IDEA Collider | Joseph Owens, Google X

IDEA Pharma: IDEA Collider Mike Rea talks with Joseph Owens from Google X

https://vimeo.com/269261531

Joseph Owens: Good to see you. 

Mike Rea: Thanks for coming over. So, this is just for the benefit of everyone who has seen the previous live streams with me talking [in the camera]. This is hopefully the start of a new series of live streams and recordings where we're going to interview people that I find most interesting among the folks that I come across.

Joseph Owens: Well I hope you find some interesting folks.

Mike Rea: Well you're about as interesting as it gets. So, for those of you who don't know Joseph - Do you want to do a quick introduction?

Joseph Owens: Yes. I'm Joseph Owens. I am a Neuroscientist at Google X, which is now actually just called X. It's the R&D factory for Alphabet, which is the parent company for Google. I am a Neuroscientist and a Management Consultant by training, by way of McKinsey and Northwestern. And right now, I'm on a team we call The Early Pipeline and we're looking for big ideas that would eventually be companies that would rival Google, basically. So, for Alphabet, we are de risking to bet on Google by creating other bets. Before that I was a Consultant at Google actually, in the Ad side. And then way back when, I mentioned my PhD was in neuroscience of sleep. So back when I was a Consultant, I was one of the experts on why the job was not very good for you.

Mike Rea: And just for everyone who knows my background, Joe and I got over the McKinsey thing quite quickly. We've settled that conversation. So, one of the things that was most interesting actually in the conversation was really -- one of the things that pharmaceuticals struggle with is scaling innovation. And I know that you've had thoughts on that before you joined Google, and clearly since you've joined Google. It'll be interesting to hear whether you think pharma's going the wrong way, in terms of its approach, or do you think that there's a different approach possible?

Joseph Owens: Well, I don't know if I can speak so well towards what pharma is doing specifically, but I can speak towards some of the things that happened in Google that are good and some of the things that I think we're improving. One of the things that Google was blessed with was, and I think it was really funny because we both knew this analogy, which was, it was a windless tree. And so, it had so much revenue for quite a long time that basically it made sense to plant as many flowers as you could. And so, by spreading bets as widely and sometimes even duplicative, you have the opportunity to let things bloom and let things figure out. As you have businesses that are more related to each other -- a great example is DoubleClick, which is programmatic advertising. The pipes for that are so complex. Having three different versions of that doesn't really work. And in some cases, we've made smart acquisitions -- DoubleClick was actually an acquisition -- and in others, we've built our own from the ground up. I think for innovation to be learned from Google, I would say it's knowing when to pull -- it's giving the engineering directors -- so Google is an engineering-led company and so the equivalent in pharma would be like the scientists or the people closest to it -- some leeway to make a call on whether they're going to let their [directs] just sort of experiment. From what I do know from you and others from pharma, that experimentation is probably not -- the degree of experimentation is probably radically different. And it is software, so you have to remember that some of that experimentation is a little bit cheaper from an opportunity [inaudible 03:51] point of view. But Google engineers are pretty well paid.

Mike Rea: That metaphor of the windless tree, I think I wrote something about that like two or three years ago. It was based on the observation and the biased biome or biosphere or whatever the name is -- the trees grow to a certain height without wind but they fall over quickly because they need distress of the wind to grow. And I think that was an appropriate metaphor for companies and pharma’s that are doing very well despite much pressure from anywhere else. They haven't really needed to think about that innovation thing. But I wonder whether in pharma we spend too much time -- make it a quick call, "Right, well we've done the science already, now let's go to market with this thing." We stop experimenting at that point. So, I wonder whether that's a lesson to be learned. 

Joseph Owens: So, Google has made a lot of changes around how it proceeds to launch, and specifically, how it measures that. Because of its size, it's pretty hard, just statistically, to figure out whether something's successful because it's got the Google brand with it. So, it's like, what is the adjustment factor for Google [to] launch this. And I think that's been something teams have been figuring out -- how to actually [re-weight] the metrics to see whether this would have been a success on its own. And there's some interesting programs in Google right now. There's a program where they're actually encouraging entrepreneurship within Google. So, people have great ideas and they might want to leave. They're allowing them to form their own teams and startup and pitch them to internal sort of VC-like group. Not necessarily with upside for the individual, except for just being able to pursue this thing that they see is really important. And so, it's a way to catch some of those folks that might otherwise leave and start other things. Because everybody has that entrepreneurial spirit.

