Through five years and 700,000 miles of testing, Google left one big question about its self-driving car technology unresolved: How would it fit into the rest of the auto industry?
Finally last week, in the form of a cute-as-a-koala electric buggy, came an answer: It won't.
Everything about the new test program that Google co-founder Sergey Brin revealed last week, from the design of its two-seat prototypes to the decision to test them on city streets instead of highways, points toward a personal transportation system that directly challenges the driver-centered, ownership-based business model the auto industry has relied on for a century.
"If you look at a vehicle purchase today, it's the second largest purchase most people in America make, and it's a resource that basically sits idle for 95 percent of the time," Christopher Urmson, the director of Google's self-driving cars project, told reporters. "It's kind of a poor capital investment, in some senses."
Google's idea of a killer app, therefore, is not a more accurate map or sensor to assist the driver but rather a vehicle that targets a pressing need for urban mobility without the need for a driver or driving at all. It envisions a network of self-guided, battery-powered vehicles standing ready in urban areas, able to be summoned with a tap on a smartphone and capable of taking their passengers to any destination.
If that sounds a bit like Uber with no one at the wheel, that may well be the point. Last year, Google led a $258 million investment in the on-demand taxi service through its Google Ventures capital arm.
Uber CEO Travis Kalanick acknowledged the connection at a conference last week in California. "When there's no other dude in the car," the blog TechCrunch quoted him as saying, in reference to the driver, "the cost of taking an Uber anywhere becomes cheaper than owning a vehicle."
Google ultimately may try supplying its software and maps to the mainstream auto industry, but car-sharing could offer the company and its technology a more direct route to the marketplace.
"All along, we've all been speculating about Google," said consultant Richard Bishop, who led the U.S. Department of Transportation's vehicle automation program in the 1990s. "We've known they're building very capable technology, but the question has been: What are they going to do with it? Here's a concrete use case -- the autonomous taxi -- that has significant potential, whether Google is the one who does it or not."
To prepare its autonomous cars to handle all sorts of urban driving scenarios, Google has commissioned a fleet of about 100 custom-built prototypes, which Urmson said will appear on California's public roads by year end.
Though current testing regulations require that a driver sit behind the wheel, Google's vision -- as depicted in its publicity materials -- is for the cars eventually to have no steering wheel or pedals.
Automakers such as Daimler AG and Volvo also are aggressively pursuing autonomous vehicles to make cars safer, improve mobility for blind and elderly people and make traffic jams and commutes less tiresome.
But none of those automakers envisions eliminating the role of the driver altogether.
Google does. It says it isn't confident enough about the handoff of controls to maintain a role for the driver in a self-driving car. Human factors research suggests that once people gave up control of a vehicle, they would be too trusting, Urmson said, and wouldn't be prepared to retake the controls quickly when they were needed.
Rather than taking a shot at "debugging the human," Google decided to go for a fully autonomous car.
Autonomous, that is, within certain human-defined limits. Google's custom-built prototypes will be electronically limited to 25 mph and will never go on highways. They will be designed as "neighborhood" vehicles under U.S. regulations, exempting them from many crash-safety standards. And they will have emergency buttons for pulling over and shutting down.
"We've built in occupant protection and pedestrian protection and tried to tailor it to the environment we'll actually be operating in," said Ron Medford, the project's safety director, who was deputy administrator of the National Highway Traffic Safety Administration before joining Google in 2013.
Kara Kockelman, a transportation engineering professor at the University of Texas, said Google's decision to contain its experiment to cities makes sense. The risk of injuries there is lower because traffic moves slowly, she said, and the higher population density means self-driving vehicles could spend more time carrying passengers and less time sitting idle.
"In many ways," Kockelman said, "cities are where we expect to see them first."
Rather than partnering with a major automaker, Google signed up Roush Enterprises of suburban Detroit to build its prototypes, according to a report last week by Crain's Detroit Business, a sibling publication of Automotive News.
That's largely in line with Google's corporate ethos. Brin, the company's co-founder, acknowledged last week in an interview with The New York Times that car companies are working on partially autonomous driving features, but he suggested that the pace of change was too gradual for Google.