. And if you have places that say, we’re anti-robots, we’re anti-future, then we’re not going to go there first.
Q: Most of the technology, the brains technology, not the mechanical stuff, how much of it came from the Pittsburgh, CMU folks?
A: It’s hybrid. There’s a bunch of it that’s from the Pittsburgh team.
Q: The question is, Google has been working on this for however many years. You guys (Otto) are a year old or something like that.
A: We started in January. How did we catch up?
We did not steal any Google IP. I just want to make sure, super clear on that. We built everything from scratch and we have all of the logs to make that, just to be super clear.
But this is my fourth autonomy stack. I did the DARPA Grand Challenge.
Q: You were part of the original DARPA team that went to Google?
A: No DARPA team went to Google. There were the Carnegie Mellon folks with Chris Urmson and a bunch of his people. Then there was Sebastian (Thrun) with Mike Montemerlo and Hendrik (Dahlkamp) and those people, and then Blue Team from Berkeley with myself and a couple other people. I’m a Cal guy.
So there was DARPA Grand Challenge, one (autonomous driving stack). There was a pizza delivery robot from 2008, where I built a Prius to deliver pizza from downtown SF to Treasure Island. That was the second stack.
Q: I’m not familiar with that.
A: Prior to Google starting the project I was at Google, doing Google Maps. I was just bored and wanted to do robots. So I asked Google “Can I make a robot car to go deliver pizza?” And they were like, “Sure, just make sure it doesn’t say Google,” because they were worried I was going to hit something. So I built that and we delivered pizza for a TV show. It’s online, you can see it if you want.
And that got Sebastian and other people excited and we then started a team at Google to do similar things.
Q: So it was your team that spun up Sebastian into doing (self-driving cars)?
A: I wouldn’t say my team spun up Sebastian. My team showed it could be done. Sebastian got excited, and then Sebastian spun up the team, including myself and Chris and a bunch of others at Google.
So that’s three, is Google. Four was Otto, and so I guess this is five.
How did we get to where we are? We understand what not to do and where not to waste time because we have experience from having tried it before and it didn’t work. And we have experience in trying things that do work, so we are just doing the things that do work and focus on that.
Q: Is the knowledge about self-driving cars spread so much that here you are doing first a startup and now Uber, Google is doing their thing, Elon is doing whatever he’s doing, NVIDIA is doing what it’s doing, guys in Israel, it seems like there’s a lot of people who now know how to do some (of the technology needed for SDC) …
A: Some things, yeah. I think you should, it would be interesting to trace back the lineage of where everybody got started and how much they’ve learned. But effectively before DARPA Grand Challenge, there was really no… there was some defense contracting stuff but it was a bad experience I would say, there was no one trying to do it for real.”
Since the DGC there are people who have been iterating on this and trying and learning and failing. And the teams have merged and un-merged and information has been shared. There is an idea of what’s generally going to work and what’s generally not going to work. So it’s really about now picking an approach and polishing it until it actually works
Miguel Helft is the San Francisco Bureau Chief for Forbes. Follow him on Twitter at @mhelft.