Mike Rea: We spoke with that a little bit [inaudible 05:54]. We covered that. It was one of the things that I thought about it over time. You look at [inaudible 06:00] with a lot of people who've left Genentech because they had to, to go and pursue their other interests. Interesting that you mention that there's no actual incentive for folks internally other than the progression of their careers.

Joseph Owens: Yeah. I think it's interesting. If you look at -- I think it's Maslow's Hierarchy of Needs or whatever -- once you're paid a decent rate -- the monetary incentive, and you're comfortable -- If you're engaged with your work and you feel like it's doing something worthwhile, I think the monetary upside can be a bit overvalued in that scenario for a lot of people I've seen. There's great engineers who have families and are comfortable and would be great on startups, but they're not going to do that. And so, I think it actually taps into, maybe it's a slightly different slice of who would do that, but they're willing to do it because if they can keep their Google job and go for it. There's other people who want to strike it rich -- go [inaudible 07:04].

Mike Rea: But that spirit of innovation is encouraged within people that join Google?

Joseph Owens: Oh yeah, definitely. So, Google still has, and I'm a product of this, the 20 Percent Program. I was a Strategy Consultant and I came into Google. I wanted to learn the main core business, how ads work. It's a lot more complicated than you think is. 

Mike Rea: It's become a big issue recently.

Joseph Owens: Google touches a lot of surfaces that you might not be aware of. Like how ads get populated across all of these different elements of the Internet, how they're sold, how they're traded in real time. All of these things. And like we talked about with the windless tree, they hire a lot of consultants, specifically from McKinsey, to come in and help them make those calls. While I was doing that, using my core McKinsey skillset, I started a 20 Percent Project. And so, for those that want to and have the inclination and [inaudible 08:01] and we formed a team, were running for six months or so, we had engineers, businesspeople, marketers -- our own little thing -- pitched it to all sorts of people. And that's actually what landed me at X. We had that drive because we saw an opportunity that Google should be working on this thing, and we can't talk about the thing. But we said we want to make sure that Google or Alphabet, actually, is working on this thing. And we pushed it until they took recognition of it. 

Mike Rea: So just let me walk through the basics of the culture -- they enabled you to put together a bunch of people to pitch it to someone else. What's that look like?

Joseph Owens: First off, to do a 20 Percent Project you need to be doing well in your role. But then the idea is -- It's based on good psychology, which is, you can't focus on one problem all the time. It's sometimes switching over to a different problem that actually helps you. And you can pull things over from that. Maybe you're going through a lull where you're bored with the implementation of your project and you're waiting for that next interesting part of your project, but you're still the right person to do that thing. Use some of that spare mental energy, connecting energy, whatever it is, on that 20 Percent piece. And instead of it being seen as lost time away from core work, it's more of an acknowledgement that you can only do core work very, very well for four or five hours a day. The idea that we can do more than that -- my background in psychology, you can't. You check your email, check the stocks, you check the news, go have coffee. And then when you look at your time across the day, whatever. But if you have something that was really driving you on the side, and you can keep up that -- The other thing is momentum. You're keeping up the momentum with the one thing that carries over the first. 

Mike Rea: So, the 20 percent time isn't like Friday, it's spread across. 

Joseph Owens: It's spread across, yeah. I don't think it would be effective if it was like everybody takes one day and puts on a different hat. There might be teams that do that. The other 20 percent thing I've done in my time is, I teach a mindfulness course. And that's actually a two-and-a-half-day course called Search Inside Yourself. And there's an org that runs that, a not for profit that now runs that. And we teach it to Googlers. And so, once a quarter I go and do that. So that sort of is a different day. Now that I've turned my 20 percent job into my full-time job, I have a different 20 percent job. 

