Landing Moonshots with Google's Innovation Chief Dr. Astro Teller

0:00The term moonshot has become so overused it's almost lost its meaning. For a moonshot to actually be a moonshot, it has to be beyond hard, stretch the limits of our imagination, and have a very high chance of failure. But if it works, if it works, it could change the world. That's where Astro Teller comes in. He's the co-founder and CEO of X, the secretive lab behind Google's most remarkable innovations. From pioneering self-driving cars, to bringing the internet to all people in every part of the world, even finding ways to better predict natural disasters. So how do they do it? Astro Teller may just be the patron saint of innovation because he has figured out some remarkable counterintuitive ways to inspire people, push the limits, and actually deliver on some incredible, well, moonshots.

0:57This is a bit of optimism.

1:02astro i'm you know i'm a huge fan of yours so the the fact that you're here and we're doing this together is just sheer joy and thrill for me likewise simon thank you you know you do what are affectionately known in the rest of the world as moonshots the organization formerly known as google x now x not to be confused with formerly known as twitter x two different organizations. But Google X used to be the secret lab at Google where they would attempt to do what are known as moonshots, which is crazy outlandish ideas that only a company with the resources and brainpower of Google could do. Were you there from the very beginning?

1:43I was. So what happened was that the founders of Google were thinking about what became Google X, even for several years before it existed. They recruited someone whose name was Sebastian Thrun to run the place as they were starting to get it up and running. This was almost 15 years ago now. And they asked him if he wanted a co-founder. He and I had known each other for a long time and had been looking for an opportunity to work together. I was running a hedge fund in San Francisco at the time. And Sebastian came and asked me, would you like to come do this thing? And I said, no, because I was really dedicated to my investors.

2:26And then he came and asked me again. And so I went to my board of directors and I said, okay, I'm going to say no again. But before I do, I just want to hear that you're all as serious about what we're doing as I am. And they said, what are you crazy Astro? Go do that. We got this covered. So I made my CTO at that hedge fund, the CEO, and I came and helped start Google X. Okay, so now let's get into the real nitty gritty. Thank you for that. But let's get into the real nitty gritty because I think every company writes on its fake values list that's written on the wall. And I say fake because, you know.

3:04I know exactly what you mean. It's just a list of stuff on a wall. But they all have innovation, right? Which I think is funny because every company thinks they're innovative, but most companies are not. And even companies that say, even the companies that are, that are driving to innovation, The innovations that they're thinking of are, I think, fair to say, even if they're good, small in comparison to the stuff that you're doing over at X and have always done. Just so that people understand some of the size of the thinking that's happening there, just so we can contextualize it. Give a couple of the brands that have spun off that people might know that came out of this brilliant lab.

3:44Sure. So some ones that people may be familiar with, Google Brain, which started at X, was arguably the original epicenter of what is now this modern day explosion of deep learning. And it was the first place where really large networks were industrialized, essentially. A year or two after we did that, it was going so well. In that particular case, we chose to move Google Brain back to Google. We decided that was the best way to harvest the value that was coming from that thing. Let me use one or two examples to show you the breadth. We started something which internally was called Chauffeur a long time ago.

4:26It was our exploration into self-driving cars. That project has grown up. It's graduated from X. It's now a standalone business inside the Alphabet family called Waymo. And so if you go to San Francisco now, you can get a ride from Waymo Cars, and it's a transportation as a service experience. There's just no one sitting in the front seat. And you don't even need to be whitelisted. You just download the app, order a car. It appears it takes you around. For listeners of yours who've tried it before, it's really exciting for the first four or five minutes, and then it's boring. And that's awesome.

5:03I've had friends who've taken Waymos in San Francisco, and I was like, what's it like? as if it's the first time on a plane. And they're all like, yeah, whatever. It's an Uber without a driver. And that is the highest compliment that I can imagine because inside of Waymo over this 14-year arc, we had to first solve all the safety problems. But then we had to solve like, hey, let's make sure we don't make the driver inside seasick now that it's safe for people outside the car. And then there were all these like third order effects, like the body language of the car. We don't realize that little things that we do when we're driving signal to people like, hey, I'm going to merge now.

