a16z Podcast - a16z Podcast: Innovation vs. Invention at Google I/O

Episode Date: May 20, 2016

Innovation or invention? Platform or app? Vertical or horizontal? Strategy or tactic? Does the smartphone eat VR? And (not to get all existential about it or anything but), what is an app, really? a16z partners Benedict Evans, Connie Chan, Kyle Russell, and board partner Steven Sinofsky explore these tensions in this episode of the podcast as they share some quick reactions to Google I/O, Google's annual developer conference, where the company announced a number of new platform products -- for VR to messaging to the smart home. Maybe most new things are really old things, but maybe those distinctions don't matter as artificial intelligence leaps into how we live our lives, automating (and anticipating) things in a new way...

Hi, everyone, welcome to the A6NZ podcast. Today, we're sharing some quick reactions to Google IO, Google's annual developer conference where the company just announced a number of new platform products for VR to messaging to the smart home. A6NZ deal investing partner, Kyle Russell, discusses the news with A16Z's Benedict Evans, Connie Chan, and board partner Steven Sinovsky. Over to Kyle. So, Stephen, yesterday, Google announced quite a few different products and services. Several cynical commenters on Twitter came. out with the idea that, you know, hey, we've seen a lot of these things before. Oh, a computer you control with your voice in the home, a bot that you interact with to, you know, interact with different services. We've seen this before. Why is Google showing all of this off now? None of this feels new. But you push back against that. Why do you feel so strongly that, you know, this was actually a really exciting event for Google? Well, because it was a really exciting event for Google. There's no doubt about that. I mean, they showed an unbelievable amount of stuff,

and it was coherent and clear and innovative all at once. And I think what people get confused by is that, like, how does innovation really happen? And there's this view that, like, innovation is the same as invention. And that, like, you one day you don't have something and the next day you do, and that was innovation. But that's also an invention. Like, there's a day without Velcro, and then there's Velcro. But not everything is Velcro. And, in fact, most things are very incremental steps.

And in fact, most new things are like 90% old stuff, like with a slightly different perspective that comes from how you think about the collection of things, your own expertise, your own experiences. And so when you look at something like the work that they showed in messaging, yeah, like it's a messaging app. And there's a lot of messaging apps. Actually, Google already has several of their own messaging apps. But because of the way they think of innovation is tied to research, they see building on each idea and almost starting over each. time. And that doesn't sit well often with people who are like, I just want the new thing and the big thing, or I want it to be completely different. And so anything you look at, word processors or photos or social networks or browsers or smartphones have all been like, wow, the new one looks a lot

like the old one. And even when the iPhone came out, like, well, there were phones that browsed the web and there were phones that had touch and there were phones that did all the things that the iPhone did, including making calls and sending text messages, but it didn't, people didn't, for some reason, that counted as a whole new thing. And everybody wants everything to appear like it's this iPhone thing in retrospect. And so I looked at it and I just saw, I just saw like innovation happening broadly. And what was really going on for me was just this underlying theme that they, it's not, it wasn't hidden or subtle. It was they've decided that the next generation of platform is artificial intelligence. It's a huge differentiator. It's a huge

thing where they've taken a different approach, and it's a huge thing. Now, they may or may not be right, but it's super clear. Right. And they're also pulling on their main strengths of having access to a ton of data, having the best computer vision technology out there. So it totally utilizes what Google's good at. Yeah, I mean, I think the, you know, when we kind of sat and looked through all the stuff that they've built, there's this huge story around what was Google now and is now Google Assistant, and that's around everything that Google might know and might be able to suggest. I mean, I was thinking about this sort of historically that there was web search before Google,

which was just text indexing, and it didn't work very well. And then there's page rank, and suddenly web search works. And then they start thinking about kind of knowledge graph, and you search for the Taj Mahal, and it knows that there's a more slim in India, and it knows that there's a curry restaurant around the corner. And then it gets to the point, like, okay, you're doing this search at 8 o'clock at night in East London. Now, why have you searched for Taj Mahal? You're probably looking for the curry restaurant. And then now what we're starting to see, like in the thing, in the demo of the

messaging app, and in all the different aspects of this, it's like, now we're getting to the point that, like, we know you're going out to dinner with your friends tomorrow night. What shall we suggest? And so they're having gone, kind of progressed and kind of iterated on what the search box was and might appear from that. Now they're kind of going before the search box and thinking, okay, what are you actually going to need? What are you going to ask for? What questions might you have? And so that's really the sort of the kind of expression of AI.