Mike Rea: So, people internally, they have the permission to spend that 20 percent time. Are they looking for each other?

Joseph Owens: Oh yes. There's so many ways in which people find people. I just happened to be really crazy interested in this one topic. And someone introduced me to somebody else who's interested in that, and then found one more person, and then steamrolled from there. I said, "Google has to be doing this." And we just pushed it. 

Mike Rea: And then you mentioned X as a special place. What's the special sauce about X that's different from Google itself?

Joseph Owens: So, X is meant to build new companies. There is [triad] of criteria. One is that it would affect enough people. So, we think of a billion people, which means it needs to not be just U.S. So, it can't be just a U.S. business.

Mike Rea: Basic rule of thumb.

Joseph Owens: Basic rule of thumb. And it needs to be some sort of radical advancement of technology that has some real breakthrough way of solving a problem. And so, that's the three criteria, and you put X marks the spot in the middle. And it's always in the eye of the beholder, obviously, how breakthrough something is. But it has to be for good and it has to be a self-sustaining business. So, it's not [inaudible 12:07]

Mike Rea: So, the really interesting thing about the business side is that those rules of thumb are not market size. The problem [inaudible 12:17] and the benefit, those are interesting rules of thumb. 

Joseph Owens: Yeah. And all of the things that happen at X will touch regulation because they're [inaudible 12:28] business model plays, they aren't regulatory plays. We have a rule, you can't break the laws of physics but you might, for a little while, break the laws of man maybe, at least as they currently exist. So, for example, we have a project that literally launched drones into the air, called Wing, in New Zealand. And because the laws of man there were a little more friendly towards flying things, that was a good place to literally launch. And so, you have to figure those things out. Same thing with [inaudible 13:05] which is the inertial project at X, which was driverless cars -- figuring out how to get those safely on the road and get enough miles driven to train the AI. Obviously had [inaudible 13:17].

Mike Rea: Someone was telling me, actually yesterday, internal [betting] that happens at Google. Is there a market internally on which projects are going to succeed and which ones aren't?

Joseph Owens: I don't know of one. But I've only been there for three years. I've been at Alphabet for three years, about a year and a little at Google and a year and a little at X.

Mike Rea: Again, someone from pharma had heard this and felt, "This is an interesting way to see which projects are likely to succeed and which ones aren't." Because internally there's a culture of -- you know stuff internally that maybe senior management don't know about. 

Joseph Owens: Yeah. I would be interested in that. There's a company called Steam; they do video for gaming. So, it's a very engineering -ed company somewhat. It's much smaller than Google. I think what they do are sort of a platform for online games, but then they also I think, sort of video. Could get that wrong. But an engineer friend of mine told me that they vote with their feet. So literally, their desks are attached to their chairs and people just move their desks together to work together on whatever the project is. I like that model. You can see what's working and what isn't, based on where people are moving. But the betting on things, we have a different version of that at X which is, before we kick off and get really running at speed on a project or even an idea in our early pipeline, we create kill criteria. And so, these kill criteria are what would be convincing reasons to stop working on this. Because the most valuable thing is our time. And those are easier to make before you spend a lot of time on a project because you're not as invested. You haven't hired as many people and all of these things. And you try to make them as objective as possible. And the way we do it is we just [sense] test with other people. "Is this significantly better than what exists?" or "Will this out compete the current thing on the market?" And that allows you as you get further down, if you realize you're not meeting that -- and you choose when you're going to check in with your ill criteria beforehand. So, it's like good statistics, it's a priority bet. And that allows for a more objective decision later on down the road. So, it's a way to manage your bets. I guess. 

Mike Rea: And then one of the things I was really keen to recover was you mentioned the "Thank god its Thursday" and I described to you this  environment where in pharma that we spent so much time moving towards this six-monthly review with senior management of very polished, carefully curated slides that they're allowed to see. Can you describe a little bit more about this?

