Everything you need

Google’s top secret moonshot factory, X

X – formerly Google X – aims to make technological breakthroughs and take crazy ideas seriously. Will its gambles pay off?

Eric “Astro” Teller is X’s Captain for MoonshotsDAMIEN MALONY

Gandalf arrives by rollerblades. It’s morning at X – formerly Google X – and Astro Teller (X’s Captain in Moonshots) glides past in a grey robe and pointed hat carrying oatmeal. Star Fleet officers walk past their desks, sipping coffee. Star Fleet officers queue for breakfast. This is not normal, as it’s Halloween. But X is a strange place. The block is surrounded by self-driving cars that loop around it. In the lobby hang sections of stratospheric, which are designed to transmit internet to remote locations. Robots move around sorting the recyclables. Teller compares X to Willy Wonka’s chocolate factory. It seems only right that there should be costumes.

It’s difficult to describe X even when you are standing in X, a former mall in Mountain View. It is part of Alphabet, Google’s parent company. However, in the metaphor “Other Bets”, X is closer to the gambler than Deepmind. Its stated goal is to pursue “moonshots”, which it refers to as trying to solve the great problems of humanity by inventing new technologies. X, which is now a separate company, Waymo, has developed self-driving cars (Wing) and internet balloons, Loon (Loon). It also built delivery drones and contact lenses that measure glucose (Verily), and technology to store electricity with molten salts (Malta). It tried to make carbon neutral fuel out of seawater and replace ocean freight with cargo-blimps, but it was ultimately defeated. One time, it debated the idea of laying a huge copper ring around North Pole in order to generate electricity from Earth’s magnetic field.

 

Although it might seem absurd or unbelievable, almost everyone uses something created at X every day. Google Brain, the deep learning division that informs everything, from Google Search to Translate started at X. Also, GCam camera software, which is used in Google Pixel phones, indoor mapping in Google Maps, and Wear OS, Android’s operating system, was developed at X.

These are irrelevant. These are symptoms. Teller states that side effects can be caused by trying strange things or doing things that are not likely to work. He keeps his rollerblades under the table, as he is a creative organisation. They save him eight minutes each day between meetings. He explains that X is not a company but a radical way to think, a method of pursuing technological advances by taking insane ideas seriously. X’s job does not include inventing new Google products. He is responsible for creating the inventions that could become the next Google.

X was once a joke in Silicon Valley (and Silicon Valley). Its self-driving cars have covered 10 million miles on public roads and run an autonomous ride-sharing company in Arizona. Loon’s balloons offer internet access to rural communities in Peru and Kenya. Wing, X’s drone delivery program, delivers food and medicine to Australian customers. Despite the fact that Alphabet is still being affected by protests from employees and leadership changes – in Dec 2019, founders Larry Page & Sergey Brin resigned, leaving the company to Google CEO Sundar Piichai – X faces renewed scrutiny to show that its moonshots were more than a luxury or costly PR stunts. 2020 marks the tenth anniversary of X. Will its gambles pay off?

X is located in a former Mountain ViewDAMIEN MAALONEY shopping center

 

Alphabet isn’t the only company to have set up a laboratory to pursue moonshot ideas. Bell Labs was founded by AT&T and Western Electric in 1925. It brought together engineers and scientists from various disciplines to improve the field of telecommunications. Bell Labs was the inventors of the transistor and the first photovoltaic cells. They also won nine Nobel prizes. From Xerox PARC, Lockheed Martin Skunk Works, and DuPont Experimental Station, corporate research laboratories have been a key part of the development of breakthrough inventions. There are corporate research labs at Apple, Microsoft, Amazon, and Microsoft. Google has many, including Google AI (formerly Google Research), Robotics at Google and Advanced Technologies and Projects. This group works on smart fabrics and AR.

Corporate research labs can be flawed. Companies that are driven by quarterly results often overlook transformative ideas within their own organizations. Xerox PARC created the graphical user interface. But we don’t work at Xerox laptops. As startups become corporations, bureaucracy may take control and their ability to think creatively decreases. Teller says that companies move from experimentation and process over a period of 20-30 years. “Processing is a way to achieve surprise down to zero. Experimentation means to immerse yourself in surprise. Both are impossible.

