HANNAH BATES: Welcome to HBR on Leadership, case studies and conversations with the world’s top business and management experts, hand-selected to help you unlock the best in those around you.
Astro Teller says there IS a vetted approach to creativity and decision-making for innovative projects that he and his teams use at X, Alphabet’s R&D engine.
Teller is the captain of moonshots at X, which he helped launch at Google in 2010. His mission there is to invent and launch new technologies that address serious problems in the world and create the foundations for large, new businesses for Google’s parent company, Alphabet. They’ve worked on a pill that detects cancer, cars that drive themselves, and mega-kites that collect wind energy.
In this episode, Teller shares key lessons for managing the process of developing breakthrough innovations.
You’ll learn how he decides to keep investing in a project, and how he knows when it’s time to pull the plug. You’ll also learn what qualities he looks for when he hires for innovation projects and assembles teams, and why humility is so essential.
This episode originally aired on HBR IdeaCast in March 2023. Here it is.
ALISON BEARD: Taking big swings isn’t always easy in business, especially when you don’t know what will pay off – or how long it will take. Today’s guest has a high tolerance for that kind of uncertainty – exploring big problems, experimenting with solutions, failing, trying again, sometimes succeeding, sometimes not.
Astro Teller is Captain of Moonshots at X, Alphabet’s dedicated innovation factory. He helped launch it after cofounding a number of other companies, teaching at Stanford University, and studying computer science there and at Carnegie Mellon. His teams at X work on everything from getting remote populations online, to monitoring ocean health, and to using machine learning to improve supply chains.
I spoke with Teller during a live virtual conference – HBR at 100: Future of Business – where audience members were also able to submit questions. Here’s our conversation.
ASTRO TELLER: Thanks for having me.
ALISON BEARD: First I have to ask about your name. It seems like kismet that someone called Astro would become the head of a moonshot factory. So how did that happen?
ASTRO TELLER: I agree, it seems like fate, but it was a typo at Stanford. I didn’t want to leave any blanks on my application to Stanford. I don’t have a middle name, so I felt like an idiot that I wrote my last name, comma, first name. I had to leave a blank because I don’t have a middle name. Almost nobody called me Astro in high school. It was a not friendly nickname from the soccer team. Basically they thought my flat top looked like a patch of AstroTurf, so I wrote in Astro. And then I think there was a type-in error because I didn’t have a middle name, and so someone typed in Astro as my legal name, and just stuck.
ALISON BEARD: And the rest is history. So, you co-founded X within Google, and the mission is to solve real problems and have a real impact, not create gadgets or technology for its own sake. So, how do you identify the problems that you want to work on?
ASTRO TELLER: So, as you were just saying, X’s mission is to invent and launch breakthrough technologies that can help tackle a huge problem with the world and create the foundations for large sustainable businesses for Alphabet. As a result, our remit is very wide. It needs to be something that can really help the world, something that can be good for Alphabet, but there’s no specific industry that it’s in. We’ve sort of caught different waves over time. The wave that I would say X is in right now, we’re largely focused on sustainability, but it used to be robotics more, mobility.
And in any particular area, really we say, what is an idea that sounds like science fiction but would be really important if it turned out to be true? And how cheaply can we ask the question: is this just a bad idea or is it once in a generation opportunity?
And we look at thousands of these things every decade. That’s our job is to sort of have a very wide funnel and then to filter very aggressively. So, at the wide part of the funnel, we’ll look at almost anything as long as it has those basic characteristics: that it could be a breakthrough technology, it could help the world tackle a really serious problem, and build a foundation for a large sustainable business for Alphabet.
ALISON BEARD: And so how does that filtering work? How do you narrow down with so many good ideas?
ASTRO TELLER: At the beginning, we’ll try anything that has those characteristics. And on day one, we don’t need anything except that it has those characteristics. It’s a hypothesis to test. But afterwards, what we’re saying is for every dollar that we put into this machine, we don’t care if the answer is yes or no. The answer is no for almost everything that we look at. The question is how cheaply, how wisely can we get to the answer? Is this a great idea or one of the bad ones? And again, most of them are bad ones.
And so, we’re always looking for evidence. How can de-risk this? How can we learn, turn uncertainty into risk? Because we don’t even mind risk. But what we’re buying down, especially in the first couple of years of this thing is, what is this really? What does it want to be when it grows up? What are the hard parts actually about this thing? And that how can we get it into the world very early on so that we start to learn even faster about the ways in which it might not work, so that we can kill it and get onto the next idea?