Joseph Owens: Yeah. So, speaking completely for myself, and I think it's well known that this exists out in the world, but the company is -- Steven Levitt, the guy from Freakonomics wrote about this in the early days of Google. Larry and Sergei and others of the founding team decided to have a meeting every Thursday with the company. And I think the first meetings were around a ping pong table, which is also like their boardroom. And that tradition has carried on. That was on Fridays. As the company grew to have enough [inaudible 16:47] components they moved "Thank god it’s Friday" to Thursday. And crazily enough, Larry, Sergei, now Sundar, Susan Wojcicki -- all these folks get up there and talk about the state of the company, weekly.

Mike Rea: Every week.

Joseph Owens: Weekly. And it's kind of funny because Sergei also does a lot of things at X, and so he is often out of breath from making the one mile, mile and a half, from X to doing the same thing at X -- going over to main campus. And I think I said this to you. That's really good, but if you're a product manager, your product's coming up this week. You're going up in front of the CEO and chairman of the board and whatever, and they'll tell you what they think. I think that level of transparency is something Google obviously has struggled with this last year, because of the leaks. And I won't talk about that. But maintaining that transparency, it's amazing. I came into the company, my first day and they give you a computer and you're on the Intranet and you're like, "I can see this?" In any other company I wouldn't be allowed to see that. And that trust in a first day Googler -- Well, maybe I'd go look at that, I'm like, "Wow!" In my last business, we were doing that a different way. "Maybe I should let that person know," And I often do. When I get launch notices from people and they haven't -- from PMs and they'd go out to all of Google -- and I see something that I have a point of view on, I'll let them know. I'll just reply to that launch notice. Not to everyone, but to the PM and say, "Hey, I noticed that you guys did something here." And I think there's a lot of people that do that. And it's not liked a trolling sort of way. It's like there's something I really care about, that maybe you should know about it. And they might ignore you. They might not. But sometimes you get really long responses. They're like, "Oh I'm so glad you pointed out that. I was really struggling with how to weight that decision. And I'd love to have coffee." whatever exactly. 

Mike Rea: And that was what struck me about that idea of Larry and Sergei and their [comfort] to do it. I'm the type to do it as well. I think we spoke about the [inaudible 19:13] book about the beginnings of Pixar and pulsing and the way that -- Pulsing sounds nice and gentle but sounds like there it's also not. You do get your animations ripped apart by everyone -- the magazine, then Pixar. That's not a destructive thing but it's a constructive, enabling, empowering way to --

Joseph Owens: As an employee, you can get an answer. If there's an issue that you believe is important enough, you can stand up at the mic and ask the heads of the company, from the beginning. You might face social feedback on that. I've never heard of anything of someone's manager getting mad at them for saying something like that. I think I would have heard that if -- Someone would tell you, "Hey, don't get up to the mic." And then they take internet questions from around company. And then they take my questions and they alternate.

Mike Rea: Okay. I've spoken to a few people in pharma about whether they could imagine a pharma CEO standing there every day, every month, every week. 

Joseph Owens: It goes with overall cultural transparency though. So, if they get at --

Mike Rea: Is it just transparency or is it something about the connection to the product or the ideas or the --

Joseph Owens: Yeah. I think you've got to be willing to go both ways. You have to defend your project, the people getting up there and talking about whatever they're launching or whatever or the bad news cycle on their project, whatever it is. That's one side of it. But then them asking like, "Hey, we did this launch and --" The thing that was in the way might have been you. Can you tell the audience why you made that decision? I don't know if I would go up and there do that, but people in the audience will. They'll say like, "Why did you make that decision?" 

Mike Rea: Right. Okay. It's interesting because part of that same conversation that we had around whether they could imagine pharma CEOs doing that, people tend to go with the ones that they've worked for, that they could imagine being that. And actually, at the same time, those people also seemed to be the most empowering and best leaders -- the people talk I'm talking about in pharma -- people like Bob Levinson at Genentech have a [proof ability] but also deep -- you'd follow them anywhere with the science. So, I wondered whether that was a --