It's getting to the point that the computer could kind of know. Like every, and one way, I think one way to think about AI is like the computer should know this. What should the computer know? And this is now going to the point, well, the computer should know I'm going out to dinner tomorrow night. Yeah, I think the messaging app just embodied the concept of eliminating friction. And I really loved how they really leaned into that concept.

And if you think about all the other Google services like Google Calendar or Gmail, you can just start getting super creative and brainstorming what this can end up looking like in a year. And I think, too, for me, like one of the things that's super interesting is just how, how you know, if you put your cynical hat on, you come up with some master plan where they're trying to absorb all the world's knowledge and get in your way and invade your privacy and do all these things. But if you take that hat off, you just go, wow, there's a bunch of problems that they're trying to solve. and they're trying to make substantial leaps in the way that, you know, life is automated. And that's really, really interesting because we spent the 20th century automating manufacturing of things and making of things. But we didn't do much to automate living. Like, you know, the GE Carousel of Progress talked about like washing your dishes.

Okay, so that was a machine that saved labor. But all the things that we do all day now, there's so much of it that should be a lot easier and it should be more anticipatory. And it turns out, you know, we do a lot of patterns in life. And the self-driving car is a recognition of like most driving is very pattern-based. So why should it also be pattern-based yet high risk or pattern-based and slow or pattern-based and you just have to think about it constantly? I think that when it does launch, there will be use cases where it goes wrong. Like you can just imagine the photos that they interpret incorrectly occasionally or the suggested phrase that just you can just imagine a whole tumbler being created around this.

Well, they will. it's just they're all going to be stupid computer tricks like and people like I read a whole article about don't forget that there are all these problems with AI and like it was like it all boiled down to a phrase that AI can be wrong and I'm like thinking myself right and everything that came

of a Gutenberg press wasn't accurate the fact that there's a tool doesn't mean that all problems go away I mean I remember using when Google Maps launched and there was a sort of a period of a couple of years where Google Maps didn't work at all in the UK unless you put the postcode in it was just totally unable to work out with what the address was

And there was like a kind of, there was probably maybe a Tumblr or something else called WTF Google Maps of like crazy results. How could it possibly have thought what the answer was? And you know what? Here we are 10 years later. It works. Well, and like I, you know, I certainly live through like all the autocorrect and spelling dictionary and grammar checking kind of things that were stupid computer tricks. And that was our phrase. That was like literally we had a kind of bug that you could open up with the descriptor's stupid computer trick, which was like the algorithm went nuts.

For those in our audience who didn't watch the live stream of I.O. Yesterday or follow along on Twitter, I want to quickly lay out the different products and services that were announced. So there was Allo, which was a brand new, separate from Hangouts messaging application, specifically for mobile, iOS and Android. There was on top of that Google Assistant, basically the modern revisioning of Google Now, their service that's been built into Android, built into the Google app on iOS. There was also things like instant application. where in the same way that you load a web app today by just going to a URL and it shows up, and it's just a website that you go to, you can do the same thing with native applications, loading them in chunks. So looking across that, you know, Benedict, something that you talk about a lot is, you know, tech Tourette's, these buzzwords that every couple of years come up and they're the focus of what all the big incumbents are talking about. You know, today it's AI, bots.

No, bots is last month, dude. Oh, no. But there was some, you know, there were aspects of those buzzwords in what Google talked about yesterday. Yeah, I mean, I think, I mean, to kind of Stephen's point, I think one, one should think, look at any of these big tech companies in terms of what is a tactic and what is a strategy. And so you look at the Google messaging app, Allo, it's like, okay, Hangouts was a bit of a mess, and it was a bit fragmented and it didn't do anything that kind of modern messaging apps do. Okay, so let's just kind of reset that and we'll make a new app and we'll call

Allo, fine, done. That's like having a better notifications panel. It's just housekeeping inside the operating system, inside the product group. But the kind of the fundamental strategy for Google is what they call Google Assistant, which is basically an expression of AI, an expression of Google turning the search model upside up on its head or looking at it from another angle and thinking how do we get to the point that we can actually answer questions as opposed to giving your list of 10 blue links? How do we get to the point that we can even predict those questions as opposed to giving you 10 blue links? And they give you a demo inside Allo that says you can do that inside Allo so you can message somebody and say, you know, what restaurant should we go to?

