Chris Urmson at SXSW. Source: Author
Chris Urmson took the stage at South by Southwest in Austin, TX last week to provide an update on Alphabet's (GOOG 3.28%) (GOOGL 3.85%) self-driving car project.
For all the press the initiative gets, concrete plans are hard to come by, so the session was one of the most anticipated events on the conference calendar.
In his address, Urmson broke down the mid-February accident in Mountain View, CA -- the first instance of a car in the company's fleet being responsible for a collision. He showed the crowd how the cars "map" the world and answered some questions about how soon we could see this technology hit the mainstream.
What are the biggest obstacles?
For all the hoopla about consumer attitudes and the regulatory environment needing to be accepting of driverless cars, Urmson maintained that the largest obstacle is still on the technical side. Specifically, the systems still need to improve to handle the infinite number of situations a car might encounter and meet the challenge of predicting the behavior of other actors on the road.
Once there are fewer technical hiccups, Urmson and team will begin to focus more on integrating driverless cars into society and making sure they are being used in a way people are excited about.
Will the average person be able to buy an autonomous vehicle?
"When you look at what's in them, there isn't anything particularly interesting there ... there's no unobtainium," Urmson said, referring to the parts and components that go into the tech fueling the autonomous driving system.
It seems the cost is in the research and development and the extraordinarily deliberate simulations being run to get the technology road ready. The actual hardware under the hood isn't all that expensive, which is great for consumers.
It's still too early to nail down a number, but Urmson expects that once the technology is at mainstream scale, it will be priced so that it will be relatively accessible to everyone.
When? When? WHEN?!?!
The timeline for driverless cars is tough to pin down, but would-be adopters shouldn't be too discouraged by estimates that peg consumer availability several decades out. According to Urmson, driverless cars could be available in as few as three years to some individuals -- emphasis on some.
The disparity in ranges is largely due to the engineering challenges different climates present.
Two of Google's test cities, Mountain View and Austin, are an engineer's dream for weather -- no snow and little rain. But the project's third pilot city, Kirkland, WA, exposes the vehicles to all of the elements. Engineers are still working through the complexity that inclement weather, particularly snow, adds to the car's digital mapping system.
Urmson explained that the release of this technology will almost certainly be incremental.
"Places where the weather is good, where the roads are easy to drive, the technology might come there first," Urmson said. "And then once we have confidence with that, we'll move to more and more challenging locations."
Of course, before you pack your bags for the desert, lawmakers need to solidify regulations around autonomous vehicles.
Is Google actually getting into auto manufacturing?
The significant investment Google has made in autonomous vehicle technology has rightfully spurred curiosity over how the company plans to monetize it.
In his address, Urmson made it clear "we don't want to make cars." Instead, it seems the company is focused on remaining a platform player and partnering with incumbent auto companies for manufacturing.
A screenshot of Google's digital mapping system -- essentially how the car sees the world. Source: Author
OK, there's an old lady, an oncoming car, and a box of doll heads in the road, the Google car has someone tailgating it and can't avoid them all -- what happens?
One oft-cited concern with autonomous vehicles is that they require programmed decision making, effectively a predetermined moral compass.
In order for the car to understand how to process this kind of situation, there also has to be a formal hierarchy in accident avoidance. For the Google Car, Urmson explained the fleet is programmed to avoid collisions with other objects in the following order:
- Vulnerable road users -- pedestrians and cyclists
- Other moving things on the road -- vehicles
- Static objects in the road -- the box of doll heads
The truth is, the cars in Google's fleet are designed to be extremely defensive drivers, and because of their notoriety, Google's fleet has encountered road conditions most drivers would never come across.
Google's cars have experienced people running up and rolling over the hood, other vehicles speeding against traffic on one-way streets, and even a crosswalk game of human Frogger, so routine accident avoidance decisions are comparatively simple.