X is not a corporate research laboratory (it uses “Moonshot Factory”)), but its mission was unclear when it was established in 2010. X was originally a result of Chauffeur (Google’s self-driving vehicle project), which was led by Sebastian Thrun, a Stanford roboticist. Page and Brin admired Thrun’s Streetview work and turn-by–turn directions in Google Maps. They offered X free rein to allow him to pursue similar unconventional ideas. Thrun says that the original title of Director of Other was “Director of Other” at first. “We wanted technologies to be pushed in many directions, including self driving cars.

X’s existence was kept secret for at least a year. Keycard access was denied to other Google employees. Google employees were denied access to their keycards, despite the fact that bottom-up management is a principle of Google and employees can spend 20% of their time on their own ideas. X was free-wheeling and intellectually anarchic, even though Google has a culture of top-down management. Project Chauffeur engineers worked alongside Loon, Google Brain and a few other similarly ambitious projects. Thrun states that he wanted no paperwork, PowerPoints, financial reporting, oversight and no bureaucracy so that people could concentrate on the challenge. Page and Brin were the main contributors to many of the initial project ideas. They took an active interest in the building and eventually moved in. (Teller once described X as Brin’s “batcave”.)

Teller assumed control of X after Thrun left X to go online with Udacity in 2012. In many ways, he was the obvious choice. Edward Teller, the father of hydrogen bomb and co-founder of US’s Lawrence Livermore National Laboratory, is his paternal grandfather. His maternal grandfather, a Nobel-prize-winning economist, was his father. Teller says, “I was the dumbest in my family.” “My family believed being smart was all that mattered. These terms didn’t work for me. It forced me to look for other ways to succeed.” Astro, a nickname of Eric, founded an AI hedge fund and sold a company that made wearable sensors. He co-authored a book on relationship advice and has written two novels. To compensate for mild dyslexia at school, he would solve every problem twice using different methods. He says, “Then if the answer came out the same way, it was the correct answer.” He learned early on the importance of experimental thinking from his experience – to “just try things fast, approximate at the beginning and approach the problem differently.”

It had very little structure when Teller took over at X. It was like the Wild West. We started projects because we were interested. Obi Felten joined X in 2012 from Google. To formalize the moonshot, Teller hired Felten to work in Google’s product marketing. While engineers were pushing the boundaries of artificial neural networks, high-altitude ballooning and high-altitude ballooning technology, Felten claims she handled “literally everything that wasn’t tech.” Legal policy, legal marketing, PR, partnerships. They hadn’t had a business plan before for any of these projects.” Her job title was “Head Of Getting Moonshots Ready to Contact with the Real World”.

Some X projects did not survive first contact. Google Glass was the first, which was a wearable computer that fit inside a pair glasses. Brin was enthusiastic about the idea and pushed X to make the prototypes a product. Google received a lot of attention when Glass launched in 2013. Glass-wearing skydivers parachuted to the top of Google’s annual developers conference. They were worn by models at New York Fashion Week. They were featured in The Simpsons, and Vogue.

Glass received poor reviews in the real world and was mocked (“Glassholes”) as well as outrage over potential privacy invasions. Teller said that Glass’ real failure was when it tried to talk about it only as a learning platform. The public responded to it like a product. The worst part was that we talked about it in this way ourselves. It was not a finished product, and that was horrible.

 

In 2015, glass was no longer available as a consumer product. Glass is still available, but it’s used mainly in the manufacturing and manual industries. Teller states that sometimes it doesn’t work because the technology isn’t ready. In these cases, we need to stop doing it and pause it. Teller believes that eventually, a Glass-like device like the one he described will be popular. Apple is said to be developing AR glasses that will debut in 2022. “There is no way to take moonshots, and you can never be too late. We prefer to be too early than too late, based on our definition of what we do.

Nick Kohli, Loon’s flight operations Manager

Every week, X’s most brilliant minds meet in a conference room to kill each other’s crazy ideas. An idea that is considered a moonshot must meet three criteria. It must solve a major global problem, invent breakthrough technology, or produce a radical outcome at least 10X better than the current state of affairs. Although hoverboards and jetpacks are fun, they don’t contribute to a shared good.

