X is Google’s “moonshot factory,” a place to work on projects, like autonomous vehicles, that have the potential to radically reshape the world. To get insights into how the operation works, and why it celebrates failure, The Wall Street Journal’s Rebecca Blumenstein spoke with Astro Teller, captain of the factory. Here are edited excerpts of their conversation:
Blumenstein: How do you decide which projects to tackle?
Teller: There has to be some huge problem with the world that we can name and say that we’re excited about going to tackle. Then there has to be a radical proposed solution for solving that problem. And there has to be some specific set of technologies which look really hard, that if we could get working, would enable that radical proposed solution.
Blumenstein: You take a certain amount of excitement in killing projects.
Teller: All things being equal, I would rather that a project succeed than go away. I would reframe the statement this way. If your project isn’t going to make it, if it has an Achilles’ heel, do we want to find out about that now, or should we spend tens of millions of dollars and find out about that two or three years from now?
Blumenstein: There is a lot of impatience in the market for Google’s car.
Teller: Getting a car to not drive off the road on a highway is not hard. Working better than a human in all of the situations that a human does that is four or five orders of magnitude harder. We’ve done 2 million miles of driving. To not drift out of your lane, good. But when someone walks out into the road holding a stop sign in their hand, you have to stop. If our car can’t do that and 10,000 other things like that, then we, nor is any company, ready to be on the road.
Blumenstein: You’re in the process of graduating the car. What does that mean?
Teller: We separated their [car business] finances from X’s finances on January 1st. They have a lot of corporate legal stuff to do before they’re fully graduated. When Verily, our life-sciences business, graduated last year, it took them an entire year to check all these different boxes. Even once they were mostly graduated, they were still relying on us for a lot of services. We’re going through that with the self-driving car group. I would say the same thing about the cars themselves. You don’t want us to, all of a sudden, have every car in the world be driving themselves. You want every step to be rational and clear and borderline boring, because that’s probably what’s safe.
Blumenstein: You’re launching balloons that deliver high-speed internet access to the developing world.
Teller: There are about four billion people in the world who have poor or no internet access. What if you could make something like a supercheap satellite, but hang it under a balloon, launching it for a thousandth the cost of putting up a normal satellite?
But the winds are very unforgiving in the stratosphere. You can’t put a big gas tank on it and a propeller. The wind is too strong, the balloons make too big a surface area, too big a sail, basically. What if instead of fighting winds we went up and down? It takes almost no energy to go up and down a few kilometers. What if we taught the balloons to get super smart at knowing which way the wind would be blowing and how fast at different altitudes?
We had a balloon stay over Peru for about three months. And then come home, back to California.
Glass half full
Blumenstein: Let’s talk about Google Glass, a product that many say wasn’t a commercial success.
Teller: More of our professional and personal lives are mediated by technology. Figuring out a way for there not to be cognitive dissonance created by our physical lives and digital lives is very important. Google Glass took good steps in the right direction.
We could’ve done a better job of signaling to the world and reminding ourselves that the thing we built was a learning prototype. It wasn’t a finished product.
Edited excerpts from an interview at The Wall Street Journal's WSJDLive 2016 global technology conference