And then you can ask Google Assist, well, what are the restaurants near there? But that will also work in Messenger and that will work on the web and it will work in anything that you're doing on your Android device and anything that you do on the web and to some extent that will probably be part of the Google keyboard on iOS as well. But that's kind of the point is what Google does is understand knowledge and understand data and find ways of manipulating that. And that's cars and that's computer vision and it's probably drones and it's Google Assist and it's search and it's everything that flows out of that. And I feel like Google Assist is almost like, just like page rank doesn't really exist anymore. PageRank is 200,000 that different algorithms.

Like Google Assist is like the umbrella for everything new that Google is thinking about how they understand what you're doing. And that, to me, is like the fundamental thing. And then there were like two other interesting things. One is the VR project. And then the other is the Google Instant Apps. Let me build on just your thesis there just for a second because I could put it in terms that people make immediate splits over it, which is it's a platform in an app. And so strategy or tactic. And the thing where people get confused is like,

What we saw yesterday was the Google platform, like I.O. is for developers. And so they're going to show some apps along the way. And it's interesting. Yeah, there's a new messaging app, hello, and a new video app. And even within Google, those are just the apps. And it's the platform. They're a platform company intrinsically. And that's also worth contrasting it with, you know, a device-centric view or contrasting it with an enterprise-centric view or a social graph-centric view. But the core is it's just a platform and platforms have lots of apps and to view the whole importance of the day through just what the app does can miss the whole you know the iceberg of like that's a platform underneath it yeah exactly everything google does plugs onto the kind of core competency you know whether it's gmail or advertising or web search or self-driving cars they all plug onto that core platform of understanding data right so when it comes to the app messaging application alo one do you think it's going to be you know successful given their kind of I almost don't care.

Well, okay, so that was my follow-up is it doesn't matter if really what they're... Well, this is kind of the point. I think it has a shot of doing that. Well, it does, but this is kind of a point I was making earlier, which is like, if you were on the Hangouts team and you look to Hangouts, it's like, okay, there's kind of a mess. We've got this different stuff and it's not great. Okay, well, we'll reset our messaging strategy and this is the thing we'll do.

But that's kind of like, you know, Stephen 15 years ago saying, you know, the charts module in Excel, we kind of we need to rebuild that. It's like, well, that was on the roadmap. Maybe it's this year, maybe it's next year. But that's not the strategy. The strategy is office, an office and windows. And the same for Google. It's like, okay, the messaging thing was kind of a mess.

We sort of the messaging thing out. But the point is the data platform. I also think in the Western world, messaging is still fragmented, even though there are leaders that have hundreds of millions of users. It's not like Asia where there's really two players that dominate, right? In the States, there's still a chance for a new messaging app to get significant uptake. Not only a chance. I would just go out on a limb and say it's absolutely certain that there's going to be

defragmentation of, like, just the app called messaging. And that's, again, back to the whole innovation thing. I mean, we're in this very highly fragmented state, which in the tech world doesn't last very long. Now, maybe unlike previous generations, it's not going to get to one because the mobile world is just too ginormous that it won't have one. Even if it's global, there'll be one in China and one in the U.S. and one in Europe or something.

This is basically we're talking about the laws of thermodynamics here. Yeah, yeah. But, like, when people, like, that was. That was sort of one of the little Twitter battles yesterday for me was like this notion that we're done with messaging. And I kind of wanted if there were more characters, I would have said, like, so like I message is all that humankind can accomplish in the world of communicating with other humans and were completely done, which like was literally by my pitch for why we needed a Windows word processor in the late 80s was because people thought DOS word processors were done, which was weird because they were six years old. And six years ago, we were using like special weird word processing hardware. And so this is super, super early in this messaging. And if you look at the ones we all use every day, like they're all wildly deficient relative to each other.