5:43No, really, like leave me some space. All of these things, it turns out we have to get good at. Like where are you going to drop someone off when they're arriving somewhere? There are so many things. It's sort of an AI hard problem. It's been an incredible ride. And the fact that it's boring is the largest compliment we could ever get. I never thought about that, of course, because when you're merging into traffic, you sort of nudge your way forward, which says, you better slow down because I'm going. Whereas an AI model would just be like, oh, there's a space, I'm going to accelerate now, which as you said, it's almost what's so fascinating about it, it's like these automotive nonverbal cues.

6:22Right. So of course, that wasn't on the list of problems to solve till we were actually out in the world and people are saying, this is the thing we really care about. Okay, great to know, and then we can go solve that. Now, everybody wishes they were more innovative. And I think leaders everywhere are trying to, quote unquote, get their people to be more innovative. What have you learned after 15 years of being at X that you didn't know before, but is part of the real formula? Is it how you hire? Is it the culture? What is it specifically that you do that lets people dream big, but also know how to make that stuff?

7:02How do you get innovation out of people? So let me take a step back and make sure that we're all on the same page about what we're trying to do so that when I describe how we're trying to do it, it will make sense. So what we call a moonshot has three basic parts. Number one, there has to be a huge problem in the world that you want to solve, that you want to make go away. Number two, there has to be some kind of science fiction sounding product or service that you're proposing that we would make that however unlikely it is, we could actually make it, get it to work, get it to be cheap enough, whatever.

7:34If we could make it, we can agree it would probably make that problem with the world go away. And then you have to have a proposal of some breakthrough technology for how we can take at least a run at making that thing that's sort of at least an order of magnitude better than the solutions that are currently out there. If you have those three things, we would call that a moonshot story hypothesis. You're probably wrong, but it's cool by itself once you get to that stage. Then the question is, how fast, how cheaply can we verify that you're wrong so we can get on to the next thing? Because in the early days, they all sound crazy.

8:14We can't pre-pick which ones are the right ones. So the best we can do is just be better than other people at filtering the ideas. That's what we try really hard to be good at. And so if you, Simon, worked at X and you come up with the teleportation proposal and you have like, hey, I think I have a way of making teleportation work. And here's this crazy idea. If we can try it, but it's going to take four years and$10 million before we find out if we're on the right track. Nope, not interested. But if you say, hey, crazy idea. Here's how great it would be for the world if it worked. I know that it's probably not going to work.

8:51but I think we can learn something for a few tens of thousands of dollars in the next month or two, oh yeah, you've got our attention. I don't care how unlikely it is if we can learn for that cheap and that fast. So that's the game we're playing, is churning through ideas, throwing most of them away for good reasons with evidence. And so the most fundamental thing we're doing is playing the long game and trying to have really high amounts of audacity, so we're willing to go on these unlikely journeys, but then with equally high levels of humility, knowing right from the beginning that it's probably not going to work, and being able to be sanguine enough about it that we can be proud of going on the journey and then proud of turning off the journey when it turns out it's not the right journey, instead of resisting that reality or going through a long grieving process when it turns out not to be the right thing to do.

9:48And so all of the cultural work at X is about trying to help people get from this sort of gambler's mentality of innovation over into this card counting mentality of innovation, where we can feel together as a community supporting each other in this process where we are passionate about being dispassionate. We pursue each of these ideas, hoping it will work, but also realistic that most of them won't. and hungry to discover where we're wrong. This is a big deal and it's worth repeating. And you said it, which is moving from a gambler's mentality to a card counting mentality, which I think is a great way of putting it.

