What We Still Get Wrong About Self-Driving Cars
There’s a lot of hype surrounding autonomous vehicles and when they’ll be ready for the real world. In many ways, the technology is further away from prime time than most people realize, especially in urban areas.
On Tuesday, two industry veterans—Anne Widera, a former Google manager who also previously worked at Uber’s autonomous vehicle group, and David Lu, another former Google manager who was most recently at microchip maker Nvidia—discussed the state of play on a conference call with subscribers of The Information. One of their arguments was that it will be difficult to compare the performance of robot cars to those driven by humans. They also talked about how there’s no breakaway leader in the field—not even Waymo—and how the Chinese market for autonomous cars will proceed differently from that of the West. Below is a partial transcript:
Amir: What are the biggest misconceptions about autonomous driving?
Anne: One common question that I get is, “When are self-driving vehicles going to launch, where are they going to be available?” Kind of like it’s some type of typical tech launch where, you know, Google or Facebook will hit the switch and all of a sudden they’ll be available to everyone. This is a physical product that is often very location dependent, and launch and roll out is going to be a slow thing. So you might have a self-driving car launch this year, like you could arguably say that Waymo has already launched self-driving cars in Phoenix, but that doesn't mean they're broadly accessible by a large swath of population. It might take three years, five years even, before you go from out on the streets to your grandma’s taking a ride in a self-driving car.
David: People drive differently in New York than they do in Boston than they do in Dallas and even San Francisco. And certainly there are commonalities. But being safe in any one of those areas is a challenge unto itself. Every single self-driving launch demo, pilot, whatever you want to call it, with a human driver in the vehicle is substantively not interesting. Having a human safety driver in there means that the robot and the self-driving system are not responsible for safety. So all the news you hear about folks being like, “Oh, we launched a...self-driving fleet in some XYZ town or a retirement home or whatever,” that's not really fair. They launched a fleet of human-driven taxis that sometimes have robots controlling them.
Amir: We’ve not yet developed robots that operate at human-level performance, but what is human level? And what is safe enough to be around humans? Companies don’t really define what it is.
Anne: Classically, you think about incident rates...a certain number of crashes, fatal crashes or injury crashes or minor crashes per million miles or hundreds of thousands of miles. But as David mentioned, everywhere is different, so it’s not like that kind of national level statistic might be applicable to a local area you’re trying to launch in. This is still a very open and difficult question in the space...which is a little bit concerning given how close we are in many ways to getting the technology on the road.
Read More - $ The Information
|