Yeah. And that's like, so that's a state that's completely unstable. And you could view that as, wow, I message isn't on Android. Or you could view it as Android really doesn't do emoji well enough. Or you could view it as none of the ones in the U.S. do all of the features that they have in Asia. And if you think about adding in orange, incorporating Google Drive, maps, email, calendar. Buying stuff?

Yeah. Looking across these products that we've talked about so far, you know, Google Home, where it's going to be this Amazon Echo competitor, the messaging application with Google kind of discussions baked in, where you can talk back and forth with this Google bot. The connection to the underlying data platform is pretty clear. But there was another set of announcements that felt a little bit removed from that core competency, that core focus, the core strategy.

As I mentioned before, the Instant Apps, but also the big VR news, that they're making this reference design for a headset, that all of the big Android OEMs will be able to build a top of, make their own headsets essentially competing with the gear VR, this idea that's a plastic. Paul Samson. The instant apps is a really big deal. Yeah, that's kind of a meds a big deal.

Yeah. That's huge. Well, okay. I think it's a really big deal. So, no, so, you know, not to say that that wasn't a big deal. I was just saying that wasn't clearly connected to this underlying search and it knows more about you. It's a different part of the story. Yeah, both VR and instant apps are not part of that, their other thing.

Agreed. Right. So instant apps, it's like, I mean, you know, this is, you know, you can make all sorts of jokes about this. Like, basically, this is the reincarnation of Java. It's like, I go to a web page and then native code downloads and starts running. And it is actually, as Oracle is currently suing Google over exactly this question of whether Android is basically Google, it's basically Java.

And so, yeah, I mean, just to explain what happens, so the last two or years, what we've had is that if you tap on a link on iOS or Android, and that's a link to a service that also has an app, then if you don't have the app installed, then it will go to the app. the web page. If you do have the app installed, it will go to exactly the same piece of content inside the app. And that's just sort of been baked in over the last two or yours on our iOS and Android. Now, what Google have now said is, and really interestingly, it goes all, not just in the next version of Android, but on every version of Android back to, I think, KitKat, which is like the vast majority of Android devices, you tap on this deep link.

If you don't have the app, it will automatically pull down the app from play and start running it kind of chunk by chunk. So in principle, you tap on the link to a story in the New York Times. The New York Times app appears on your screen and is running on your computer even though you haven't gone to the Play Store and downloaded it. And then doing that by kind of breaking the app up into parts and kind of pulling them down one by one in the background while you're using it. And that changes what an app is, pretty obviously, and it changes one of the sort of fundamental problems on the smartphone, which was this sort of binary question of is the app installed or not and how do you get people to install your app? And are they on the

web page or the app, and you send them out, and you have to pay Facebook loads of money to get people to install your app. And so, you know, and it's a huge kind of technical achievement or sort of technical change. And it really challenges the question of this binary split between apps and the web. And if you think that they own the Google Play Store, too, the natural linkage for people to upgrade to the full app, it makes a lot of sense for Google to be able to pull that app. It's not even that you're upgrading it. It is a full app. And there's a kind of what I'm not quite clear about is kind of what happens when you leave it. Does it still remain on your home screen or not? Do you have to get it? Yeah. Well, there's a lot. There are a lot of interesting details about how does, you know, how do you really have to build your app so that it works well and what are your data charger? There's a lot of stuff you could go all conspiracy theory on that it's not going to work or that it's. But fundamentally, what they're, you know, again, back to their solving problems because they always start with the hard technical problem and work backwards from that. And fundamentally, they're still trying to solve the problem of, you know, the local device has a certain set of runtime capabilities and code has to exist and come from other places. The web and the browser solved a problem in a certain way by putting the browser runtime on every device you could imagine, and then the code that you could download was JavaScript and HTML, but it turns out to not be rich enough for a whole bunch of interaction styles and a whole bunch of things. So this is a, but it was a better trade-off at the time. Now we have more bandwidth. We have incredible compute power on the devices. So like, what's another run up the hill to solve a problem of getting chunks of code on device? And so this seems, you know,