Joseph Owens: Well you have to remember that Larry and Sergei were grad students at Stanford, in information sciences. So, the transparency piece is there. The depth of engagement is there. These are future academics made into CEOs. And I think Larry's written about this bunch, about what that transition was like for him. What they're gifted with is all these great people who can teach them these things. And so, as they were going through -- I think I've seen this written in a number of books about Google -- One of the things I did before I applied to Google was, I read all the books about Google, at least the ones that are available. When I was at McKinsey what I did was a lot of reorgs. And so, I worked on helping organizations be more effective, because I liked the novelty of that problem every time. And reading about their early history and seeing the problems they faced in changing their worldview -- I was a PhD student. That is a very different [person] to being an executive. And so, the attitude there is you have journal club. And I got to say, TGIF is not that dissimilar from journal club. Journal club, you get up, you talk about some data, you beat it up. The goal is everybody gives their opinion. And if someone is silent then you're losing out on something useful. Because all the researchers in the room are going to have different takes on that data, or maybe they have statistics or genetics or whatever it is. It's not that dissimilar. It's bringing a little bit of that academic culture into corporate; I think.

Mike Rea: There is something about pharma which I think we [could] change. We've got [inaudible 23:30] with people with project teams to say, "Well, what are all the things that could go wrong here?" Remarkably, it's the first time we've ever been asked, typically. And then they have this long list of things that could go wrong. They're not just about the product succeeding or failing on its basic parameters, but everything else that needs to be thought about it to get it there. If they're not being asked, those things will still happen and we're just going to ignore it until they do. Is there something that's enabled -- Let me describe it perfectly just from the beginning -- it always was that way.

Joseph Owens: Yeah. If you're not working [in] your culture at the beginning then you're going to have whatever culture you get. The changing it though, is that what you're asking?

Mike Rea: Well, I was wondering because one of the approaches that you have, clearly, is that you stop other cultures that are separate -- that you've created companies within Google that are different.

Joseph Owens: Yeah. That was one of the things that kept me awake the most when I was working on the 20 Percent Project. I said, "Okay, we've got five people on this. Whatever we do right now that's the beginning of the whole proto-companies culture. And those are big weighty problems to think about. So how are you making decisions as a group? How are you choosing the direction? Are you going to be monolithic based on that one engineer or are you going to be consensus driven? Those decisions are made on those teams as they form, and a lot of big projects in Google started out that way. I'd say there's an example that teams can learn from, which is what's happened at Google and maybe what's happening on their own teams. And then when they make these new teams -- like the 20 percent ones for example, or the new bets at X, or the acquisitions -- there's a lot of freedom given to them to make those calls. I think it's an experiment that keeps happening over and over.

Mike Rea: And we also discussed the accidental versus on purpose nature of the organization within Google. Which you can say about the way that it's organized and your observations on how controlled that is versus uncontrolled. 

Joseph Owens: I think Google, last couple of years, they made the switch to be a holding company, I think quite wisely, while I was there. And the reformation that happened because of that has objectively been good for at least the short-term stock price. And starting to compare some of these projects against each other, and to make some of these calls. I think those things happen in cycles. And so, they're on that cycle of it. I think the culture probably still has this exploratory way. And so, if you go through one cycle of comparing things and choosing which ones of the best ones, you'll go through a growth phase. I think the inertia is clearly there for it to be a [inaudible 26:54] thing, not like a, "It was doing this and now it's doing that."

Mike Rea: And there was an observation that you mentioned along the way about how much people want to work for Google, as opposed to somebody else.

Joseph Owens: Yeah. So, Google maintains a pretty amazing reputation, at least as a place to work, in the world. I always saw it on lists with McKinsey and other consulting companies. And I feel like those are pretty different jobs, which is interesting.

Mike Rea: Those rankings are usually done by [inaudible 27:24]

Joseph Owens: Yeah. [inaudible 27:26] does rankings too. If you want to be a world class software engineer and you want to have some of the best tools at your disposal, and obviously the [inaudible 27:43] places to work, and smart people -- I think I've got a little bit off the question -- but the attractiveness to do that, I think it's quite high. What was the question again?

Mike Rea: Well,  it's linked to that. Because we had the conversation around the pharmaceutical innovation index, on whether that leads to retention of people over time or the ability to recruit. 

Joseph Owens: Yeah, I think there's everything at Google. So, there's enterprise businesses at Google, there's consumer businesses at Google. With the cloud bet, that's a very different business than the hardware bet. And one of the things Google has is a lot of ability to move around. I think that's what I was mentioning. And so, you might work two or three years in one role and then you might change ladders as I did. I went from a strategy consulting ladder -- I'm actually on the engineering ladder now. I don't know that that happens that frequently, but I definitely see people who might go from, say, a sales ladder to PM ladder or a program management to product management, or one type of engineering to another type of engineering, as they change their skill set. One of the things I do as a 20 Percent Project is, I work on what's called G to G which is Googler to Googler training. And we have loads of that. We have an engineering school. If you want to get ML training, there's weeks of training you can go take to start teaching yourself to be an ML engineer. There's Python 101. There's everything you can imagine if you want to spend that effort to train yourself. Now there are tools that are available for online training and any person training, you can literally change your career while you're at Google. And I've seen a lot of people do that.

Mike Rea: And you can start your 20 Percent Project from any one of those ladders? You don't have to be on --

Joseph Owens: Yeah. You can be a salesperson and be the PM on your 20 Percent Project. Or like me, you can be a strategist on your normal ladder and you can be a scientist on your 20 Percent Project. 

Mike Rea: Because one of the things that we haven't spoken to anyone yet about is about the rankings that lead to companies being perceived as more innovative, and whether that leads to the ability to attract and retain all the time.

Joseph Owens: Google made a big bet on hiring ML engineers and that looks like it's paying off.

Mike Rea: That's machine learning?

Joseph Owens: Machine learning, yes. Sorry. Everything where you teach a computer to label things. That's all ML is. So, it's saying, "I give you a lot of data --" and then computers are very good at saying, "That is A and that is B." Assuming that you have good enough examples of A and B. That is all machine learning is. And Google made a big bet on that because they get a lot of -- it's an information technology. We're categorizing and making available the internet. And so, all that tagging, that's kind of the grass of machine learning. You have videos on YouTube that are labeled, and voice recognition and all these things. These were the data we were taking in. And so, not being [inaudible 30:39] ML was pretty obvious, you're not going to work. And then we happen to have servers. So, the other thing that's happened to make machine learning capable these days is something called deep learning. And that's only possible with the amount of server space, basically. The amount of little, literally, processors to throw at the problem to run these iterative models. And without that you can't do the kind of machine learning that we do. And so, we had both of those things and then we are where we are.

Mike Rea: Which is interesting, the ability to understand and deconstruct at the same time, is important. And then clearly within the health space, I know we spoke a lot about the essential problem of hundred-year-old disease definition still being part of the fabric against which we're developing new drugs and new ideas. 

Joseph Owens: I just read the outgoing NCI directors book on cancer, which was I think, Curing Cancer. And it's a labeling problem. Initially, when you go into labels, if the label's too general, well, the machine can't learn to label below that. At least, it can't learn on its own. There are machine learning techniques called clustering and unsupervised learning, and those can begin to do some of that. And we're not -- Google is not the only person doing this. Unsupervised learning without the gold standard labels with it, and clustering these things out and then saying, "Hey, this is a cluster. Let's go study that." And yeah, these were all what we were calling cancer. But [now in terms of] mass childhood lymphoma -- and this is sarcoidosis or something, whatever it is -- 

Mike Rea: But we're getting there, or we're starting to get towards that in cancer. I think probably because it's had a molecular target for such a long time and people have explored the genetic mutation mode and so forth within the tumors. My concern is that you get into areas like mental health, that we're still using broad categories like schizophrenia or major depression --

Joseph Owens: Now you're getting into my wheelhouse. I'm not going to begrudge the people who hammer these things out in committee to make the DSM. That is exceptionally hard, based on what we have. Because we don't have data. We have an empirical wisdom and we have research going in lots of different directions. Because we just don't know very much. We don't know very much about the brain. We have to admit it. We don't know very much. And I won't compare neuroscience to cancer or anything like that but taking one of these labels and deconstructing it. And then, we have loads of studies. I was doing genetics too, where we say, "Wait, why does one disease and another disease and another disease, all radically different labels, run of the same family? Are these normal curves and we're just picking out the ends of the curves? Are these bimodal curves under certain environments?" Picking that stuff out, I think computers will be very good. But we need more labeling data. So, the move right now -- and there's a lot of folks doing this -- is to get passive monitoring. One interview in a doctor's office is not enough. And if you can move towards passive monitoring and long-range continuous datasets -- And then folks are very wary of doing that.

Mike Rea: It starts to feel healthier as a way of -- if you take something like schizophrenia, we know there's genetic components, we know that there's typically socioeconomic components as well, and then the family environment components. But then also, the interventions that we've had are pretty broad brush and pretty crude measures, in terms of their effect. And if you look at the construct that you're describing of an appetite, to want to break it down into micro subsets --

Joseph Owens: I mean, it's been variously called personalized medicine, lots of different titles for it. But subcategorizing disease for neurological -- I mean, all of this -- is the next wave, I think. And hopefully we'll destigmatize it. 

Mike Rea: And then you put together, what I see from the outside, as a kind of long bet that someone like Google is prepared to take on. If you look at mapping the roads and self-driving cars, there was no business in that for a long time. There's a long-time bet. Parmer is in that same sort of 20 or 40-year cycle of discovery to development to revenue. Do you see any parallels or any differences between them?

Joseph Owens: Oh yes. I think pharma is interesting because it starts, and at the early stage if you can kill something and save you a lot of money down the road -- because the last trials were the most expensive, theoretically. We're at a point which is very different, where we say, "Let's take in as much data because we don't know what it's going to lead to." And that's a very different decision to make at the beginning. So, to take the mapping example, "Let's go out and put cameras on backpacks and on cars and take in this data." I don't think they knew exactly what product that would turn into. But when I did my interview at Google they said, "What product at Google do you admire?" And I said, "Apps. It changes my life every day." Every day I set out with confidence. I can get to where I want to go. Every day I can take a request from somebody to go meet somewhere I've never been. And I [inaudible 36:16] take that request. "Hey, come meet Mike in this building you've never been to." Didn't bat an eyelash. Before maps -- get on the internet, look up where I can find it, find a map, whatever it is. That's a radically different decision multiple times a day. I think when they first sent the cars out -- there's no way they're foreseeing that everyone would be making different decisions. At least that has, the luxury of having Google Maps. 

Mike Rea: Yeah. And that's one of the things that we talk a lot about. That idea is that exploration and value early. Because no one knew where the iPod would lead, in terms iPhones and apps and a bunch of other things. And certainly, if you try to do what pharma tends to do, which is to try to put like five decimal place forecasts around a Phase 1 asset, you're already limiting --

Joseph Owens:  False precision is --

Mike Rea: Yeah. And how does that get approached at Google? Do use those rules of thumb all the way through or is it something that someone else [inaudible 37:14]

Joseph Owens: I don't think I've been there long enough to see things from genesis to multiwave implementation at scale. So, mine would be snapshots across different projects. The decisioning that happens to kick something off at Google is, I think, laxer. So, it's experimentation. If you're really excited about, "Well, I believe in you because I hired you." or "we hired you" and you're coming to me and saying, "I really [inaudible 37:50] this." Your energy is the voting factor for a little while. At the point that you start needing additional resources, then you start to make decisions. So, then you're making prioritization. Some of the similar, "Let's make a business case for these things." pops up and you say, "Here's a design brief, here's a PRD -- product requirements document -- and here's the case." At the beginning of the product requirement document it would be, "Here's why we need the thing. And here's what the thing has to look like." From the ones I've seen, they're not trying to get to decimal places of precision. And that's probably a little bit because of the luxury of resources. I think things are allowed to flourish a little bit. 

Mike Rea:  That's an interesting word -- flourish and thrive. Because one of the things that pharma tries to do is to project ten years into the future and then bring it back to today with a huge degree of accuracy, despite us all knowing whether it's wrong every time we do it. And then the project's not allowed to flourish, despite all the evidence that most of the great drugs have got where they are through serendipity. They've pivoted at some point in their life cycle. 

Joseph Owens: My example for that is -- basic research is -- Carrie Mullins goes and works on a project which, its title, would every time get defunded. He's going to go measure proteins and enzymes in hot geysers. No one cares. No one cares about those organisms. But then you get PCR. His project, if it came up for a vote, everyone would, "De-fund." And then he's driving down the highway and he thinks -- So we have some core principles; more data, better; diverse data. So, try not to just collect data in Silicon Valley, these sorts of things. Build for scale down the road. Because everything we want to do is going to serve Google's customers. And so, things probably move more slowly than they would at a startup because we're building for scale early. That could be a headwind of saying, "We'll go get a bigger dataset than maybe a startup would want to launch their thing." Or,  "Build your pipes a little bit stronger than a startup might."

Mike Rea: So, there's some value to their being Google?

Joseph Owens: Right. But then you're slower. I mean, those are tradeoffs. 

Mike Rea: And probably the last question, because we could carry on for another few hours, would be just really around how you personally see health -- the intersections -- health and Googling -- the kind of technology that sits behind those.

Joseph Owens: So, health for me,  for Joe Owens, II vacillate between extreme jaded [inaudible 40:54]

-- like I said, we don't know anything about the brain. And we're at that moment of Newton where we can't see Einstein. When you're Newton, you can't see Einstein. [inaudible 41:05] physics, you can't move into relativity. And I feel like we're -- that moment on brain -- But in the same sentence I have to say, there's so much science sitting on the table that hasn't been brought to people's lives. We have doctors that have no time to do the thousand things that have been recommended for them to do. Well that just sounds like a platform and it's fixing the issues that happened. If that's an operations problem, that's a more McKinsey [hat] problem. So, if we can take all of these recommendations that we have for our health, and figure out a way to [massage out] the way we live to meet them, well then maybe we don't need -- we don't actually have to know all those things about the brain to actually operationalize some of that. So, I think science [hat] kind of terrified, consulting [hat], I feel like we just need to do some stuff. 

Mike Rea: So, some systems thinking?

Joseph Owens: Systems thinking, yeah. And design thinking. Changing the way, we live to be healthier. We know what to do, we just haven't done it. 

Mike Rea: Yeah. Some of those things are [inaudible 42:09]. If you've got socio economic problems --

Joseph Owens: Google, it changes every day of my life with maps. It could change every day of my life with my health. And I think there's people at Google who are seeing that. They have been thinking about it. 

Mike Rea: And actually, [inaudible 42:29] is really around the ethics with that as well. I'm aware that there's a kind of internal ethics, people looking at whether you can do harm as well as good.

Joseph Owens: Oh, yeah. First do no harm, is the first rule of medicine. So, if an IT company wants to get into medicine, they've got to follow that. One of the things we have going right now is a lot of people thinking about machine learning fairness. Do you collect first data sets? Things that work on one population won't work on another, unless you figure out the little bits. And so again, that'll be a tax for speed but it'll be in the effort of fairness. And ultimately, scale. So, the ethics of that -- I would say you probably don't see people out in the world talking about this, but teams talk about fairness a whole lot. [inaudible 43:23] would be a great example. And the computers are sometimes really good at this. They have an example where there's a woman out in the street shooting a duck with a broom. There's no way you're going to train your data [to solve] that because you could even conceive that the car would ever see that? And so, figuring out ways to end all of the niche cases via getting really diverse data sets and really good transfer learning, that's -- Computers actually may have a better shot at scale -- oh, sorry, in fairness. 

Mike Rea: Excellent. So, I think I've promised everyone that this would be phenomenal, and it has been, Joe. It's probably obvious, we could carry on for another couple of hours and debate this. And I hope you get a chance to [inaudible 44:16]. 

Joseph Owens: Yeah. That would be great.

Mike Rea: Thank you.

Joseph Owens: Of course. Have a good day.

Mike Rea: Thank you very much.