Skip to main content

Astro Teller: Google X 'encouraged too much attention' for Project Glass

Astro Teller: Google X 'encouraged too much attention' for Project Glass


Failure is a feature

Share this story

Speaking today at SXSW, Google X's "Captain of Moonshots" Astro Teller took to the stage to talk about how the secretive lab is opening up and talking to the wider world more often — and earlier. It's a shift for Google X, which has traditionally kept its projects close to the chest before revealing them to the world in splashy ways. Google Glass, Project Loon, robotics, Project Ara, and self-driving cars are all "moonshots" that don't seem to have immediate business models today, but could theoretically be huge in the future. "When we say moonshots, what we mean is that we're shooting for things that are 10 times better," Teller said.

But Teller doesn't see those projects — or Google X — as secretive. Actually, he sees it as Google X putting out its projects as soon as possible to get feedback. Teller said he learned early on in his career at BodyMedia that he needed to get user feedback right away, not at the end of product development. "The longer you put off that learning, the most you will unconsciously avoid getting the news," he said, characterizing it as "failing at the beginning."

"We did things which encouraged people to think of this as a finished product."

Perhaps the most high-profile thing Google X has produced is Glass, the wearable headset that provides a minimal version of augmented reality. Google pressed forward with a fairly large "explorer" program to try to get feedback from early adopters. But Teller believes that in marketing that program, Google made a mistake. He called the program itself a "great decision," but argued that it wasn't positioned well. "The thing that we did not do well, that was closer to a failure, was that we allowed and sometimes even encouraged too much attention for the program," Teller said. He added, "We did things which encouraged people to think of this as a finished product." Instead, Google mostly wanted to learn what the social norms and uses would be, and didn't intend to make people think it was actually launching as a consumer product anytime soon — but that assumption was clearly widespread. Google Glass has now moved out of Google X and is now under the leadership of Tony Fadell.

Without entirely discounting the privacy concerns related to Glass, Teller did seem to imply he thought they were overblown. "Google Glass did not move the needle, it was literally a rounding error on the number of cameras in your life," Teller added, saying he was not making any apologies for Google Glass. But the conversation it started about privacy is important, and he hopes that it expands to the actual cameras that are everywhere, not just the prototypes on a few thousand people's faces.

Teller also gave an overview of other examples of early failures — some of which were actually intentional failures. One example of intentional failure is Project Loon. "We designed our early balloons to fail," Teller said. Google X designed the balloons with latex, which ensured that the cold in the high atmosphere would explode them. That meant they wouldn't drift into countries where Google hadn't yet secured permission to fly them.

Turns out fluffier socks are better

Teller also talked about the problems with leaks on the new designs of the balloons (an endeavor we've detailed in this feature). There was no way to simulate actual conditions in the stratosphere, which made it difficult for Google to know what was causing the leaks. Things got desperate: "We literally ended up doing a detailed study on the fluffiness of the socks" on the engineers who build the balloons, Teller said. Turns out fluffier socks are better than thin socks.


Teller pointed out that with self-driving cars, there was no way for anybody to collect the list of all the things that could go wrong — the only way to do it was to start using them in the world. He told a story about how the car was able to ignore a duck being shooed across the road by a lady in an electric wheelchair, exactly the sort of thing that nobody could ever plan for.

"The assumption that humans could be a reliable back-up for the system was a total fallacy."

Teller says that Google is trying to get "stumped like the duck" again, even though there is a much easier solution for getting to market: highway driving. In the fall of 2012, Google had already given out Lexus SUVs to non-Google employees to use — and on the freeway, they worked perfectly well. "We probably could have made a lot of money selling those," Teller said. But something surprising happened: "Even though people had sworn up and down 'I'm going to pay so much attention,'" and there were cameras watching them too, "people do really stupid stuff when they're driving." Basically, "the assumption that humans could be a reliable back-up for the system was a total fallacy!" Teller said. That's why Google began to pursue a car that doesn't even have a steering wheel or a gas pedal.

Teller also described a program for creating delivery drones. Google X settles on a "tail-sitter" design for its drones, meaning they would take off vertically. The failure there, though, was that after a year and a half, "80 percent of the team" knew it wasn't going to work, but the team also really wanted to get the project "out in the real world." But because Brin wanted a vehicle ready in five months, the team had to "double down on their failure," Teller says. But nevertheless, the team still learned from the experiment even though it knew it was essentially doomed. And now the team is moving forward with a new design.


Teller also described a wind turbine that the company is working on that doesn't require a tower. Instead of building a giant tower, propellers are arrayed in a circle and tethered by a cable to the ground. Larry Page wanted the team to "crash at least five versions of the vehicle," which was his way of trying to push the team to learn faster. So "we picked one of the windiest places in North America" to test a prototype, Teller says. Despite the high winds, the prototype didn't crash — and ironically Teller says the team was disappointed by that.


Teller says he doesn't regret anything about these mistakes, except for one thing. "I just wish we could have made these mistakes faster."