particularly interestingly innovative, the fact that there is app stores, the fact that they own the runtime and the Java environment and the eclipse tools that they said 90% of the people are using for the top 125 apps, like they could actually solve this, which would be a huge change in computing. Yeah, and what excites me about instant apps is it really does eliminate friction for the end consumer. And it also reflects to me that Google understands something that WeChat has known for years and it's critical to the success of official accounts on WeChat,

which is if you can eliminate the need to sign in and eliminate the need to input your payment credentials, that is huge for increasing the velocity and the conversion of transactions. And so Google demonstrated one example where you could pay for parking within just a couple of clicks. With Android pay, without the app, but just staring at the app set.

But just staring at the parking meter. So this is, I mean, there's a fascinating comparison And on the one hand, with Facebook's attempt to do bots, and on the other hand, with the rumor that Apple will do Apple pay on the web, on the mobile web. I don't even believe that that's a rumor anymore. Yeah. So you have this kind of question. So Facebook launches bots.

And part of the argument for bots is, well, it's a pain in that you don't want to have to get people to install your app. So maybe you could just interact with them inside the messaging app. WeChat solves that problem by actually giving you web views, which give you like a nice, rich interaction model. Facebook tries to do it with AI and with a chat model, which may work for some use cases and doesn't work for some other use cases. Google comes along and says, screw that, we'll just give you the damn app. You know, never mind all of these workarounds. I mean, it reminds me, I was on a, I remember, like, in, I think, 2002 or three, talking to the CTO of Motorola, and he was talking

about how hard it was to put a hard disk into a mobile phone to compete with the iPod, because, you know, you drop it and it breaks, and people wouldn't accept that in a mobile phone. And in the same time, Apple, of course, and they're not putting it, they're not thinking about hard disk, so putting Flash to make the iPod Nano. And as I was like, you know, never, Don't solve that problem, you know, jump over it and solve it in a completely different way. And I think that's what Google has done. Yes, they have, you know, in effect, the whole AI platform, which is partly around bots, if you want to call it that.

But actually, they've just said, no, no, no, you know, that's not how you solve the problem. If the problem is, how do you get the app onto the phone? Let's solve that problem. Let's not solve it by making some other thing. Well, and also let's take advantage of a whole bunch of learning and a whole bunch of changed aspects of the technology landscape that make a different approach. You know, like it's not virtualization. You know, it's not like rebuild a new app runtime.

It's not different set of abstractions. They're trying to figure out a new way to solve it. The people were speculating that you'd solve this with like a kind of a Citrix approach where you have the app running remote in a VM and you kind of screencast it. So no, screw that we'll put the whole app on the phone. There are some cases that where a WebView can be beneficial versus parts of an app, like giving developers the ability to do these one-off marketing campaigns or to react much faster and just change something.

thing same day. Yeah, I think that's right. I mean, it comes back to Stephen's point that we had 20 years of the web browser mouse keyboard and that was like the monoliths for how the internet worked and now we've broken that apart and we've got all these different shards all kind of floating around and landing in different places and being juggled and changed and moved around. And Apple and Google every summer kind of throw all the pieces up in the air again and say, okay, now it's going to look different again and again and again. And we haven't got any kind of sense of resolution or stability here because we keep having this kind of iterative innovation. To round out this discussion, I want to jump to my favorite topic,

virtual reality. Google, while it didn't come to market with a high-end headset that you connect to a computer, it was the first of the major tech incumbents to release something that's actually used by millions of people, cardboard, actual headsets made out of cardboard that you slot your phone into for a very basic virtual reality experience. While it was amazing that they got in the hands of so many people, it wasn't exactly an extremely compelling experience, especially if you have tried some of these higher-end headsets. But now they're kind of pushing in that direction. You know, with a daydream reference design

that they're going to give access to all the different Android OEMs, they can make their own headsets, basically at this gear VR tier, as well as optimizing the Android operating system for virtual reality use, it seems like they're making a much more serious effort than cardboard seem to be. Well, I think the basic story here is smartphones eat VR.

