字幕列表 影片播放 列印英文字幕 Every year, more than a million people in the world die in road traffic accidents. And many of those fatalities are caused by us, by the mistakes we make as drivers. But there's a solution. Okay we're ready. Okay, engaging. And it's to relinquish control of the wheel to a computer. We've dreamed about it for decades, and now we finally have the technology to get us very close. The rest is up to these two. They're teaching autonomous cars how to drive. My name is Daniela Landey. My name is Steven Lin. And we're autonomous vehicle operators. Daniela and Steven work for a start-up called Aurora, and they're part of a team of about two dozen specialists who drive the company's fleet of self-driving cars on the roads of Pittsburgh, Palo Alto, and here in San Francisco. Aurora's software learns from their expert example, so to get this job, they had to be excellent drivers to start with. And to become even better, they then went through six weeks of training. Today, I'm here at the Sonoma Raceway to experience just a small part of that training and to see if I have what it takes to do this job before I hop in the back of a self-driving vehicle myself. Nobody actually, in my opinion, knows how to drive a car. You all drove here today, so I'm not, don't, I'm not bashing you, right. But do you know how to handle a car when a car does something out of the ordinary? That's what today's all about. All right, let's go have fun. Among the first drills was emergency braking. All right, Daniela, come on down. Since Steven and Daniela have already completed this kind of training, they're here today to brush up on what they already know. Excellent. And to show me how it's done. Good job! All right, next car, full throttle all the way to me, don't brake early. Full throttle, go, go, go, go, keep going guide the brake. Go, go, go, go, go, stop And then it was time for a series of high speed lane changes. Godspeed. Imagine you are forced to suddenly swerve on the highway to avoid an obstacle? Get that brake timing down. Release, release, release, there ya go, awesome. All right, here go, 50 right lane. 50, right. 50, right lane, that's not 50. Go, go, go, go, go, go, go, go, go, go, go And turn. And for an especially tricky challenge, the instructors had us recover from a skid. Show off. Oh, oop, Oh my God! Ah! Let's just say it's a good thing no self driving car will be learning from me anytime soon. The next week I went to go see Daniela and Steven at their office in San Francisco. Come on in, welcome. Founded in 2017, by engineers from the early autonomous vehicle projects at Google, Tesla, and Uber, Aurora is developing technology to power self driving cars. And so far, it's signed deals with automakers including Volkswagen and Hyundai. Daniela and Steven have invited me here to be the very first reporter to sit in the back of one of their self driving cars. It says ready up here. Ready? We are ready. Okay, we're engaging. Now we are in auto. Oh, whoa. Yep. Tracking the car ahead, stopping. So the car right now is stopping on its own? It's stopping on its own, its thinking, its making decisions. This is really cool, getting to see what the car sees. [Steven And Daniela] Yeah. What are these yellow boxes? The yellow box right there is, that's a bicyclist. That's a cyclist, oh cool. So the car recognizes that this is a bicyclist and if you see these kinda blue boxes, that means the vehicle recognizes it as a vehicle. And the red boxes are pedestrians? Pedestrians, exact, you're hired. Aurora does a lot of its testing virtually within the safety of a simulation. But to teach the system how to react to all the crazy things that happen in a city as nuts as San Francisco, you need to take it out in real life. What's the biggest misconception your friends have about your job? That we just sit behind a wheel and do nothing. Auroroa's operators take turns as pilots and copilots of the cars. Today, Daniela is the pilot. I'm looking at everything, I'm thinking about everything, planning for everything. If the car decides to bail out in the middle of the intersection, what am I gonna do? See how her hands are ever so slightly touching the wheel? Her foot hovering just above the brake? She's ready to take over in a fraction of a second, whenever she or the car senses she needs to. And that's a bailout. Oh. And I took over. The car saw something that was not accurate and it gives me control. As Daniela's copilot today, Steven's job is to keep an eye on his laptop that shows what the car see. Getting a false positive on the left, flickering. He alerts Daniela to the things she might not see with the naked eye and prepares her for what the self-driving system is about to do next. It's gonna want a left lane change. Okay. As the copilot, Steven is also taking notes on when Daniela needs to intervene. Later on the engineers will pour over this data, so they can figure out what went wrong. This is all part of the painstaking process of teaching a computer how to drive. It's brain, it's a little baby brain that learns-- That you are nurturing, step by step. Um, hm, yeah, I call all of them my baby robots. I swear, because we're teaching them. Each operator at Aurora has responsibilities outside the car, too. Steven assigns operators to drive the routes that engineers needs them to drive. And Daniela liaises with the team that builds the maps the vehicles rely on. Because driverless technology is so new, operators come from all kinds of backgrounds. And the job doesn't require a special degree, or prior industry experience. Daniela was a surgeon in Mexico and worked for the fire service after she move to the states. And Steven was a military police sergeant in the army. Operators at Aurora earn about $30-$40 an hour. And as full time employees their benefits include health care, equity, and fitting of the tech company stereotype, free lunch. You know you guys are teaching cars how to drive and if you're successful that is going to replace a lot of jobs out there, Uber drivers for example. That's millions of jobs, how do you think about that? We've had interactions where people tell us you're taking jobs away from people, but what I like to say is, imagine you're living in a world where you don't have to worry about drunk drivers, you don't have to worry about people who are texting on their phones, and, you know, just being reckless. Yeah, we might lose some cab driver jobs But will it save millions of lives? Right. In the long run. In the future, our cars aren't going to need the kind of intensive teaching that Steven and Daniela are doing now. But when it comes to the most high stakes life or death responsibilities that we'll soon entrust to computers, we're gonna want human experts to be vetting those capabilities for a very long time. We're still gonna be testing, for perhaps new features, new avenues, new ways to use self-driving technology in other type of vehicles, so I don't believe this job will disappear anytime soon.