On a foggy Monday morning in Silicon Valley this week, I took a ride in Google’s self-driving car and nothing happened.
That’s a big deal.
That’s because the company announced its self-driving car project, which was created in 2009, has racked up over two million miles of driving experience. It’s a significant marker for Google — no other company has that many miles of fully self-driving experience.
But that landmark actually doesn’t mean much on its own, because the real milestone is how well the technology has developed in that time.
“It’s a nice round number, but it’s really about the quality of those miles,” said Google’s head of self-driving tech Dmitri Dolgov. “We’re building a driver,” he later added.
Indeed, because after riding in Google’s car, it was clear that two million miles in four cities, along with the millions of miles a day the company does in simulated rides, has helped the technology develop from what is basically a nervous teen student driver to the equivalent of a more experienced licensed person who drives every day.
Mistakes are still made and collisions happen — Google has been involved in 14 real ones so far, 13 of which were caused by other drivers — but the ride we took in its self-driving car was uneventful, which is the goal the company has long been aiming for consistently.
My ride started like your basic self-driving car test drive. A Lexus equipped with Google’s proprietary sensors and cameras pulled up to the front of its Mountain View headquarters building, with Dolgov in the driver’s seat. Another engineer sat in the passenger seat manning a laptop that was running the self-driving software.
My colleague and I sat in the back and were instructed to buckle our seat belts, as the car took us around the local and unpredictable roads on the San Francisco peninsula.
Almost immediately, a bicyclist crossed the street in front of us from the left; at the same time, a woman with a stroller walked off the curb preparing to cross on the right. The car next to us, driven by a human, stopped abruptly.
But the Google car didn’t jerk forward. Instead, it slowly came to a stop to allow the bicyclist to cross and then went along its way.
It was almost eerily similar to the archetypal situation ethicists often ponder: If a self-driving car came across a woman with a baby, a bicyclist and an elderly woman, whom would it risk hitting? In this case, nothing happened — again, a great outcome.
Later on the road, a pickup truck swerved into our lane and the Google car fairly smoothly and just slightly veered left to avoid it.
While it might seem table stakes for any car experience, this is a major feat for a fully self-driving car to navigate complex environments without having to disengage a single time, even if there are safety drivers in place in case a human is needed to take over. But in the face of obstacles human drivers often encounter, the robot car handled itself efficiently and the safety drivers never had to take the wheel.
Two million miles in more than seven years is a lot. According to a Google spokesperson, the average human only drives 13,000 miles a year.
In the first few miles, Google operated on freeways and the car was simply gathering fresh intel, learning about the environment and figuring out how to recognize and navigate around objects. Then, the company worked on getting the cars to understand how those objects behave in normal situations to teach the the car how it should react.
Now, Google is working on finessing the software’s ability to predict the behavior of other objects, such as pedestrians, cars or bicyclists, and react appropriately even when that behavior is not normal.
In all those miles, Google’s car has seen a number of anomalies, from a man with a chainsaw to a man on a horse in the middle of a road to sand storms in Phoenix. And when one car experiences it, all of its cars experiences it through the cloud.
But while driving experience is irreplaceable, no single company can experience all the edge cases a car will ever encounter. So, though initially the task was simply to get the car to drive, now the task is to make the car more human. That means, a car needs to be able to think, understand and communicate like a human driver would.
For example, it’s relatively straightforward to teach a car via software that if they are stopped at a red light at an intersection that the cars to its right or left have a green light. But it’s harder to teach it how to understand and predict whether a car will follow those rules and, if they don’t, how to react to them.
Google is also working on finding ways for the car to implicitly communicate its intention to other drivers or pedestrians, such as the car edging forward at a stop sign.
However, even if issues the technology might encounter are solved, Google has yet to publicly announce any strategies for monetizing and commercializing its self-driving car technology. In other words, making the self-driving dream a reality.
Interestingly, Uber — in which Google has a major stake, though an increasingly fractured relationship — is doing just that. It has just launched a real-world pilot of self-driving cars in Pittsburgh — though it’s not on par with Google technologically — and got one of the original members of Google’s self-driving project when it acquired autonomous trucking company Otto.
So, it was probably no surprise then that Alphabet executive David Drummond recently stepped down from Uber’s board, due to these obviously growing conflicts of interest.
That competition will surely continue between Google and Uber, as well as with the many players from auto giants like Ford and General Motors to tech rivals like Apple and Tesla to a myriad of startups.
Like I said, two million miles is impressive. Until everyone else, as they surely will, catches up.
This article originally appeared on Recode.net.