After more than seven years of technical breakthroughs and 1.8 million miles of driving on public roads, Alphabet’s self-driving car is still the one to beat. But with commercial deployment still a way off, it’s also starting to look less than unique.
The project’s chief technologist, Chris Urmson, departed last week, the latest in a parade of talent to abandon the effort, which began in 2009 inside the Google X lab—now called just X and a subsidiary of the Alphabet holding company created to separate Google’s core business from its other projects last year. Alphabet and others that invested early in autonomous driving, such as Tesla Motors, have seen senior staff leave as competitors ranging from large automakers to Apple to well-funded startups have built up serious autonomous-car projects of their own.
“Others have been catching up to Alphabet; I wouldn’t say they’re running away with it anymore,” says Quin Garcia, managing director of Autotech Ventures, a Palo Alto investment firm focused on transportation startups. “Investors and corporations are willing to put big money into developing that technology.”
Recent advances in machine learning have made it easier to cover ground that took Alphabet's project years, says Garcia. And experienced engineers such as Urmson can be sure of strong backing even if they decide to go it alone, he says. (Urmson said in a blog post last week that he doesn’t have a new project lined up.) The $581 million General Motors spent earlier this year to acquire Cruise, a startup working on automated driving, has stoked confidence among entrepreneurs and investors that competing with Alphabet can be viable.
One founding member of Google’s original autonomous-driving project, Anthony Levandowski, left the company in January. In May he launched Otto, a San Francisco startup working on autonomous driving for trucks. The company has already modified trucks to drive themselves in tests on California freeways. Otto has more than 60 employees, according to LinkedIn, including several veterans of Alphabet’s and Tesla’s autonomous-driving teams.
Alphabet’s other competition includes Zoox, a startup based in Menlo Park, California, that is working on an autonomous vehicle intended to be used for a driverless taxi service. A June filing shows that Zoox has received $100 million in funding, and Business Insider reports it is set to double that figure.
One reason investors are backing self-driving startups is a belief that despite Alphabet’s early lead, the company is not well positioned to commercialize the technology.
The project’s leaders have said they are only interested in cars that can handle every type of driving without human supervision. To that end, they are developing cars that lack steering wheels and brake pedals for a human to ever use. And Alphabet wants to license its technology to a car manufacturer rather than make cars itself. Last year the former head of Hyundai’s American division, John Krafcik, was appointed as the project’s CEO.
That strategy faces two big challenges, says Robert Seidl, managing director at Motus Ventures, an early-stage venture capital firm focused on transportation. For one, “car companies don’t trust Google,” he says. “They don’t want to be left just bending metal and getting the low-margin part of the business.” And second, says Seidl, neither Alphabet nor the automakers are in a good position to launch a fleet of robotic taxis, which he and other investors argue would be the most plausible way to make money with fully autonomous vehicles.
An on-demand ride service could quickly recoup the cost of developing and manufacturing the cars by putting them to work for many hours a day, says Seidl. A robot car service could also be viable even if its vehicles were limited to operating in a certain area, he says. Autonomous vehicles rely heavily on detailed 3-D maps that must be kept up to date, so releasing a consumer car capable of driving anywhere would be a big challenge. Urmson said at MIT Technology Review’s EmTech Digital conference in May that he expected Alphabet's vehicles to come to certain urban pockets first.
Shahin Farshchi, a partner at the venture firm Lux Capital in Menlo Park, which invested in Zoox, says that neither a company built around answering Web searches nor one that specializes in selling cars for private use is likely to build out a good transportation service. “The notion of Toyota and Honda going into the autonomous-transport business is like saying Boeing and Airbus are going to be effective as airlines,” says Farshchi. (GM’s decision to invest $500 million into Uber competitor Lyft and work on a trial of self-driving ride service could be recognition of that challenge; Alphabet has a stake in Uber, but the ride-sharing company is aggressively pursuing its own self-driving technology.)
By insisting on waiting until fully autonomous passenger vehicles are ready, Alphabet could also miss a chance to capitalize on the early years of the technology’s emergence. Meanwhile, other companies could make money from selling self-driving technology for, say, highway driving, which is simple compared with getting a vehicle to cope with complex urban situations (see “Driverless Cars Are Further Away Than You Think”). Other companies, including Otto and Tesla, believe it’s possible to offer automated features for private or commercial driving—features that work only some of the time.
Urmson has been a vocal critic of the idea that partial automation is safe, saying that Google employees loaned self-driving prototypes in 2012 quickly became dangerously trusting of them (see “Lazy Humans Shaped Google’s New Autonomous Car”). He has also said Alphabet’s project is closer to being road-ready than some may think. At EmTech Digital in May he repeated his frequent claim that he was aiming to make it unnecessary for his son to get a driver’s license. That would mean Alphabet’s cars will hit the road in 2020. Whether he still believes that, or whether those cars will have to share the road with other self-driving vehicles created with Urmson’s help, is unclear.
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
How to fix the internet
If we want online discourse to improve, we need to move beyond the big platforms.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.