Astro Teller And His Google[x] Moonshot Updates & Failures

Astro Teller told a full auditorium at the Austin Convention Center, during SXSW, that the last 5 years of Google[x] has taught him to fail and fail at the beginning. He admitted that it is hard. That the longer you work on a project, the less you want to know the world’s reaction to it. He continued to tell the crowd how a number of Google[x] projects have experienced a series of “bumps and scrapes.”

What is a Google Moonshot? They are projects that solve a huge problem, have a radical solution, and involve breakthrough technology. Moonshots must produce a massive value to make things better, not only by 10% or 20% increments, but by 10X.

Teller continued to tell us about past, current, and future projects:

Project Loon

  • Moonshot : Balloon-powered internet for everyone. Astro said 4 billion people in this world are not connected digitally. Project Loon helps them to be connected.
  • How : A network of balloons to act as cellular towers in the sky
  • The Failure : These balloons needed to be so high that they couldn’t be tied down. The team was essentially making a cell tower that needed to be 1% of the standard weight, produce more power, and to function in 90 degrees below zero. Finally after creating a device, the balloons that were used to keep the device high above the earth started to leak and eventually burst. When their balloons failed, they had to fetch it (sailing in the pacific, scavenging the Arctic).
  • Lesson Learned : Turns out it was the socks that the techs were wearing when they were manufacturing them. They found the issue, tackled it, and now the balloons last for 6 months at a time.

Project Driverless Car

  • Moonshot : We spend millions of hours driving. Those hours are wasted. The goal is to make a car drive safer than a human.
  • How : Through artificial intelligence that learns on the fly through it’s surroundings. Astro told the audience about an incident that happened 4 or 5 months ago. Google’s driverless car was driving on yet another test run in the suburbs of California. While driving on the road the car “saw” in front of it a women in an electric wheel chair with a broom. She was reaching into the middle of the road trying to chase a duck to shoo it off the road. The car came to a stop, waited til the woman was out of the road, then continued on it’s way. There is no way for the Driverless Car team to sit in a conference room and say “we need to watch out for women in wheel chairs, in the middle of roads, trying to shoo ducks”. Thats why there is AI. Now that specific Driverless Car computer knows what that specific incident looks like. Astro continued to say that they show the computer this moment and 10,000 other similar moments, and it learns! The more bumps and scrapes in the project the better. It is not about not getting them, its about extracting the value from them.
  • The Failure : When the product was starting to become more and more developed the team started to work with safety drivers. These safety drivers were assigned to ride with cars during the test runs. The safety drivers were there to make sure the car was driving properly and to help in emergency situations. The Google[x] team had cameras watching the human’s every move. What the team found was that the humans trusted the car more and more over time and started doing stupid stuff during the testing. The humans would try to take over in weird moments, or not pay attention at all. The safety drivers, in the end, made the cars drive worse. They assumed humans were a good back up system to the driverless car.
  • Lesson Learned : The Driverless Car team made a car with no steering wheel, brakes, accelerator, or other parts so the humans riding in them wouldn’t be tempted to “take over” the ride. Astro said that regulations did not require cars to have a steering wheel so they cut it out. Oddly enough it was required to have rear view mirrors, so they kept those on.

Project Google Glass

  • Moonshot : Connecting the offline and online worlds.
  • How : Through a smaller and slimmer head-mounted and hands-free display design.
  • The Failure : They made one great decision on google glass, in that they did the glass explorer program. They made a bad decision because they allowed the explorer program to get to big for it’s own good. “We allowed and somewhat encouraged too much exposure to the program….We did things to encourage people to think that [Glass] was a finished product,” Astro said.
  • Lesson Learned : The Glass explorer program led them to learn a lot on the tech front. Including battery issues and improvements. It also led them to social discoveries and how social norms could be built. But Google learned that there is a fine line between rushing into production and perfecting a product.

Project Genie

  • Moonshot : Help the field of construction and architecture. It would design and architect buildings. It was a building genie.
  • How : The invention is a cloud based collaboration software with “planning applications to help architects and engineers in the design process, especially for skyscrapers and large buildings. The platform includes planning tools of expert architects and engineers and advance analytics and simulation tools.”
  • It is now called Flux, a stand alone business. It is the only project that is its stand alone business.
  • Flux automatically answers zoning questions, and correct building codes for a specific area.


Project Wing

  • Delivery of things via self flying vehicles
  • Goal : Why are we sending 2-3 pound packages in 3000 pound vehicles across towns in traffic
  • First road block, could they possibly use someone else vehicle so they didn’t have to build a vehicle? No. There wasn’t anything that fit what they were looking for, so they had to build one. It had to be a vertical take off vehicle because they couldn’t have run ways when out on delivery.
  • They built one. They call them tail sitters. It is a hover able and wing engine vehicle. They are perfect for flight, but they are harder to control.
  • After 8 months, 50% of the team knew it was a failure. Then after a year and a half 80% of the team knew it was a failure. The lesson they learned, admit when something is not working, while also trying to get out to the world as fast as possible.
  • Sergey Brin saw the stall in development and said “You have 5 months….go.”
  • They finally got out to the real world to a select group in australia. They learned alot and because of Sergey’s deadline they have a new and improved vehicle.


Project Makani // Failing to Fail

  • An energy source
  • Goal : Harvest the power of the wind at a fraction of the cost of normal wind turbines.
  • They discovered that the higher they got the higher the wind speed and the more consistent the speeds. There is an enormous advantage to going up when harvesting wind for power.
  • A normal wind turbine is about 100 meters and 300/400 tons. They are an incredible amount of money.
  • How : Next month they have a turbine that will fly, and once it is released the length of the tie down (80ft), the wind turns it’s 8 propellors. The kite like contraption is the most efficient wind energy producer.
  • Larry page said, “make sure you crash at least 5 versions of the vehicles.” So the team picked the windiest place in America to test the product. But they failed to fail. They never crashed it.

Leave a Reply

Your email address will not be published. Required fields are marked *