Safety boon: A Cruise driverless vehicle in San Francisco. The firm was forced to halt operations in the city after one of its cars hit and dragged a pedestrian. — The New York Times
I took my first ride in a driverless car last week courtesy of Waymo, Google’s autonomous taxi service, which has received regulatory permission to test in the District of Columbia (DC).
The cars are mostly plying the streets for training and validation purposes, but Waymo invited a few journalists to check out the cars.
Mine had a human supervisor in the driver’s seat (as required by their current regulatory status in DC) but he never touched the wheel.
Instead, the car whisked me seamlessly from Google’s office near Georgetown Law School down New Jersey Avenue, through Lower Senate Park, before turning north onto First Street and successfully executing the nightmarish left turn into the traffic circle outside of Union Station and then back to the office.
In all, the ride last about 15 minutes.
I can’t say how well the software’s abilities generalise to other routes, but the driving was completely flawless across a range of scenarios – including the unexpected arrival of some fire trucks at one point – that appeared to me to pose a high degree of difficulty.
Since the point of the exercise was to court support from columnists like me, I was expecting the conversation with Waymo’s team to end with them asking for us to advocate for some type of regulatory relief.
But they said they had no actual imminent plans to enter the DC market, and are working build their existing offerings in San Francisco, Phoenix, Los Angeles and soon Austin and try to find a way to make the service profitable.
Meanwhile in China, Baidu Inc is operating a much larger fleet of vehicles across a larger set of cities.
A key difference, according to the New York Times, is that the Chinese government wants self-driving cars to succeed.
Which is why you see abuses there such as censoring online discussion of safety problems with self-driving cars.
That’s not something the United States could or should do. But the broad concept of trying to encourage autonomous vehicle development is sound and yet America is moving in the opposite direction.
Waymo’s progress is encouraging, but its two major competitors, Argo AI from Ford Motor Co and Volkswagen AG have shut down. Cruise, from General Motors Co, was forced to halt operations in San Francisco after one of its cars hit and dragged a pedestrian.
The accident was tragic and awful. But the pedestrian in question was originally hit by a human-driven car, which is how she came to be knocked into the path of the Cruise vehicle.
Neither the city of San Francisco nor the state of California nor anyone else reacted to the accident by suspending human drivers’ ability to pilot cars.
I was in San Francisco the weekend before my Waymo ride and saw many Waymos driving safely along Market Street.
I needed to get to Sausalito, which is outside their service area, so I hailed an old-fashioned artisanal taxi outside the Ferry Building by sticking my hand in the air. The driver was friendly but managed to make two wrong turns once we got to the other side of the bridge.
Walking from a meeting in downtown DC last Thursday morning to my office, I saw a human car driver make a careless right turn and hit the front wheel of a bicyclist who had the right of way.
The cyclist was unharmed beyond a skinned knee but a robotaxi wouldn’t have been as reckless as the human driver.
Safety first
Human error is part of life and the problem with the current American approach to self-driving regulation is that it lacks any objective standards or cost-benefit analysis. Robotaxis should be safe. But how safe?
Autonomous vehicles don’t get intoxicated. They never get exhausted after pulling an all-nighter. They don’t rage because of stress or get distracted by kids in the back seat or an unexpected text message.
The potential safety gains of robotaxis alone could be enormous simply by achieving the same performance standard of an average non-impaired human.
This is technology we should encourage with some genuine leniency for testing, and clear standards so that companies can be less fearful of experimenting.
Instead, we’ve unleashed the opposite dynamic. Waymo currently faces no real competition so there’s limited pressure to move forward aggressively.
Much of the regulatory opposition is unrelated to safety and driven instead by the economic interests of teamsters or taxi drivers.
Pair those interests up with car companies’ general worry about being forced to dramatically overhaul their core business model and you have a powerful lobby for an excessively cautious regulatory posture.
A general climate in Washington of hostility to “big tech” doesn’t help either at a time when the most plausible path forward is for Google to continue plowing profits derived from search engines into the robotaxi business.
This is not to say that autonomous vehicles won’t become a reality in America.
Waymo just received permission to operate on San Francisco freeways, which should let it offer a viable commercial competitor to Uber and Lyft in that market.
Their service area in Phoenix has steadily expanded over the years. They are growing, just slowly because it’s an inherently risky investment where no competitor is nipping at their heels.
Striking contrast
The contrast with the Chinese approach of pushing the industry forward is striking, with more cars on the streets of more cities, which means more data and more rapid learning.
If the basic technology works at all, which by all accounts it does, the best way to improve safety is to let the cars drive so their sample of scenarios and the quality of the decision-making gets better faster. In this scenario, a small lead could become a big one quickly.
Which leads to the other striking contrast, between the deep rhetorical investment in competition with China espoused by US politicians and the minimal amount of actual policy change.
Fully autonomous vehicles could be a boon to traffic safety, productivity (think of basic trucking functions), and even the military. In fact, it was a US Defence Advanced Research Projects Agency prize competition that jumpstarted autonomous vehicles in the United States.
Large-scale commercialisation is the best way to realise the promise of autonomous vehicles.
And if we take the competitive challenge seriously, that means adopting a regulatory posture that errs on the side of innovation and tries to encourage stable and competitive flows of investment as a national priority, not a series of city-by-city special interest fights. — Bloomberg
Matthew Yglesias is a columnist for Bloomberg Opinion. The views expressed here are the writer’s own.
