Google’s Driverless-Car Czar on Taking the Human Out of the Equation

How John Krafcik went from mechanical to digital and why he thinks you have to go fully autonomous.

The following is a condensed and edited interview with John Krafcik, CEO, Google Self-Driving Car Project.

You devoted your life to human-driven transportation, engineering SUVs at Ford and taking Hyundai (as U.S. CEO and president) to record levels of sales in the U.S. Why did you go driverless?
For a lot of my generation, perhaps especially those coming from the family situation I had—my dad was a tool and die maker—mechanical things and complicated physical objects have always been a source of undue fascination. The driverless cars we’re working on now are consistent with those human-piloted objects of my youth.

Did you find kindred spirits in car enthusiasm at Google?
There are a surprising number of car geeks here. That joins a lot of us on the team, which I think proves the point that it’s the fascination with complicated physical, mechanical devices. Mind you, there are a lot of hard-core software-coding types as well. There’s a unifying element: When I was at Ford and Hyundai, I thought it wasn’t fair for consumers to have to make the choice of what’s the safest way to equip the car. We at Google think we can keep more people safe than the way we currently get from A to B.

“We need to take the human out of the loop”

What can be learned from the Tesla fatality?
Well, first of all, it’s a tragedy. I mean, Joshua Brown lost his life. A couple of key points, though. One is, he was one of probably a hundred or so people who died that day in automotive fatalities, in the U.S. alone. You know the statistics: 35,000 fatalities, up 7 percent from the year prior. Globally, it’s over 1.2 million. It’s as if a 737 was crashing every hour of every day all year. From a macro standpoint, it’s a very, very big problem.

But we need to make sure we’re using the right language when we talk about what happened with that accident, because that wasn’t a self-driving car, what we refer to as an L4, or fully autonomous car. That was a car with traffic-aware cruise control and a lane-keeping function—an L2, where, for better or worse, it was the responsibility of the driver to be cautious. We, as humans, are fallible creatures. [The crash] confirms our sense that the route to full autonomy, though much harder, is the right route. And we’ve learned from experience what happens when you put really smart people with really clear instructions inside a car with capabilities like that Tesla one.
 
Back in 2012 we had a technology that was very similar. We let Google employees test it, after lengthy training sessions imploring them to pay attention at all times. We wanted to see how they were interacting with the technology. After three months we saw enough to say this is definitely a problem. People would take their eyes off the road for some period, look down at their phones and start texting while in the driver’s seat. Turning around to the back to get their laptop because they needed to plug their phone in. Right? When you’re hurtling down the road at 60 miles an hour in a two-ton vehicle?

That takes us to the fundamental conundrum of the L2 semiautonomous solutions: As they get better and better, but not quite good enough for humans to zone out entirely, then risk increases. So we need to take the human out of the loop. With L4, which is our focus at Google, the idea is, you don’t need a steering wheel or controls because we’re going to take care of everything, and you just have to say, “I want to go to that destination,” and the car will take you there.

What role do you think government should play in promoting autonomous vehicles?
It’s incumbent on them to provide the regulatory framework so that people can feel confident that the technology is evolving in the right way and is safe. That’s not an easy task. Recently, California decided an extra level of licensing would be required to operate an L4 vehicle, which runs counter to one of the major social goods we think autonomous vehicles can provide—giving folks like my 96-year-old mom the ability to move from point A to point B. Not just one driver’s license, but two would be required for an autonomous vehicle.

Now California I think has realized they went too far. One of the things we suggested, back in April, was that those working on this technology put forward recommendations based on data that would allow changes to current Federal Motor Vehicle Safety Standards requirements and the like to actually get this technology out on the road.

AlixPartners just did a survey that found 73 percent of people absolutely wanted autonomous vehicles. But when people had the option to have a wheel in the car, allowing optional full control, the acceptance rate went up to 90 percent. Would that help people get over being frightened by a robot car?
Yeah, I think if you could solve that problem of the handoff and making it crystal-clear who was in control. We all have fond memories of being behind the wheel. It’s a beautiful Sunday morning, and maybe you take an extra day, and you’re driving up the Pacific Coast Highway on the way to Carmel. And those are shining moments. If we’re really honest with ourselves, though, and try to remember the times that we’ve been driving a car where we really, truly enjoyed that experience, I think they’re few and far between. For the vast majority of time, you’d prefer to be taken from point A to point B in a fully autonomous car.
 
It’s a certain sort of joy and exhilaration when someone’s in a self-driving car for the first time. I have been in cars with folks who’ve been in the industry for some time, and seeing how quickly they adapt, relate, and then relax—it’s astonishing. It takes about five minutes or less. That trust is developed.

“This is a company that doesn’t have to sell cars today, is not burdened by legacy, and can take a clean-sheet approach”

If you can show them what’s happening with a dashboard map and how they’re being kept safe, that puts them at ease?
Yeah, here’s this 360-degree view. We’ve become quite fond of this angle where we pull up as if there were a camera—at a 45-degree angle, maybe 50 feet up in the air—giving folks that … 

Like the drone view?
Yeah. It’s led some people to ask the question, “Who’s shooting that video of our car?”

Google X is trying to reimagine everything. Does it create a Big Brother atmosphere?
I hope it doesn’t. As I looked at this project from my prior position it felt pretty cool. This is a company that doesn’t have to sell cars today, is not burdened by legacy, and can take a clean-sheet approach. It’s a wonderful service to the industry.

In this very connected world that we are creating, how do you protect people’s privacy?
At this point, we’re testing cars with test drivers in them, so we haven’t had a lot of privacy-related situations come up.

But these cars will create a lot of data. Should the government get it to help design smarter cities? Should companies or courts have it to resolve accident disputes? Who owns the data?
All questions to be determined.

Is the balance of power shifting to Silicon Valley from Detroit?
That’s too bold a statement. We demonstrate that with the fact that we just opened up an R&D center in Detroit.

Does Google want to build its own cars?
Google realized that it’s really hard to build a car. We built the little prototype that gave us a taste of the complexity. And we had [automotive engineering company] Roush and Conti [Germany’s Continental Automotive Systems] helping us. The 100 hybrid minivans from [our new partner] Fiat Chrysler more than doubles our test fleet.

What will transportation look like 20 years from now?
I do think one of the fundamental shifts the transportation world will pivot around is a movement by all the players to recognize that the unit of economic optimization will have moved sometime during this period from vehicles sold to trips provided. Fully autonomous cars are going to be more expensive. Society will find ways to better utilize those assets. Cars are used just 4 percent of their time—96 percent of the time they’re sitting in parking spaces, and each car has somewhere on the order of three or four parking spaces reserved in its name in our great land. That’s a shame, especially for cities.

We’ve heard that Alphabet has set aside $10 billion as a long-term bet on the self-driving car program.
Really? Somebody said that? I don’t know who said that.

Sergey Brin has predicted the death of car ownership.
The same person who said $10 billion might have said that.

You said we’ll always have cars that are a blast to drive. So which is it?
We’ll always have cars that are a blast to drive, absolutely.

Will we be allowed to drive them on public roads or will they be show ponies?
No, we absolutely will be. I don’t think in our lifetimes we have to worry about driver’s licenses being taken away from people or the freedom to drive our vehicles where we want. I don’t think that’s a future that any of us—or anyone—wants.

No one is taking your Porsche 911 away from you.
No, no!

Subscribe to Buisnessweek