10:26Most innovation, especially in big companies, happens where somebody comes up with ideas or a group of people come up with ideas. Some executive who has to sign off on it wants the ROI, wants to know what the bet is, and they pick what they think is going to be the best bet. And if it fails, it probably takes too long, they write it off, and then they repeat that system. And it doesn't produce high quantities of great innovation. And as you said, you might hit it now and then, but it is a gambling. It was a little bit of luck or a lot of luck that it worked in the first place. Whereas what you're doing is you're creating a mentality of trying, not just evaluating on paper and taking bets.

11:06And you're allowing the reality of, is this possible or not, to dictate if we go affords or not. So egos kind of aren't really involved. You don't really have camps being pitted against each other that my idea is better than your idea because the data will prove whether it's viable or not. And I assume that it makes people much more cooperative that you're all trying to help each other make your ideas work because nobody's going to be rewarded because you're not gamblers. What you described is our aspirational state. The actual work that's going on here, and you've spent time here at the Moonshot Factory, so you know what I'm talking about, is trying to get our way step by step towards that aspiration.

11:46Because you're right, and it sounds great, but it's really hard for us to do. Because that's not what the rest of the world has encouraged us to be like since we were six years old. The rest of the world, if I say, I don't know what the answer is to your question, or I say, your idea is better than mine. And we can all agree abstractly that that's the right way for us to be and that the best innovation factory would have those kind of characteristics. But that is not how you jack up your career in almost all of the world. So getting the culture to reinforce all of these pieces that are aligned with the audacity, with the humility, with taking long time horizons.

12:27Part of the problem is we train particularly engineers, but arguably other skills too. If you're making version 17 of some satellite, I'm sure you know what you're doing. And so you just want to be left alone for two years so you can make the next version and you can pretty much predict what's going to make the next version better. And it's going to be incremental. It's not going to be radically better. If we're shooting over the horizon, the temptation is there in all of us to say, give me enough people, enough money, enough time, and leave me alone. Let me build it right the first time. I get that temptation.

13:04The smarter people are, the more that temptation is high in us. But when we're shooting over the horizon, you don't even know if you're going in the right direction. So how fast you head there is like not the question. It's like being able to work the sextant that's important, not how fast you row in your boat. So what I love about this is, I mean, so I know that we call moonshots moonshots because, you know, we always like we could put a man on the moon, but we can't dot dot dot. Like, I get it. It's like that we landed somebody on the moon in 1969. 1969, for heaven's sakes. It is the great things to point to of human ingenuity that we use as a point of reference.

13:44And though that's true, it's your description of those three characteristics that define how X goes about pursuing ideas. is it seems to mirror exactly what they did back in the 60s when Kennedy challenged everybody, which was it had to be for some greater good. And in his man to the moon speech, he says, in order to serve mankind, in order to do something great to advance the human race forwards, he literally talks about why we're doing this, the audacity that we're going to put a person on the moon and bring them back again safely. And then the trials and trials and trials to prove that I think we can do this.

14:21Like we have the rocket power. We can figure this out and give it enough time, less than a decade with a computer, the power of a Commodore 64. And we did it. So that's, I think the most impressive thing. It's not the accomplishment. It's that your three criteria match the actual moonshot. Yes, thank you. And we're adding on, I mean, so obviously we're trying to learn from all of the great work that's happened before us. Yeah. Bell Labs, Xerox PARC, the Apollo space mission, as you were just describing, Bletchley Park during World War II, and the invention of the computer. There's a lot to learn from.

15:02One of the things that distinguishes us from the eponymous moonshot is efficiency was not the goal. There was a national security and national pride issue, which trumped efficiency. and that helped and then made it sort of a special case. We do not have that benefit. We are and should be held to an efficiency standard. That's in principle what distinguishes or part of what distinguishes gambling from being a card counter. Like everything that we're doing here, yes, those three circles and the intersection, it has to have those three things, but it's also how are we going to pursue this as efficiently as possible?

