SAN FRANCISCO -- At GoogleX, failure isn't just a good thing. It's something their engineers strive for.
"We need failures," said Astro Teller, head of GoogleX, the company's secretive innovation lab. "If we are going to build something, we need it to fail and fail quickly so we can learn as much about it as we can. If something doesn't fail, how are we going to learn from it?"
Teller spoke to a packed room Friday at Google I/O, the company's annual developers conference.
His subject was moonshots and failures. The company, he said, needs both.
Actually, according to Teller, everyone needs both moonshots and failures.
"We need to be reminded about the risks we're taking and the long-term things we're looking ahead to," said Teller, who's official title is Captain of Moonshots. "We can all work on moonshots. Working on things that aspire to be 100 times better, rather than 10 times better, is something really worth working toward. When you aspire to make the world that much better, you have to come at it from a new perspective and not depend on what people have done before."
During his talk, Teller touched on various projects the Google X group is working on, including self-driving cars, floating balloons in the stratosphere to bring Internet connectivity to remote areas and the increasingly maligned Google Glass wearable.
Google stopped selling the Glass prototype in January and pulled it out of the public spotlight so the device could be reworked. That move led to speculation that Google planned to kill the entire Glass project.
Teller said that's not the case, and that the computerized glasses have been moved from under the umbrella of GoogleX, and into its own niche at Google.
"Google Glass, I think, is making really good progress," he said. "It's graduated from GoogleX. We'll hear more about that."
He added that GoogleX made one great decision and one really bad decision on Glass.
"The thing we did right was get it out into the world," Teller said. "The Explorer program was the right thing to do. We were trying to learn about the social issues around Glass. But there were ways in which we were doing that, like putting it on a runway, that made people think this was a finished product and not a prototype. We left people with some confusing messages there and I wish we'd done that differently."
Teller also talked about Google's autonomous car project, noting that the car turned into far more than it had initially been expected to be.
For instance, two-and-a-half years ago, Google's auto engineers thought they were basically done with their car efforts. They had added some autonomous tools to the car that they thought would help drivers.
However, when Teller allowed some Google employees to drive the cars home, telling them to make sure they were always engaged with the car, he found that many of them weren't paying attention to their driving at all.
That discovery prompted the GoogleX engineers to get back to work and make a fully autonomous car that required no driver intervention.
"Pretty good most of the time is not good enough for a self-driving car," Teller said. "The cars performed flawlessly. The people did not. People don't pay attention when they're actually supposed to be driving. They're putting on makeup. They're texting… We learned that if you are going to have a self-driving car, it has to go from Point A to Point B with no human intervention."
The GoogleX team also worked on Project Wing, which entailed building an autonomous drone for delivering goods. The team chose to deliver heart defibrillators to people who call 911 to report having a heart attack.
Teller said they figured they could get the potentially life-saving equipment to the person in need much faster.
What they learned, however, was that people generally don't know how to use defibrillators, so the whole project wasn't as helpful as they had hoped it would be.
"They're not easy to use and if you can't use it, it's not as world changing as you'd hoped it would be," Teller said.
This story, "GoogleX head talks moonshoots and need for failure" was originally published by Computerworld.