Obi Felten: How Google's X Wants to Change the World

I think I've learned a lot about how to deal with failure and accepting that it's inevitable.

At X we talk a lot about failure because we work on moonshots. So by its definition they are high risk and daring and audicious. So many of our projects fail, especially in the early stages. My job at X is to run foundry which is the earliest stage of export acts. I start several projects a year and about half of them close down again. So I'm constantly starting new things and I'm constantly thinking about, you know, which ones should survive, which ones should close etc.

We set at the beginning of each project success criteria where we define what would happen in the world. How would the world be a better place if we're successful, right? What if we give internet to everyone in the world? What would that enable? But we also set very clear failure or kill criteria. And the beauty of that is if you agree those upfront, and often the team writes their own, so I want them to write the list. If those criteria then fulfilled it becomes a much less emotional decision to close the project down. And it's a decision that the team can do themselves. So sometimes those are technical criteria like we want a certain efficiency in a process. Sometimes they are economic. And the example where that worked really well was a team that was working on a new kind of fuel. The project was called Foghorn and the idea was to make a carbon-neutral fuel by extracting CO2 from seawater combining it with hydrogene and making a fuel that you can burn in a conventional car. And when the car burns the fuel the CO2 is released back into the atmosphere, so you have a closed carbon cycle. And the team focussed both on technology, they made a lot of progress on the technical side. They actually built a prototype together with Xerox PARC. I have a little viol of fuel that was made by the prototype that sits on my desk back in Mountain View. But they also had a kill criteria around the economics which was we wanted to make the fuel price-comptetitive with conventional fuel because nobody is going to pay three times as much for the petrol in their car. And that was actually a thing that killed the project. Because we started the project when oil was at 100 dollars a barrel. And when oil hit 30 dollars a barrel the team looked at it and said: We can never make these economics work. And so they decided to close down their own project. So what's interesting I think is what happens after that. So first of all I gave them all a bonus for their brave decision to shut down their own project. And I remember when I first raised this the HR team said: Really, you like rewarding them for failing? And I said: I'm rewarding them for the decision and for the rigorous analysis that went into this. The fact that the project didn't survive is beside the point. This is exactly the kind of behaviour that I want to encourage and that kind of braveness and honesty that I want to encourage. Then they presented at an X All Hands, they talked about all the reasons, all the things they explored, all the reasons why the project wasn't going to continue. And afterwards their colleagues came up to them and they said: That was incredibly honest and incredibly brave, thank you for sharing these lessons. And then they also wrote it up as an article that they submitted to a peer review journal. And so I have the secret hope that maybe this project is only deep-frozen and not really dead because somebody might read that article and say: Oh, here's an idea that, you know, X overlooked and here's how we could move it forward.

A start-up is constantly living with the fear of running out of money and literally being obliberated entirely. So they try and get to something like a minimum viable product as fast as possible, even if that minimum product is a long way away from their original vision. We actually want to tackle the toughest problems first because we want to find out as quickly as possible whether it is possible to go all the way to the long-term vision. So I think that's a sort of a somewhat fundamental difference. And it's way contra-intuitive because we are inclinded as human beings to work on things that we know we can solve, we make some progress, we feel good about that. But to as that doesn't make any sense because if you can find out quickly that this project is never going to work that's better than spending a year solving all the easy problems to then fail on an impossible problem a year in. And I have an image for that. We sometimes talk about training a monkey on a pedestal, so imagine you want to train a monkey to recite Shakespeare while standing on a pedestal, where do you want to spend your energy? Do you want to spend your time building the pedestal or training the monkey? So in that example it is very obvious that you should spend all your time on the monkey because you already know how to build the pedestal. But that's not what people do.

How do you make the car drive itself? How can we make the balloon float 100 days so it can actually provide internet connectivity in a meaningful way? But there's a matter moonshot in this, and that's actually building the moonshot factory itself. So we talk about X as a moonshot factory and it's almost a deliberate contradiction in terms because people think you can't make a factory of innovation, right? Innovation is weird and special and unpredictable. We actually think that some of the aspects of innovation you can factorize. So that's both a process like I sometimes think about the X as a moonshot factory, so I put small projects in at the beginning, and then something pops out at the end? No, like Verily, our life science company that's spun out a year ago or Waymo which spun out of X at the end of last year. So there is definitely a sort of factory process for the products to go along with it. But there are also factory processes in a sense that there are certain things that we have learned about how to run these kinds of projects that we can apply to the next one. So my team is called Foundry, it's the early stagers, so I have constantly projects starting, I have constantly projects moving on. And we have learned a lot for instance about the cultural transition that the projects go through as they go from a research experiment and Rapid Eval which is the team that thinks of new ideas that X should be working on and all the way to a team like Waymo which has become its own company. And there are different shocks that happen as you're going through these transitions for instance, just very concretely for me like when I take new projects in they switch mode, they switch from being sort of a research experiment that thinks a little bit about where my market would be, what product would I have to running much more like a start-up. While they are in Foundry they have to think like start-ups. Who is my user? Why do they want this? What problem exactly am I solving for them? That whole sort of user and market view comes in during the Foundry stage as well as progressing the technology and buildung a tech prototype. And how do you do that and how do yo get teams to go through these transitions? For instance that's someting to me that you can productize. The other part is what I talked about how do you deal with failure? You can make very deliberate choices about not penalizing teams for failure, how to reward honesty and decouple some of the personal success from the project success.

It's not that Europe can only learn from California. It's actually a two-way street. And there are certain things that California is not very good at, that Europe could really bring to the table. One example is European cities have been growing for hundreds of years. They have tons of experience of how do you live with millions of people in a very small space. So things like how do you optimize multiple different forms of transportation to co-exist is something that a European city should be much better at than an American city which primarily, with a few exceptions, but primarily depends on cars to get people around. But the starting point for the cities is very different. So we get these city planners from Europe come over and they say, we're writing the 2050 plan and we're going to try and imagine what the city of the future looks like, and please let us know what you think selfdriving cars are going to do to cities. We have the Americal cities who come to us and say: Hey, we hear you're looking for a pilot city. Why don't you come and try out your cars? We know they are not finished yet but we are willing to take that risk. So you have the long-term planning and you have the very short-terminism, if you can combine the two, it would be incredibly powerful. So it's only when we got both these cultures more merged and have influences from both come together that I think we can solve the really big problems in the world.

Top on my list would be climate change because that's one of those problems that is going to effect everything and everybody. And it's a very diffuse problem that can only be solved if multiple parties work together. So that's governments at all levels, it's companies, the technology companies but also others. Other ways that's not going to be solved in our lifetime and it's the kind of problem that if we don't solve it in our lifetime then everything else will be lost and we might as well all go home. Another problem that's somewhat related is actually food, food safety. People always say Malthus was never right, agriculture always became more efficient before we ran out of food. But if we look at population growth projections we may actually run out of food. And there is certainly no way that we can produce meat in the same way that we're doing today, and a billion more people eat meat. It would also be like an eco desaster for the planet. So these are the kinds of problems that I think we will have to solve in order to really make a big difference in the world. The other thing I think about a lot is that what are the by-products of the technology? So if you introduce technologies like automation into society it produces fears. People are afraid of losing their job, so how can we think about what the effects of that technology are not just 5 or 10 years down the line but maybe 30, 40, 50 years down the line. So maybe the whole way we think about work will be very different that we don't think I go to school and then I go to university and then I’m going to have a job for life. I mean no millenials will think about a job for life in any case. But maybe we don't even think about picking up a particular profession and sticking with that. I may be personally biassed because I have a philosophy degree and I now run engineering teams, so that's a very long way from what I originally studied in university. But I think that people if they have the right agency can re-invent themselves multiple times during their lifetime. So as a society we have to think about that. What are the skills that we have to give them or even just the attitudes, right? It's not about teaching someone coding, it's actually about giving somebody the ability to create their own destiny and think about what they are going to do next with their life.