And like right now, VR is kind of a branch of the PC. in much the same way that you could argue the games console was sort of a branch of the PC. Certainly the Xbox was, but it was kind of a branch off to one side of the PC. And now, you know, if you need $3,000 worth of kit, it's kind of a branch off the PC. But clearly the mass market use case is it's not a PC with an umbilical cord and $3,000. It's, you know, it's a couple of $100, and it's probably, that probably means your smartphone, and it probably doesn't mean a cable. And so it's pretty clear that the future of VR is something to do with your smartphone.

and that probably means Android and iOS. And it probably doesn't mean custom hardware from particular people. It probably means something more or less standardized in the way that the smartphone got more less standardized. And the PC got more less standardized with differentiation within that. And so clearly Google, as the kind of the owner of the Android platform, is the people to drive that.

I sort of said earlier, like poor Samsung, you're kind of looking at the instant apps and you think, poor Facebook, you look at Daydream and you think poor Samsung. Because, you know, this is the logical place for this, be as a platform owner. And so that's what Google have done. And so there will be VR APIs on Android and there will probably be VR kit on the iPhone. So Connie, as our expert on China, here we talk a lot about what these big Silicon Valley-based tech companies are working on when it comes to VR. How do you think the Google Daydream News affects how VR is going to roll out

in China, given that it's going to be baked into the Android operating system, which everyone's building on top of? It's exciting, I think. That means, in some cases, poor Samsung, because you'll have a lot more competition. Yeah. I think also this is one, again, it's just the stark contrast and how they're approaching it. They're going to build what they view is the really hard part of the problem and sort of allow a barred set of partners to work on the part that they believe is easy.

In this case, it'll be, wow, they're all going to be smartphones that are going to differ by a little bit. So we'll build the hard part and let them. And then for the VR enthusiasts, it's going to be very interesting because there's fast forward this 12, 18 months. And you could see the reviews and the columns. And it's going to be the Samsung VR kit, the Xiaomi VR kit, and just like 12 different ones that are like the same but different.

And people are going to evaluate those. And then there'll be the Oculus one. And then there'll be the HoloLens one. And they'll be the whatever the Apple one is and that we don't know if there is one or not. But that is itself is going to show the stark contrast between these approaches. And it really is this analogy of like they're going to. about on an ecosystem of partners that are going to attempt to deliver an experience versus this all up.

And that's this pendulum that the whole tech industry has swung back and forth between vertical and horizontal for generations. You'll have Apple and Facebook trying to do the full stack. And you'll have Google saying, well, we're going to provide the hard bit and everything else will be a commodity around it. I do think that will end up, though, at least in the short term, having a big range of quality for the pen user experience. And it reminds me of, you know, years ago when a developer had to support all kinds of phones.

Yeah, I mean, I was looking at the Facebook, the Google event, of like, there'll be the phones that are certified for Daydream. And it's like, you know, back in the day when, like, games came with a sticker on the side that they were certified for... NVIDIA. Well, actually, it was like, you know, it was like 3DFX. It was like, you know, or Sound Blaster. Like, this game supports Sound Blaster.

You know, which VR headset do you have? Well, you know, today we say full stack because we, that's like a good metaphor for how the stuff is written. But, like, this is just the history of our industry. And it's particularly bold, given how early it is in AI and machine learning and in the VR space that they're willing to say, we think this is the commodity part and this isn't. And I think that is super, I think that's super interesting. And it's going to definitely, to Connie's point, man, these are going to be so wildly different in quality. Plus, people are going to take the ones that don't have the sticker and still try to make it work.

And so there's going to be a whole enthusiast world of like, you know, overclocking and root kidding and forcing whatever random handset you have to do that, it's going to be messy and innovative at the same time, which is pretty much how innovation happens. Yeah, it's going to be a bit like phones in the early 2000s, actually. Well, there'll be all sorts of shapes and sizes and colors and things, and then there'll be the resolution point, and then it explodes, and then there's a resolution point again. Stephen, as you said, as we were starting off, really excited about everything that was announced yesterday.

With that said, thank you all for making time to chat with me about everything that Google talked about, and yeah, that's it for this podcast. Cool. Thanks. Thank you.