For people who drive for ride hailing firms like Uber and Lyft, the irony could not be clearer: the more they pick up and drop off passengers, the more they are contributing to the companies replacing them with computers.
Uber and Lyft are among several companies rushing to develop a Level 5 car. As defined by the Society of Automotive Engineers, a Level 5 car does not require human drivers and can go anywhere and do anything a human driver can do. Such vehicles won’t even require a steering wheel or pedals.
Tech giants and auto manufacturers ranging from Google and Apple to Ford and Audi are also hoping to be among the first to bring such a vehicle to market.
But despite all of that money and brainpower, some experts say self-driving cars are still several years away. In a report, consulting giant McKinsey & Co. estimates it will take a decade or so before we see such vehicles on the road because engineers still need to perfect the necessary software.
“The software to complement and utilize the full potential of autonomous-vehicle hardware still has a way to go,” the McKinsey report said. “Development timelines have stalled given the complexity and research-oriented nature of the problems.”
But Kumar Chellapilla, a top Lyft engineer who oversees its Level 5 program, said Lyft (and by extension Uber) has a specific advantage over its tech and auto competitors: information.
Specifically, data from the 10 billion miles driven by nearly 2 million Lyft drivers since the company’s inception in 2012, he said.
By collecting and analyzing the data from its core ride hailing business and searching for patterns, Lyft can teach its autonomous vehicles to “drive more human like,” Chellapilla recently told attendees at UC Berkeley’s Institute of Transportation Studies. “What do humans normally do?”
“Humans are actually good drivers,” he said. “Humans can provide the answer to ambiguous problems” autonomous vehicles will encounter on the road.
So each time a Lyft driver transports a passenger to a destination, he/she is helping the company to perfect self-driving technology.
In addition, Chellapilla said, Lyft insures all of its vehicles and can use the data to measure risk and compare how safely autonomous vehicles drive when compared to humans.
So the key to self-driving cars is not necessarily technology but knowledge, Chellapilla said.
When artificial intelligence programs first challenged chess masters in the 1980s, humans soundly defeated machine. Finally, in 1997, IBM’s Deep Blue program narrowly defeated world champion Gary Kasparov because researchers over the years were able to add cumulatively more knowledge about previously played chess matches to the computer’s memory, Chellapilla said.
“How do we use our knowledge to win at self-driving?” he said.
Despite the still formidable technological challenges, Chellapilla said he is confident self-driving cars will reach consumers.
“It’s only a matter of when,” he said.