As 2018 dawned, expectations for self-driving vehicles were sky-high:
- Waymo, Alphabet's self-driving car company, was testing fully driverless cars and preparing for a commercial launch in the Phoenix area.
- Uber was racking up millions of test miles for its own cars.
- Tesla was selling a "full self-driving" upgrade for its vehicles—though the feature wasn't actually shipping yet. Several times in 2016, CEO Elon Musk predicted that the technology would be ready in about two years.
Self-driving technology seemed to be right around the corner. But then the industry was battered by bad news.
In March, Uber was forced to drastically scale back its testing activities after an Uber vehicle hit and killed a pedestrian in Tempe, Arizona. The same month, a Tesla customer died when his Autopilot-enabled Model X car slammed into a concrete lane divider.
Waymo hasn't suffered such a catastrophic event. But its long-promised December launch of a commercial service in Phoenix, called Waymo One, was a disappointment. Waymo canceled plans to offer a fully driverless service, leaving safety drivers behind the wheel. And Waymo only offered the service to people who were already in Waymo's closed testing program. Waymo has long been seen as the industry leader, but its halfhearted launch raised questions about how much progress it has really made.
"I've been seeing an increasing recognition from everybody—OEMs down to various startups—that this is all a lot tougher than anybody anticipated two or three years ago," industry analyst Sam Abuelsamid told Ars. "The farther along they get in the process, the more they learn how much they don't understand."
In the self-driving world, there's been a lot of discussion recently about the hype cycle, a model for new technologies that was developed by the Gartner consulting firm. In this model, new technologies reach a "peak of inflated expectations" (think the Internet circa 1999) before falling into a "trough of disillusionment." It's only after these initial overreactions—first too optimistic, then too pessimistic—that public perceptions start to line up with reality.
Driverless cars seemed to reach peak hype some time in late 2017. Then in 2018, the industry plunged into the trough of disillusionment, with some people wondering if driverless technology might be decades away. But today's extreme pessimism seems as unwarranted as the extreme optimism we saw a year ago. Maybe in 2019, the public will start to develop more realistic expectations for self-driving technology.
Uber, Tesla, and Waymo all struggled in 2018
The biggest, highest-profile companies working on self-driving cars have all had a difficult 2018.
It started with Uber. In March, an Uber self-driving vehicle struck and killed pedestrian Elaine Herzberg during public road testing in Tempe, Arizona. Uber halted on-road testing right after the crash and didn't resume it for nine months. Uber laid off its Phoenix-area safety drivers and abandoned the Phoenix market. Only in December did the company finally resume testing in Pittsburgh. In the rebooted testing program, Uber's cars are restricted to a top speed of 25 miles per hour and only drive a one-mile loop between two of the company's offices.
Tesla not only saw the death of one of its customers in California in March, the company suffered several other Autopilot-related mishaps. Tesla cars have crashed into parked fire trucks and other stationary vehicles—though thankfully no one else was killed. As we explained in June, Autopilot simply isn't designed to detect and avoid stationary objects when the car is moving at highway speeds.
To be fair, the driver-assistance technologies offered by most other carmakers have a similar limitation. The difference, however, is that Tesla has portrayed Autopilot as being a step away from full self-driving capability. To this day, Tesla's Autopilot page has a big banner at the top that says "Full Self-Driving Hardware on All Cars."
In January 2016—almost three years ago—Tesla CEO Elon Musk tweeted that "in ~2 years, summon should work anywhere connected by land & not blocked by borders, eg you're in LA and the car is in NY." In other words, he was suggesting that a Tesla car would be able to drive itself thousands of miles with no one inside as soon as early 2018.
In late 2016, Tesla began offering a "full self-driving" package for $3,000 that would supposedly enable this capability—though the company made clear it wasn't quite ready yet. Customers who signed up for this feature when it was originally offered have now been waiting for more than two years. Tesla quietly stopped taking new orders for the full self-driving upgrade in October 2018—seemingly an admission that the technology was still far away from completion.
"They still have a long way to go," Abuelsamid said about Tesla.
One likely reason for the slow progress is turmoil on the Autopilot team. Tesla's Autopilot team lost a number of key engineers and managers in 2017. Then in April 2018, Tesla lost its third Autopilot chief in 18 months.
What about Waymo?
Uber and Tesla's mishaps in the first few months of 2018 left me unfazed, because, as I wrote in June, I thought Waymo was in a class by itself. Waymo—previously Google's self-driving car project—had begun working on self-driving technology a good five years before many of its rivals launched their own efforts. By June, Waymo had already logged millions of miles of real-world testing and seemed to be building up the infrastructure necessary to launch a large-scale taxi service.
In the first half of 2018, Waymo ordered 82,000 cars for delivery over the next few years, and the company began aggressively hiring customer service and operations personnel. I argued that Waymo wouldn't be spending all that money unless the company knew its technology was nearly ready.
But the disappointing launch of Waymo One in December forced me to rethink that logic. Waymo probably still has the world's most advanced self-driving technology. But the halfhearted rollout of Waymo One—with safety drivers and an extremely narrow customer base—suggests that the company is nowhere close to being ready for a large-scale commercial deployment.
[contf] [contfnew] 
Ars Technica
[contfnewc] [contfnewc]
