astro teller the unexpected benefit of celebrating failure

In 1962 at Rice University, JFK told the country about a dream he had, a dream to put a person on the moon by the end of the decade. The eponymous moonshot.

No one knew if it was possible to do but he made sure a plan was put in place to do it if it was possible. That's how great dreams are. Great dreams aren't just visions, they're visions coupled to strategies for making them real.

I have the incredible good fortune to work at a moonshot factory. At X—formerly called Google X—you'll find an aerospace engineer working alongside a fashion designer and former military ops commanders brainstorming with laser experts. These inventors, engineers and makers are dreaming up technologies that we hope can make the world a wonderful place.

We use the word "moonshots" to remind us to keep our visions big—to keep dreaming. And we use the word "factory" to remind ourselves that we want to have concrete visions—concrete plans to make them real.

Here's our moonshot blueprint. Number one: we want to find a huge problem in the world that affects many millions of people. Number two: we want to find or propose a radical solution for solving that problem. And then number three: there has to be some reason to believe that the technology for such a radical solution could actually be built.

But I have a secret for you. The moonshot factory is a messy place. But rather than avoid the mess, pretend it's not there, we've tried to make that our strength. We spend most of our time breaking things and trying to prove that we're wrong. That's it, that's the secret. Run at all the hardest parts of the problem first. Get excited and cheer, "Hey! How are we going to kill our project today?"

We've got this interesting balance going where we allow our unchecked optimism to fuel our visions. But then we also harness enthusiastic skepticism to breathe life, breathe reality into those visions.

I want to show you a few of the projects that we've had to leave behind on the cutting room floor, and also a few of the gems that at least so far, have not only survived that process, but have been accelerated by it.

Last year we killed a project in automated vertical farming. This is some of the lettuce that we grew. One in nine people in the world suffers from undernourishment. So this is a moonshot that needs to happen. Vertical farming uses 10 times less water and a hundred times less land than conventional farming. And because you can grow the food close to where it's consumed, you don't have to transport it large distances. We made progress in a lot of the areas like automated harvesting and efficient lighting. But unfortunately, we couldn't get staple crops like grains and rice to grow this way. So we killed the project.

Here's another huge problem. We pay enormous costs in resources and environmental damage to ship goods worldwide. Economic development of landlocked countries is limited by lack of shipping infrastructure. The radical solution? A lighter-than-air, variable-buoyancy cargo ship. This has the potential to lower, at least overall, the cost, time and carbon footprint of shipping without needing runways. We came up with this clever set of technical breakthroughs that together might make it possible for us to lower the cost enough that we could actually make these ships—inexpensively enough in volume. But however cheap they would have been to make in volume it turned out that it was going to cost close to 200 million dollars to design and build the first one.

200 million dollars is just way too expensive. Because X is structured with these tight feedback loops of making mistakes and learning and new designs, we can't spend 200 million dollars to get the first data point about whether we're on the right track or not. If there's an Achilles' heel in one our projects, we want to know it now, up front, not way down the road. So we killed this project, too.

Discovering a major flaw in a project doesn't always mean that it ends the project. Sometimes it actually gets us onto a more productive path.

This is our fully self-driving vehicle prototype, which we built without a steering wheel or break pedal. But that wasn't actually our goal when we started.

With 1.2 million people dying on the roads globally every year, building a car that drives itself was a natural moonshot to take. Three and a half years ago, when we had these Lexus, retrofitted, self-driving cars in testing, they were doing so well, we gave them out to other Googlers to find out what they thought of the experience. And what we discovered was that our plan to have the cars do almost all the driving and just hand over to the users in case of emergency was a really bad plan. It wasn't safe because the users didn't do their job. They didn't stay alert in case the car needed to hand control back to them.

This was a major crisis for the team. It sent them back to the drawing board. And they came up with a beautiful, new perspective. Aim for a car where you're truly a passenger. You tell the car where you want to go, you push a button and it takes you from point A to point B by itself.

We're really grateful that we had this insight as early on in the project as we did. And it's shaped everything we've done since then. And now our cars have self-driven more than 1.4 million miles, and they're out everyday on the streets of Mountain View, California and Austin, Texas.

The cars team shifted their perspective. This is one of X's mantras. Sometimes shifting your perspective is more powerful than being smart.

Take wind energy. It's one of my favorite examples of perspective shifting. There's no way that we're going to build a better standard wind turbine than the experts in that industry. But we found a way to get up higher into the sky, and so get access to faster, more consistent winds, and so more energy without needing hundreds of tons of steel to get there.