15:42So it's not just, can we do it? Let me give you an example of something that I think is a learning that we've had that's very much in the culture here now. We call it moonshot compost. So we take a run at something. A lot of people are passionate about it. I'll use an example. We have a moonshot now for helping the electric grids of the world move into the 21st century in a variety of ways so that we can help them to take care of the grids as they are, plan for the grids as they want to be in the future, and operate the grids in real time in ways that will make them much greener and more resilient.

16:23That doesn't exist. We think we can make it. The moonshot we have now, which is called tapestry, is actually well on its way to doing that. And that is not our first run at that particular problem. And when we took previous runs at it, and they didn't work out, and I still remember the conversation I had with somebody internally where they talked about doing another run at it. And I said, fine, it's a huge problem with the world. I would love to see it fixed. Can we please make sure that we learn everything possible from all the ways it didn't work before? So when we take our next run at it, we're at least somewhat more likely to do it.

16:59We don't duplicate any of our past mistakes. And so that concept of moonshot compost is one of maybe 10 cultural memes here that people really get that when we try to learn in a space and it doesn't work, when we stop it, many of the people stay, the patents stay, the code stays, we maintain some partnerships. And so then when we try again, we're not starting from scratch. We have all of this learning and intellectual momentum that we can already build on, even if the project has a new name or kind of starting over. So again, this is something that, you know, even though it's the adage, you know, you either get it right or you learn a lesson.

17:41How many organizations are actually like pulling out the old data to be like, all right, we're going to try this again. Let's spend some time to sort of like go through so we don't repeat the mistakes. You know, a friend of mine who is a consultant says of her clients, they never have the time or money to do it right the first time, but they always have the time and money to do it again. Right. And I'm so glad you brought up never time to do it right, but always time to do it over. Right. That stuff gets brought up at X all the time because that's what they learn in the rest of the world. And measure twice, cut once.

18:11It's like a seamstress saying. When the most expensive thing you have is the fabric and you pretty much know what you're going to do, measure twice and cut once makes sense. But when you have no idea what the real problem even is, let alone what the solution is, it might be that there is no solution or that now isn't the time to find a solution that the tech needs to move forward another decade. Who knows? The whole concept of doing it right the first time is just out the window. Oh, that's such a, it screws with our heads, as you said, because we've had it drilled in us, you know. So that must be the hardest thing you have to do is when you're taking people who are not 12 years old and you're hiring them and you have to unteach them their incentive structures, their reward structures, their grade structures, everything that they did to get to the point of being smart enough, good enough, talented enough to get a job with you.

19:12And they're like, congratulations, now forget everything you've learned. You're going to try something new. That must be the single hardest thing you have to deal with. Exactly right. You know, you've probably heard that really great painters who break all the rules learn first how to follow the rules. Yes. We want the same thing in the people who work at X. If we're trying to do something really radical, we don't want people who are starting at zero. We want people with all of the skills to do things the normal way who also have the mental plasticity to put that aside, to limber up as individuals, but more importantly, as groups and to say, OK, now how can we be productively nonconformist, bringing our skills to the table, but not all of our biases and the habits?

20:02because as they say, the tools that got you here won't get you there, right? If you're going to bring to bear on some really big problem, the same things that everyone else has been trying, you shouldn't expect to get better results than everyone else is bringing. Can you give some specific guidance as how you actually do do that at a cultural level from a leadership standpoint? How do you actually get people to let go? So my experience is that at first you have to give them some intellectual architecture that gives them a boost while they sort of get into the flow of things. You can't just say, let it go.

20:42That doesn't work. Forget everything you've learned. Right. So an example, which I know you've heard, but many of your listeners won't. For many years, I was just saying we have to work on the hardest parts of the problem first. Yeah. And it just was clearly like not working for people at X. And then I was at a Wall Street Journal live conference and someone asked me about Just Like You're Doing and I said work on the hardest parts of the problem first. And they said, what does that mean? I don't get it. And I said, look, hyperbolically, if you were trying to teach a monkey to stand on the top of a 10-foot pedestal and recite Shakespeare, what should you do first?

