Tuesday, July 12, 2016

Silicon Valley-driven hype for self-driving cars

Lee Gomes

July 12 2016

Because of reports that he may have been watching a Harry Potter movie at the time, Mr Joshua Brown, the Tesla Model S owner killed in the self-driving car industry's first known fatal accident, has come to be regarded as a reckless contributor to his own sad fate.

Here is another view: Mr Brown may be the first casualty of the widespread and potentially dangerous belief that autonomous cars are much closer to being road-ready than they actually are.

Mr Brown, who died in Florida on May 7, does not appear to have been heeding an important rule in the official instruction manual for the Tesla Autopilot feature he was using: Drivers should keep their hands on the wheel and be ready to resume control of the vehicle at any time.

Instead, he seems to have been answering to a higher authority: Mr Elon Musk, a founder and the chief executive of Tesla Motors. Mr Musk is well known for his salesmanship and he used it liberally in promoting Autopilot. "It's almost twice as good as a person" was one of his claims. Another: A driver could use Autopilot for the roughly 1,300km between San Francisco and Seattle almost "without touching the controls at all".

It would not be fair to tar all of the growing self-driving industry with Mr Musk's braggadocio. Nonetheless, his technological over-promising fits into a common narrative in Silicon Valley: The major engineering problems with self-driving cars have essentially been solved and their widespread adoption is inevitable. Ask "when?" and you will usually be told, "Much sooner than you think."

Some lawmakers are even talking about scaling back investment in mass transit, which they claim will be unnecessary in a world full of robot chauffeurs.

Aware of the conventional wisdom that robotic cars are about to cause an epochal "disruption", automakers are eager to demonstrate that they are fully engaged. A result has been a drumbeat of announcements auguring the imminent arrival of robotic cars, almost as though they were the next generation of iPhones. The breathless statements are especially beguiling for members of the public without the engineering background needed to understand the challenges that remain. In other words, most people.

Volvo, for example, recently announced a programme called "Drive Me London", which in two years will put 100 self-driving cars on the road in what it says will be "Britain's most ambitious autonomous-driving trial". Envisioning myself being driven around Trafalgar Square by a robot, I called up the company for more information. It turns out that the news release did not quite tell the whole story. I was told that all the autonomous driving will initially be limited to England's major motorways; everywhere else, the steering must still be done manually, including within London proper.

Motorways and expressways are the low-hanging fruit of autonomous driving - everyone is moving in one direction at the same relative speed and there are no pesky pedestrians to get in the way. Much of what is passed off today as "autonomous driving" is some variation of this sort of advanced cruise control.

But there is an elephant in the cab with even this rudimentary form of autonomy. Many companies are planning cars that, in the event of an emergency, hand back control to the human driver. (Google, a notable exception, is planning a car with no steering wheel or brake pedal.) The potentially fatal weakness of that strategy is that it assumes "drivers" will be paying attention at the split second they are most needed, instead of being busy, say, taking a nap.

The much harder, and still mainly unsolved, autonomous-driving problem involves not highways but cities, with all their chaos and complexity. Self-driving cars still struggle with simple potholes - no one has come even close to demonstrating a completely driverless car that could do the work of a Manhattan taxi driver on a rainy day.

The sad reality of autonomous car technology is that the easy parts have yet to be proven safe and the hard parts have yet to be proven possible. We are nowhere close to Silicon Valley's automotive "Tomorrowland". The most realistic industry projection about the arrival of autonomous driving comes from the company that has done the most to make it possible. Google, while never explicitly saying so, has long intimated that self-driving cars would be available by the end of the decade.

In February, though, a Google car caused its first accident - a bus collision with no injuries. A few weeks later, Google made a significant, if little-noted, schedule adjustment. Its project director Chris Urmson said in a presentation that the fully featured, truly go-anywhere self-driving car that Google has promised might not be available for 30 years, although other much less capable models might arrive sooner.

Historians of technology know that "in 30 years" often ends up being "never". Even if that is not the case here, if you are expecting a self-driving car, you should also expect a wait.

And so you might want to do something to pass the time. Maybe go for a nice drive?


Lee Gomes is a former technology reporter for The Wall Street Journal.

No comments: