An autonomous car parked outside the headquarters of semiconductor company Ambarella in Santa Clara, in March 2018. (Lauren Hanussak/KQED)
One day we may tell our kids or grandkids about the first time we ever saw a car drive down the street without a human behind the wheel. Today in California, we are a little closer to that milestone. As of April 2, the DMV can issue permits to test driverless cars on public roads. Unlike previous testing, the new permits will not require cars to have a person behind the wheel.
KQED's Brian Watt spoke with Nidhi Kalra, senior information scientist at the RAND Corporation and an expert on self-driving cars, about what this means for our roads.
Watt: How far are we, do you think, from seeing cars regularly go by on the street without a human driver?
Kalra: We're not far at all. I would say by 2020, which is around the corner, we are going to see self driving cars. We're going to look over at the car next to us and there will be no one behind the wheel, there may not be a wheel.
Watt: Let's talk about the significance of April 2nd. Is today a big change in policy, a big change in opportunity?
Kalra: It is a big change. You know, until now there's had to be a test driver behind a vehicle even if it's driving itself. And now that changes. And that matters for a couple of reasons.
The big benefits of self-driving vehicles are that there doesn't need to be anyone paying attention. Not even anyone there. To get that [technology] on the road for consumer use we have to get that on the road for testing. That's what this allows.
It's a mixed bag, though. On the good side, we're getting towards technology that many people think is going to hold a lot of promise. On the negative side, there's a lot more risk in having a vehicle that doesn't have someone paying attention.
Watt: As we look at this emerging technology, what are the safety pros and cons of driverless cars?
Kalra: Let's do a little context-setting. Every year there are 40,000 fatalities on our roads in the United States alone. So it's a public health crisis that we sort of take for granted. Over 90 percent of those crashes involve some kind of human error. These vehicles are never tired. They're never drunk and never distracted. They don't make the routine mistakes that you and I might make all the time. That's their potential in terms of safety.
But there's a risk. One is that they might make mistakes that we would never make or they might not be able to solve all of the safety issues we have. And then there's issues of cyber security. Will these vehicles be hacked and will we have simultaneous crashes? Or will a bug in the program cause them to do crazy things? So just because we're terrible drivers doesn't necessarily mean that the vehicles will be great, but there's a lot of room for improvement.
Watt: What are the pros and cons of having a driver testing a self-driving car and not having a driver in them?
Kalra: The obvious pro is that there's a backup safety driver because this technology right now is by many measures like teenage drivers. They're pretty good. They're not perfect. And they've got a ways to go still, so that backup driver is this professional who is going to jump in if something goes wrong.
But eventually you've got to let that technology be on the road without a backup driver because that's the way we want people to start using it. The public is interested in this technology. So this is a step towards that. It's not without its risks.
Watt: What's the coolest and most interesting thing you've seen so far in this field?
Kalra: You know the coolest part of it is the brains of the self-driving cars, the stuff you can't even see. It's making sense of the world, figuring out what's going on and deciding what to do. That requires a level of computation and algorithmic sophistication that, a few years ago, wouldn't have been possible. So this technology couldn't have happened 10 years earlier, and that's really exciting that we have it now.
Watt: Now at any intersection in San Francisco, you're seeing all kinds of different vehicles. You've got a lot of fire trucks passing through, city buses, Muni, little scooters. These brains powering these cars, are they ready for all of the nuances of driving on a city street?
Kalra: It's not clear that they're ready yet, but there's no way to get them ready without getting them on the street. There's so much diversity on our roads, not to mention the pedestrians and the dog walkers and squirrels crossing. There's so much to take in but that can only happen when you actually get the vehicles on the road. And they're pretty advanced; they can tell you that this is a kind of a vehicle and that's a two-wheeler. They can put the world together pretty well.
Watt: So another thing a car might at an intersection is that between the hours of 7 and 9 a.m., say, you can't make a left turn. This all gets programmed into the into the computer, the "brains" of a self-driving car?
Kalra: Absolutely. Keeping up with not only the rules of the road, which change from time to time, but the signage and what's happening. So these vehicles need pretty detailed maps of the world and then they have a pretty good idea of where they are within that world. There's a lot of technology on location, on sensing perception algorithms, and driving execution -- a whole lot of stuff that goes inside this technology.