Our Makani energy kite rises up from its perch by spinning up those propellers along its wing. And it pulls out a tether as it rises, pulling energy up through the tether. Once the tether's all the way out, it goes into crosswind circles in the sky. And now those propellers that lifted it up have become flying turbines. And that sends energy back down the tether.

We haven't yet found a way to kill this project. And the longer it survives that pressure, the more excited we get that this could become a cheaper and more deployable form of wind energy for the world.

Probably the craziest sounding project we have is Project Loon. We're trying to make balloon-powered Internet. A network of balloons in the stratosphere that beam an internet connection down to rural and remote areas of the world. This could bring online as many as four billion more people, who today have little or no internet connection.

But you can't just take a cell tower, strap it to a balloon and stick it in the sky. The winds are too strong, it would be blown away. And the balloons are too high up to tie it to the ground.

Here comes the crazy moment. What if, instead, we let the balloons drift and we taught them how to sail the winds to go where the needed to go? It turns out the stratosphere has winds that are going in quite different speeds and directions in thin strata. So we hoped that using smart algorithms and wind data from around the world, we could maneuver the balloons a bit, getting them to go up and down just a tiny bit in the stratosphere to grab those winds going in those different directions and speeds. The idea is to have enough balloons so as one balloon floats out of your area, there's another balloon ready to float into place, handing off the internet connection, just like your phone hands off between cell towers as you drive down the freeway.

We get how crazy that vision sounds—there's the name of the project to remind us of that. So since 2012, the Loon team has prioritized the work that seems the most difficult and so the most likely to kill their project.

The first thing that they did was try to get a Wi-Fi connection from a balloon in the stratosphere down to an antenna on the ground. It worked. And I promise you there were bets that it wasn't going to. So we kept going.

Could we get the balloon to talk directly to handsets, so that we didn't need the antenna as an intermediary receiver? Yeah.

Could we get the balloon bandwidth high enough so it was a real Internet connection? So that people could have something more than just SMS? The early tests weren't even a megabit per second, but now we can do up to 15 megabits per second. Enough to watch a TED Talk.

Could we get the balloons to talk to each other through the sky so that we could reach our signal deeper into rural areas? Check.

Could we get balloons the size of a house to stay up for more than 100 days, while costing less than five percent of what traditional, long-life balloons have cost to make? Yes. In the end. But I promise you, you name it, we had to try it to get there. We made round, silvery balloons. We made giant pillow-shaped balloons. We made balloons the size of a blue whale. We busted a lot of balloons.

(Laughter)

Since one of the things that was most likely to kill the Loon project was whether we could guide the balloons through the sky, one of our most important experiments was putting a balloon inside a balloon.

So there are two compartments here, one with air and then one with helium. The balloon pumps air in to make itself heavier, or lets air out to make it lighter. And these weight changes allow it to rise or fall, and that simple movement of the balloon is its steering mechanism. It floats up or down, hoping to grab winds going in the speed and direction that it wants.

But is that good enough for it to navigate through the world? Barely at first, but better all the time.

This particular balloon, our latest balloon, can navigate a two-mile vertical stretch of sky and can sail itself to within 500 meters of where it wants to go from 20,000 kilometers away.

We have lots more to do in terms of fine-tuning the system and reducing costs. But last year, a balloon built inexpensively went around the world 19 times over 187 days. So we're going to keep going.

(Applause)

Our balloons today are doing pretty much everything a complete system needs to do. We're in discussions with telcos around the world, and we're going to fly over places like Indonesia for real service testing this year.

This probably all sounds too good to be true, and you're right. Being audacious and working on big, risky things makes people inherently uncomfortable.

You cannot yell at people and force them to fail fast. People resist. They worry. "What will happen to me if I fail? Will people laugh at me? Will I be fired?"

I started with our secret. I'm going to leave you with how we actually make it happen. The only way to get people to work on big, risky things—audacious ideas—and have them run at all the hardest parts of the problem first, is if you make that the path of least resistance for them.

We work hard at X to make it safe to fail. Teams kill their ideas as soon as the evidence is on the table because they're rewarded for it. They get applause from their peers. Hugs and high fives from their manager, me in particular. They get promoted for it. We have bonused every single person on teams that ended their projects, from teams as small as two to teams of more than 30.

We believe in dreams at the moonshot factory. But enthusiastic skepticism is not the enemy of boundless optimism. It's optimism's perfect partner. It unlocks the potential in every idea. We can create the future that's in our dreams.

Thank you very much.

(Applause)