 

Similar to the distribution of vaccinations, it is not a feasible goal. 

Teller says, “Everything must be on the table.” Teller smiles. It’s not a good idea, but it’s beautiful. However, the idea’s creativity is a plus. That person will come up with other great ideas for society.

After a moonshot has been proposed, X’s Rapid Evaluation Team – a rotating group of Xers with expertise in materials science and artificial intelligence – begins what is called a pre-mortem. Let’s pretend that everything went haywire. Phil Watson, the head of Rapid Evaluation, says “What are the causes?” Ninety percent of ideas fail during this stage. Some ideas fail because they are too costly or too difficult. Some break the laws. If an idea is difficult to kill, it becomes an investigation and is assigned a small group to further investigate it. Watson says, “We begin looking into it more systematically.” What would it take to make this work? What skills would be required to take it to the next level of success? What is the most likely thing to fail? Successful investigations are projects with a name and a budget.

“Monkey First” is one of the foundational tenets of X. If you are asked to teach a monkey how to speak, then you should not start with the easiest task, such as building a pedestal. The team should set performance targets and “kill targets”, which are thresholds that, if not met, will end the project. Project Foghorn was an attempt to convert seawater into fuel. However, it failed to produce enough fuel at a reasonable price. X killed the project and published the findings in a scientific paper. He also gave the team a bonus.

 

 

Kathryn Zealand from the Rapid Evaluation team follows me one afternoon to observe her at work. It is my goal to create a pair of assistive pants that can help the immobile and elderly walk independently. Zealand, an Australian, says that the space for ageing has not received enough investment. It’s expected to become a major trend in the future, according to demographics.

Smarty Pants are code-named Smarty Pants. They were inspired by soft robotics advances and Zealand’s personal experience with her grandmother, a 92-year old woman suffering from Alzheimer’s. Even simple movements such as standing can become very difficult at that age. They will take 30% more steps if you can help them with that one thing. Zealand says that the more people walk, the fewer other health problems they will have. One leg is covered in what appears to be 3D-printed armor. It’s wired with sensors that collect data about her gait and allow us to see how she moves.

X starts prototyping as soon as possible during any investigation. X’s Design Kitchen contains everything needed to carry out experiments in many fields. It includes a wet laboratory, milling machines and laser scanners as well as 3D printers. Zealand replies, “OK, we say, OK, what’s the fastest way to get us to a yes/no?”

We are greeted by a spacious, open atrium. Zealand has invited her mother, who is visiting, to test the system. Zealand says that she struggles with stairs. The prototype trousers is rough. There are actuators at each knee joint connected to fabric panels that wrap around the legs. Corset-style, the outer seams have a Victorian steampunk feel. A Raspberry Pi is used to control the motors. It comes in a pearlescent bag. The team from Zealand, which includes a deep-learning specialist and a clothing designer, fitted her mother in the pants, then monitored her as she climbed several flights of stairs. She is delighted to say, “It’s amazing,” as she steps down. “Normally, I would be out of breath by time I got up there.”

 

Zealand asks me if they would like to test them. After a quick fitting, I begin to walk cautiously on the first step. I immediately feel pulled up by another set of muscles. It is much easier to climb. Zealand explained that the trousers can “see” the stairs and know when to apply force. She hopes that soft robotics and material advancements will eventually allow for a lighter product that is half the weight. This could help with mobility issues. She says, “That’s probably ten years away.” It’s still early. Only half of X’s investigations end up as Projects. It will likely have been published by the time that this story is published.

 

Kathryn Zealand of X’s Rapid Evaluation Team wears sensors for Smarty Pants projectDAMIEN MALOY

X has a great advantage in being able to solve long-term problems. Teller says that some technologies require a lot of reliability in order to be safe. “There is a big difference between a 1% error rate and a 0.001 percent error rate.” While a software problem in a mobile app will not be fatal, one in a self driving car could.