21:19Train the monkey or build the pedestal? in that extreme circumstance i think we could all agree there's no point building the pedestal first because you've removed zero percent of the risk all of the risk is can you train the monkey right but when you actually look in slightly more subtle circumstances all over the world almost all the time everyone's building pedestals because then you can be half done you spent a couple years you got a bonus you got a promotion and if it's incremental if you kind of if there isn't a lot of risk and its possibility, it's just hard work, then it's fine to do the pedestal first because it's kind of all pedestal.

21:58But just me saying that, it became a meme here at X. And so now people literally put little monkey icons by the parts of what they're working on in slide decks that are like the make or break it part of what they're doing. And so that is an intellectual architecture that has helped people focus in on, are we really de-risking the hard parts now, or are we kicking the can down the road in sort of chickening ways? Yeah. Yeah. And you're right, because egos are fragile, right? And we've all been in this meeting where we spent two years and spent$10 million building the pedestal, that because we don't hurt anybody's feelings.

22:40We go, great job to the Monkey Shakespeare Project team. Brilliant job building this pedestal. That's exactly what we envision with this perfect pedestal. We just want to say good work to everybody. Keep it up. And we do that. We sort of - Halfway there. Halfway there and just want to, just a quick high five to the Monkey team for building a great pedestal. And we do this in the meetings because for good reasons, we don't want to be like, what the hell are you doing building the pedestal? Like train the monkey. Right. I'll give you another example. When, when someone cancels their own project, which is arguably the pinnacle of what we're trying to do here.

23:21It's very hard. It doesn't happen all the time. But when people do that, we stand them up at an ex all hands meeting. Yeah. We give them a standing ovation. We give them all very large bonuses. and we make a point to Xers every single time, this is the rate limiting step of innovation. Not coming up with ideas. I don't think anyone's much better than random at coming up with ideas. Not killing off bad ideas. Anyone can kill off bad ideas. It's killing off the pretty good ones. The ones that still feel possible and we want them to work. But when the project lead says, you know what, the reward risk ratio of this thing, now that we are two years in, we've learned a lot, it's just not as good as some of the other stuff going on at X.

24:10And I think we should stop working on this. I hope I find another project to work on. And that's, of course, better for X, and it's better for them as individuals. So that's another example. But it goes into compost, right? Because if you came up with Netflix 20 years ago, the technology was DVDs. That's the best you could do. It was an upgrade from videotape, but that's the best you could do. So you would put it on the pile, and then as the internet matured, you'd be like, okay, let's take it off the pile and try again now that we have the internet. Exactly, exactly. And that happens regularly here that we're too early.

24:44And sometimes we're too early and we come back to it. That's happened quite a few times. Sometimes we're too early and the world comes back to it. Google Glass, for various reasons, arguably we were too early, but we were exactly right. There's now a whole bunch of work and products being sold that are very much in the vein of Google Glass. So, you know, we were prescient, but our job is to, arguably our job is to be the right amount too early. But in order to be the right amount too early, we not infrequently have to be too, too early. Are only companies like Google able to do this because there's a lot of failure, which means there's a lot of waste, which means you need a lot of money.

25:21you know how is this methodology or mentality transferable to a startup or a lesser resourced organization like my my little company i don't think i could afford to do this or could i i mean i don't know that you could afford to do it at the scale that we're doing it but if you said to your team i would like to whatever it is i'd like to have 10 times as many podcast listeners a year from now. Instead of having one idea about how we're going to do that, we're going to spend a month coming up with 100 crazy ideas. Then we're going to spend two months beating the crap out of all of those 100 ideas.

26:02We'll have used up three of the 12 months that we have. Then we're going to spend another three months on the top four ideas and give them each a chance to like prove their weight. And then we're going to push the last six months, the best of those four ideas. I guarantee you, you would make more progress on building your listenership for your podcast doing that than if you picked some idea and just hoped you were right from the beginning and pushed really hard for 12 months. So that is exactly the process that we're doing. You do realize that one example, you have screwed a lot of people's heads right now.