Speaking at the MIT Technology Review EmTech Digital conference in San Francisco on May 24, Sterling Anderson, director of Tesla’s Autopilot program, revealed how Tesla’s ability to access each car it builds via the internet allows company engineers to design and improve autonomous driving systems.
Tesla first began installing the hardware that makes Autopilot possible in October of 2014, but it did not activate the software until nearly a year later. During that time, Tesla downloaded data from millions of miles worth of driving. The data revealed how people actually drive and gave Tesla a baseline of information about road conditions.
“The ability to pull high-resolution data from these vehicles and to update the vehicles over the air is a significant part of what’s allowed us in 18 months to go from very behind the curve to what is today one of the more advanced autonomous or semi-autonomous driving features,” said Anderson.
“Since introducing this hardware 18 months ago we’ve accrued 780 million miles,” said Anderson. “We can use all of that data on our servers to look for how people are using our cars and how we can improve things.” Every 10 hours Tesla gets another million miles worth of data, he said.
Anderson also said something even more surprising. When Tesla engineers come up with new software ideas, they can be tested by uploading them to customers’ cars without them knowing about it. Then the company can monitor how the tweaks perform in real world driving before deciding whether to activate the new features via an over the air update.
“We will often install an ‘inert’ feature on all our vehicles worldwide,” said Anderson. “That allows us to watch over tens of millions of miles how a feature performs.” While ‘inert’, the software has no ability to actually control any of the functions of the car.
Google has been working on autonomous driving systems much longer than Tesla has. But its data collection is limited to a small fleet of company owned vehicles operated under tightly controlled conditions. Tesla is able to gather data from 70,000 cars located all around the world. It monitors how real people drive under real conditions.
The philosophical differences between the approach both companies take toward self driving technology are dramatic. When a Google engineer took his eyes off the road during a test to plug in his cell phone charger, the company was aghast. It quickly added constraints to its software to prevent such behavior.
Tesla, on the other hand, has taken a much more relaxed approach. When its Autopilot software was first activated, many Tesla drivers started doing ridiculous things with it, like the German banker who filmed himself riding in the back seat reading a newspaper while his car drove him to work on the autobahn.
Tesla gives people great freedom, then dials things back when it discovers issues. Just a few days ago, a Tesla driver was filmed sound asleep while his car drove itself on the freeway. Some Tesla owners tell of hanging weights from the steering wheel to fool the software into thinking there is a hand on it.
Anderson said Tesla’s data driven strategy allows the company to keep advancing the company’s Autopilot technology. It will allow the system to operate more effectively in urban driving environments that have challenges like intersections, pedestrians, and other less predictable factors.
Tesla must be aware of drivers’ expectations he said, but doesn’t need to take them out of the equation altogether. “Autopilot is not an autonomous system and should not be treated as one,” said Anderson. “We ask drivers to keep their hands on [the wheel] and be prepared to take over.”
Anderson’s remarks illustrate once again what is widely known but not always internalized. Every Tesla owners is a beta tester for the company. It it is possible that, in the future, as Tesla becomes more of a mainstream car company and less of a tech company, not all of its customers will he comfortable with that role.