ALISON BEARD: Related to that, I’ve read about a project management concept that you call monkeys and pillars to help you make those decisions. So, tell this audience about it.
ASTRO TELLER: So, I was on a conversation very much like this one. It was about seven years ago, and up until then I had said we have to work on the hardest parts of the problem first. And that made sense to me, but for whatever reason, it hadn’t really taken at X. And so, the interviewer, like you are now, asked, “What do you mean by that?”
And so, I gave this hyperbolic statement, which was, let’s say that you’re trying to train a monkey to stand on the top of a 10-foot pedestal and recite Shakespeare. Which should you do first, train the monkey or build the pedestal? And so, this has become a joke inside X, but because it’s easy for people to remember.
In that extreme case, it should be obvious to all of us that if you build the pedestal, you could be like, “Hey, look boss, I built a pedestal.” And the boss would be like, “Hey, good job, Astro.” But you haven’t actually made any progress. There was no chance that you couldn’t make the pedestal. All of the risk was on training the monkey. So clearly what we should do first is try to train the monkey because if we can’t, the pedestal is a total waste of time, and if we can, we can always build the pedestal afterwards. So, something about that hyperbolic statement has become a meme inside X, and people actually put little icons of monkeys next to the parts of their effort, which they believe are the really critical parts to push on to understand whether or not this could actually be a really once in a generation opportunity.
ALISON BEARD: Okay. So, let’s assume that people are doing a good job of beginning to train the monkey. You have some positive results, you have some negative results. How are you deciding when a project needs further investment, you’re green lighting it to keep going, or that something should be killed?
ASTRO TELLER: It depends, I have to be honest with you, there isn’t a single answer, but let me give you some of the kinds of things that we look for. If a team says, “We know what the right thing to do is. Just leave us alone so we can build it.” And that’s so the wrong answer when it comes to moonshots, that we might stop it just because they said that, and we will definitely stop it as soon as they turn out to be wrong, which they inevitably will. A team that shows up saying, “We’re going to audaciously try thing after thing,” but say from the first moment, “We’re probably wrong,” they have a better chance of turning the loop faster, so we’re going to bet on them longer.
Sometimes projects are inherently slow in their learning loops because of what they’re trying to do. And when that happens, all things being equal, it seems like less of a good bet than a project that has somehow figured out how to get into the real world and learn something every single week. Our experience is the faster you’re learning, the more likely you are to be successful, kind of independent of how things are going this month, even this quarter. So, it’s really a measure of learning per dollar that we’re getting. The ones where the learning per dollar is high, we tend to keep betting on, and the ones where the learning is low, even if the progress looks good, we tend to slow down or just stop.
ALISON BEARD: I like that, learning per dollar, it’s a new metric. So, we’re going to dig into some of the specific projects that you’re working on in a little bit, but more generally, what sort of time horizons are you looking at when you are thinking about a successful spinoff or that a project has been completed in X’s term, and is ready to move on to the next thing?
ASTRO TELLER: Ten years is sort of what we say here at X, and that allows for us to play the long game and there’s a lot of incremental value that can be produced, and a lot of incremental goodness for the world that you can go after when you play the long game. 10 years though, I mean, I would say on average things that ultimately graduate from X to become other bets, our most recent other bet for example was minerals in the computational agriculture space. It’s our moonshot for agriculture. It was here for about six years before it graduated. So, by saying 10 years, we don’t necessarily mean that it will be at X for 10 years. We mean that certainly within 10 years, the thing that we are starting from a cold start should be pretty interesting and important within 10 years.
ALISON BEARD: So, I’d love to just better understand your place within Google and Alphabet because I think other large corporations – or even smaller ones – can learn from it. How do you have an incubation factory within your own company? So how does your funding work? How much interaction do you have with the rest of the company, and then how do you do that spinoff process?
ASTRO TELLER: We’re not the only source of innovation at Alphabet of course, but we are an innovation engine for Alphabet. So, our job is to help Alphabet, Google’s parent, have new problems and hopefully find new solutions to those problems. Over time, some of the ones that we created earlier on like Google Brain, Verily, the life science business for Alphabet, Waymo, the self-driving cars, Wing, the drones for package delivery. More recently, Intrinsic, an attempt on our part to democratize how the manufacturing process works and the automation of robotics in manufacturing.
As I was mentioning, we’ve just recently spun out this new effort in and made it a company in Alphabet, Other Bet as we call it, in the agriculture space. In each of these cases, these are still nascent businesses, and what we would hope over time is that at least some of these become large, important, good for the world and valuable to Alphabet. So we care very much what’s happening at Google, but we’re like a little sister to Google on the side, trying to make things that will ultimately be important to Alphabet and help Alphabet to continue to grow and do good things for the world.