As a Waymo driverless vehicle pulls up to the campus, the thought lingers. Waymo has logged over 10 million miles autonomously on public roads since its 2009 launch at X. Waymo has been operating in Phoenix, Arizona as a small-scale ride-hailing service. Jaguar is currently developing its next generation of vehicles. Morgan Stanley recently valued the company for $105 billion (PS80billion). Andrew Chatham, Waymo software engineer, says that Waymo’s goal “is to build the most experienced driver in the world.” It is not about building a car. Others are very good at this.”

 

We turn and pull away. Rick, a safety driver, sits in the front seat. However, the wheel turns by itself. The car is a white Chrysler Pacifica. Headrest displays allow for a live view from the top of the vehicle’s roof-mounted sensors. These include pedestrians in yellow wireframes and the purple outlines of other cars. The ride is quiet and uneventful, with the exception of a few hesitations at junctions when cars are still unable to predict driver intentions.

However, self-driving cars are still far from mass adoption. “If you look back at the first six months of 2009, there were some really great videos. Chatham states that we are now more than a decade later and the videos… it was just too easy to get there.” “Actually, getting to the point of deployment is another ballgame.”

 

X employees were able to work happily on technologies that could be decades away in the early years of Google, even though advertising revenue was flooding into Google. Thrun remembers asking Eric Schmidt, former CEO of Google and Alphabet executive chair, for $30 million to finance a project. Schmidt gave him $150million. Schmidt gave him $150 million.

One morning in 2015, Page and Brin announced that Google was undergoing a restructuring to become Alphabet. This shock news shocked everyone within the company. Reports were widespread that budgets were being cut. However, at X, it was clear what the team was doing: “[It] became more obvious that the goal is to produce new Alphabet businesses,” Felten states.

Projects that reach a certain size “graduate” to separate companies. Many, such as Waymo join Alphabet’s Other Bets. Several have been acquired or spun off by Google, including the renewable energy startups Dandelion, Malta, and Malta. After graduation, project leaders are promoted to executive positions and employees get a share of the company. Teller says, “When projects leave here they’re still not done.” There’s still a lot of learning to be done.

 

It isn’t always an easy transition. The original leaders of many X projects including Wing, Loon, and Waymo have left or been replaced since Alphabetisation. Wendy Tan White, vice-president of X, oversees projects at growth stages and says that if you want to accelerate something, it is very difficult to find someone who can do it. They would need to grow very quickly themselves.”

In its first five year, Alphabet has faced many controversies as an organisation. The company was hit by sexual misconduct allegations by its senior executives in 2018. In protest, 20,000 employees including many from X staged a worldwide walkout. Richard DeVaul (then head of Rapid Evaluation at X and one of the original founders of Loon) was one of the executives named in the New York Times. (DeVaul quit, apparently without any exit package.

Teller publicly expressed regret for the episode and his admiration of those involved in the walkout. He tells me that it made him believe in Alphabet and Google more. “It’s amazing that employees can say, “This is our company too, and that should reflect us.”

 

Widespread dismay has been expressed about Google’s participation in Project Maven (a Pentagon-based artificial intelligence project) and Project Dragonfly (a plan to launch an uncensored search engine for China). Both have been reportedly abandoned. These events have rekindled debates over the responsibility Alphabet, in Google’s famous phrase, has to not be evil.

Teller said that although these projects were not under X, he is deeply concerned about the ethics of his team’s work. After all, his grandfather worked on the Manhattan Project. “There have definitely been things that people brought up [at X] which are like, Nope, this’s evil’. He says, “We’re not doing that.” Other issues are more complex, such as projects that could lead to the loss of jobs to automation. Current X projects include the creation of all-purpose robots that can automate mundane tasks. Teller states that new technologies can cause concentrated harm and diffuse advantages. “If automation has 100 times more benefits than its downsides, then that leaves us with 99 percent positive. We owe it those who have had to deal with those problems, and it is a matter of public policy that we make sure they are taken care of.

DAMIEN MALOY

X’s early successes were a flurry, but X’s moonshots have not been able to capture the public imagination – nor financial success – in recent years. Only Dandelion and Malta have created a commercial product out of all the energy startups. Google recently merged Chronicle, the cybersecurity moonshot that was meant to create an internet immune system, into its parent company. Wing’s drones could transform the logistics industry but it’s difficult to see delivering burritos in Wing’s moonshot.