ALISON BEARD: So, let’s talk about talent. What kind of people are you looking for to help you with these moonshots, and is there enough of it around?
ASTRO TELLER: I think that there’s an incredible amount of it latent inside people, but finding people who have unleashed themselves is pretty hard. We think about this a lot, and we actually spend a lot of time and energy, even once we’ve hired people, helping them to unleash themselves. So really a lot of our interviewing process is about trying to decide if people are ready to unleash themselves rather than that they’re sort of done or perfect in any way.
The top four things that we look for: fearlessness, which tends to map to audacity and creativity and things like that. Humility, because audacity, fearlessness is critical so that you will try really out there things, but then you need humility to be able to say right after you start trying it, “You know what? This probably isn’t going to work. Let’s use evidence, verify that it isn’t one of the really great ones so we can throw it away and get onto the next one.” Teamwork, because innovation is fundamentally a team sport. And then a growth mindset. If we’re trying to build learning machines inside of a moonshot factory, if each of the teams is supposed to be a learning machine, then we need each of the humans here at X to be a learning machine.
ALISON BEARD: You have what I imagine are very brilliant, creative, probably a little quirky people. How do you decide who will work well together in those teams that you’re talking about, and do they require a different style of management?
ASTRO TELLER: The style of management is somewhat different. I would say that the difference probably is even bigger at the sort of X level. We don’t have org charts the way you would think of normal businesses working. There’s a lot of fluidity inside of X. Because if you were to come to X and start, I don’t know, flying car company or whatever it is you were trying to start here at X, you’re going to turn out to be wrong. Almost everybody is almost all the time. So then your thing stops, and then you’re going to find a new thing to be a part of. And so that fluidity kind of ruins the sort of hierarchy and politics that often goes on inside of groups.
So, I think of myself as a culture engineer, and a lot of the way that X is wired is if you ask people to do a bunch of really basic things like play the long game, show up with a lot of audacity, but also a lot of humility. If you’re asking them to practice running these experiments and then being intellectually honest about the experiments after they’ve run them, this is all really simple stuff. It’s easy to say. And just like a diet, actually practicing the diet is super hard. Everything I’ve described, even at X, ferociously hard to do.
And so, everything at X is wired around trying to make you not feel stupid about actually showing up humble and open-minded, with a growth mindset. Why are you going to kill your project if you think that your bonus or your ability to get promoted or the next thing that you’re going to get to do is going to be harmed by that intellectual honesty? Which is why there isn’t a lot of intellectual honesty floating around. And so we are like back at basics all the time, saying what do we need to do to send the hundreds and hundreds of signals necessary so that everyone at X naturally does the things we’re actually asking them to do, and that they tend not to do at most other businesses?
ALISON BEARD: Yeah, so I mean, if you’re struggling with it, you can only imagine what it feels like at more traditional organizations that want to be more X-like. So, what advice do you give leaders of other organizations, particularly outside the tech sector, about how to develop the kind of culture you’re talking about?
ASTRO TELLER: I think it really comes down to, A, how serious are you about the thing that you want? And then, if you’re really serious about it, then you have to commit to the practice of actually making people feel good about doing it. So, here’s my one-hour innovation lecture in 60 seconds. Choice A, choice B. Choice A, you can give a million dollars of value to your business this year guaranteed, or choice B, you can give a billion dollars of value to your business this year, but it’s not guaranteed. It’s one chance in 100. A, million guaranteed. B, billion, one chance in 100.
I’ve done this all over the world and I say, “Who’s choosing choice A?” Nobody raises their hand. “Who’s choosing choice B?” Everybody, big smile on their face, raises their hand. And I say, “Okay, now leave your hand up if in your wildest dreams, on their best days, your manager, your CEO, your board of directors supports you choosing choice B, even kind of a little bit.” And every hand in the room goes down and then I say, “You don’t need a lecture on innovation. You need a new manager.” This is the problem, is everyone asks for innovation, but they’re not actually willing to support the innovation because innovation is mostly about making mess. And you can try to do it efficiently, that’s what we try to do, but you can’t make the mess go away and almost nobody is actually tolerant of the mess.
ALISON BEARD: So how do you get leaders, managers to be more tolerant of the mess?
ASTRO TELLER: How badly do you want a factor 10 increase in value? The reason everyone raised their hand for choice B is because it has 10 times the expected utility of choice A, and that’s what innovation is. It is literally worth that much more. So I guess you have to decide whether you know want your 10% improvements or your 10X improvements. If you want the 10X improvements, you have to take a really long time horizon.
You have to have a portfolio because you will only get the payoff, the expected utility payoff over long periods of time, over a wide range of things. And then you have to be able to help everyone there be in it to sort of do the card counting. We’re not going to be gamblers of innovation, we’re actually going to be card counters of innovation, following a process and trusting that that process over very long periods of time will get us that 10X that we’re looking for.
ALISON BEARD: Before we get to audience questions, I want to ask you rapid fire about some specific projects that you’re excited about, because everyone wants to know what the next X moonshot is. So, Chorus.
ASTRO TELLER: People have been trying for 30 years to track all the physical things in the world so that we can improve the logistics supply chains, and it looks like we might have a way to do that that is much less hardware intensive, and that would be transformative for the world of logistics and supply chains.
ALISON BEARD: Yeah, particularly coming out of the COVID-19 crisis. So, Taara…
ASTRO TELLER Yeah. We have a way of shooting a laser up to 20 kilometers. It’s eye safe, so you could just go up like this and it still wouldn’t hurt you, and it moves information at 20 gigabits per second. So, you have to have line of sight between these two things, but you can just strap them to two poles as long as they can see each other. If a bird flies in between, then you lose one 1000th of a second of data, and it’s less than 1% of the cost of trenching fiber. We’ve been rolling them out for the last two years in Africa and India. We’re really excited about that one.
ALISON BEARD: Very cool. Okay: Tidal.
ASTRO TELLER: Ocean health. Fundamentally, humanity gets several trillion dollars a year of value from the oceans, and we’re killing the oceans faster than we’re killing our land or our air. We have to stop. And because humanity needs the oceans and derives so much value there, we have to somehow get more value from the oceans while regenerating the oceans. And that’s not going to happen unless we take automation to the ocean so that we can understand it and so that the value that we’re producing in and with the oceans is healthy for the oceans. Now, we’re starting in aquaculture, but we have a lot of other ideas about the maritime industry, about blue carbon, et cetera.
ALISON BEARD: Okay, last one before we go to audience questions. Tapestry.
ASTRO TELLER: That’s X’s moonshot for the electric grid. If you want to be able to plan, build and operate a clean, resilient electric grid, you have to start by understanding your grid. And the grid worldwide is the most complex machine that humanity has ever made. It is literally the case that it is so complex that there isn’t currently a digital map of where every wire is and where every transformer is, even for the people who are running the grid. So, when someone asks, “Hey, can I put this new solar field onto the grid?” The reason that they’re waiting in a five to 10 year line waiting to be added to the grid is because the grid operators, who are responsible for keeping the grid safe, don’t know what will happen if they plug that solar field onto the grid. So, we are trying to make the digital tools, the virtualization of the grid that will allow grid operators around the world to actually understand their system, play what if games, and ultimately operate their grid much faster in a sort of 21st century way.
ALISON BEARD: We do have a lot of questions from the audience. I’m going to try to get to as many as we can. Lizette from Cape Town is asking whether the monkey and pedestal approach can apply to other less ambitious projects as well. Should you always start with the most difficult part first?
ASTRO TELLER: Only if money is precious to you. I don’t know what to say. Look, if you know you can succeed, if you’re making a 10% improvement on something that already exists, then everything’s the pedestal. There isn’t a monkey. So maybe the order doesn’t matter very much and do whatever will get you the bonus first. I don’t know. But if what you’re doing has a lot of risk in it, if it’s a moonshot, if it’s a 10X opportunity, not a 10% opportunity, you’re probably going to be wrong and you’re going to have to stop entirely or pivot dramatically.
The faster you find out that you’re on the wrong track, the thing is, learning is not driven by success. You learn nothing when you succeed, except maybe to do that again. You learn exclusively when you fail. You have a model of the world and you find out you were wrong. And so failure is learning. They’re identical. So you should chase that if you want to go fast.
ALISON BEARD: Okay, so more talent questions. Lots of people are wondering how to really unleash talent in the way that you do at X. Gabby from New York City says, how do you help your employees do that? Juan asks, how can managers encourage people and teams to unleash themselves in more traditional organizations?
ASTRO TELLER: I want to be fair. Unleashing yourself in a traditional organization is hard if the organization, in being traditional, doesn’t totally want you to be unleashed. I wear rollerblades all day every day at the office. They’re on my feet right now. And I do that, it’s fun. But I also do that to remind people I don’t take myself seriously. I don’t take anyone else here seriously. We’re having fun together because fun and humor are the wellspring from which creativity comes.