X recently increased its efforts to tackle climate change and other threats to humanity. Teller states that climate change is, by all reasonable standards, the greatest problem humanity faces. Many climate-related projects are in progress, including research on ocean health. Agriculture is the most important, but it’s not yet named. It’s our most basic need. It is one of the most important industries in the world. Teller states that it has the highest carbon footprint of any major industrial sector.

Engineers are working in an X workshop at the second floor on several blue boxy vehicles. They have stilt-like legs, and off-road tires. These are drones for farming. They can comb a field in groups and take hyperspectral photos of the topsoil and crops. They are currently being tested on farms in California. “Collecting millions of images of strawberries, each strawberry has a unique identification,” says Benoit Schillings (a bright Belgian who manages many moonshots at X). “Agriculture is a complex optimization problem. Schillings states that the current way to solve the problem is to simplify it: We’re going put hybrid corn on more than 10,000 acres. X is hoping to increase crop yields and soil health by analysing data and making suggestions.

 

The agriculture project is a typical X-moonshot: Take a large problem and use Alphabet’s huge advantage in computing power and intellectual expertise to solve it – creating a global company in the process. Schilling giggles, “Let’s solve Agriculture” is a bit too ambitious. “We accept the challenges that I believe very few other players would be able to take on.”

Moonshots can also be seen as a cynical attempt to control industries that do not yet exist. The global market for agriculture is worth trillions of dollars. Waymo is the operating system in every car and Wing the air traffic control for every package. It’s not hard to imagine. Google was actually a moonshot to map all human knowledge. Despite all the rhetoric of X about changing the world it is ultimately there to create new businesses – and profit – for Alphabet.

Teller seems to be unaffected by the idea that creating highly profitable companies and solving problems like climate change are at odds. Alphabet and X are essentially one and the same thing. He says that things that lose money tend not to grow in size over time while those that make money tends to increase in size over time. I see profit and purpose as complementary. They are mutually beneficial to me.”

Material used in the Loon Project is examined with a polarizing lensDAMIEN MALOY

Recently, X celebrated its tenth anniversary. The senior executives were in meetings trying to plan the next decade. Tan White states, “The world’s changing.” The cottage industries that X pioneered are those in fields such as self-driving vehicles. Both startups and governments are now using the concept of moonshots extensively. Venture capital funds such as SoftBank Vision Fund allow startups to take large-scale risks on their own.

X’s true impact might not become apparent for at least a decade. It has brought in substantial returns to Alphabet – Teller stated that the value of Google Brain alone has paid for several years’ worth of X budget – but it’s not clear if the companies it graduates will survive or become the next Google. The “Other Bets” division of Alphabet lost $3.36 Billion (PS2.57 Billion) in 2018. Teller states, “We must accept that some businesses won’t make the cut.” Almost all attempts to invent something end up failing. Real breakthroughs require immense capital, creativity, and perhaps most importantly, patience.

Moonshots can be defined as the attempt to create something truly radical. That’s hard. It’s hard to predict who will do the really important things,” Nathan Myhrvold (ex-director of Microsoft Research, founder of Intellectual Ventures) told me. But the flip side is that if you have those resources but you don’t make it happen, then we won’t know if that amazing technology was worth that much effort.

As Teller-as Gandalf sits in front of us, it is hard to forget about an earlier moonshot factory, the labs of Thomas Edison. Also known as The Wizard Of Menlo Park, The self-driving vehicle, or any of X’s other moonshots, may end up transforming society, in ways that we don’t yet know. It could save the world or make Alphabet richer and more powerful.

“The real test will be 15 to 20 years later, when all the dust has settled and we can look backwards. Teller asks, “Then what are we doing?” There will always be more insane ideas that are worth following. “The world has more problems than it can handle, unfortunately.”

 

Updated at 11:50 GMT, 17.02.20: The original article stated that only Malta had created a commercial product out of X’s energy startup. Dandelion also has a commercial product. Sebastian Thrun, not Coursera, also invented Udacity.

 

Leave A Reply

Your email address will not be published.