If you can’t embrace silliness, if you can’t acknowledge that we’re all a work in progress and that most of why we waste time at work is fear, and we can’t get past fear until we can understand why we’re afraid and get really vulnerable with each other. It’s all just going to be like suit and ties and wasting time. I don’t know what to say. So I guess if you are a manager and you really want people to be unleashed, you need to first put down all your armor, take off all your masks, and then you need to start rewarding people when they do it.
ALISON BEARD: Okay, so this is related again, if people aren’t necessarily unleashing themselves yet. Sandeep from Cincinnati, Ohio is asking how do you train them? He says, I’m curious to know what training looks like at X.
ASTRO TELLER: There’s a lot of different parts of that. There isn’t a single answer, but for example, we have a program here called Thrive where we take people who we think are ready through a nine, 10 month process. It costs us a lot. We can’t do it for everybody, but we do it for a non-trivial fraction of the people here. And it is about helping them to understand better what’s holding them back, their limiting beliefs, the ways in which their fear shows up in controlling them and stifling their creativity, their audacity, their humility, sometimes. A hundred percent of us have things to get over, and it does take support for people to learn that. But I think really, I mean, there’s a lot of training. I can give you other examples, but if you want it and you don’t actually do it yourself and reward it, I don’t think you’re going to get it no matter what training you give to people.
ALISON BEARD: Terrific. So, Yusuf from Pittsburgh, Pennsylvania is trying to get at this idea of tackling these incredibly challenging problems that others have failed at, that you might fail at as you’re experimenting. So, what is your advice to companies when they’re in a situation where they need to make a decision on an idea to invest or not, knowing that people have failed before and that failure is quite a possibility?
ASTRO TELLER: Well, I’m going to answer the question under the assumption that this is a company that’s placing lots of bet, and this isn’t a small startup that where that’s their only bet. But assuming that this is a more sizable company, it’s placing lots of bets. And so it’s trying to decide, should I place a bet here? What I would say is, number one, good chance you’re wrong. That does not mean don’t do it, but just start with that in mind. Then number two, go learn why those other ones didn’t, and make sure that you at least make a new interesting mistake when you fail, instead of making the same mistake somebody else already made. What a waste that would be. I also think there’s a lot of benefit to failing more than once in the same area. When I look at the things that we ultimately make successful here at X, there’s so much moonshot compost that goes into them. Because when a project ends, the people stay, the code stays, the patents stay, the learning stays.
So, in ocean health, in agriculture, that’s not the first time we tried. That was like the 10th time we tried. And so, we actually have these sort of reverse org charts of this moonshot compost and all of these different ideas and people and how they ultimately culminated in something which looks like a really good idea and in a super exciting business. But that’s because we kept failing and keeping all of those learnings. So, if you’re only willing to fail once and learn once, and then you’re just going to run off to another space, I think it’s harder.
ALISON BEARD: So, I have to end by asking about generative AI. Your first novel, which you wrote when you were in your twenties, was about Edgar who is sentient AI. He basically becomes HAL and starts talking back to his creator. So, what do you think of the recent developments? How close are we to that Edgar sentient AI character?
ASTRO TELLER: Let me answer like this. I think that computers have been levers for our minds for a long time, and robots increasingly are becoming levers for our bodies, doing physical work for us in lots of different situations. And I see a lot of ongoing opportunity just here at X to use artificial intelligence and machine learning as raw materials that go into trying to make the world better. And in the same way that if we were 100 ago, and electricity was relatively new on the scene, everyone would be excited about electricity. And rightly so, but electricity isn’t the end of the story. Electricity is the beginning of the story. If we can now put more intelligence into things that we’re making, great, then we can find better and better ways to solve huge problems in the world. That’s how it feels at X.
HANNAH BATES: That was Astro Teller, Captain of Moonshots at X, in conversation with Alison Beard on HBR IdeaCast.
We’ll be back next Wednesday with another hand-picked conversation about leadership from Harvard Business Review. If you found this episode helpful, share it with your friends and colleagues, and follow our show on Apple Podcasts, Spotify, or wherever you get your podcasts. While you’re there, be sure to leave us a review.
When you’re ready for more podcasts, articles, case studies, books, and videos with the world’s top business and management experts, you’ll find it all at HBR.org.
This episode was produced by Mary Dooe, Anne Saini, and me, Hannah Bates. Ian Fox is our editor. Music by Coma Media. Special thanks to Rob Eckhardt, Maureen Hoch, Erica Truxler, Ramsey Khabbaz, Nicole Smith, Anne Bartholomew, and you – our listener. See you next week.
