[MUSIC PLAYING] [APPLAUSE] JASON TITUS: Good
afternoon, everyone. I'm Jason, and it's great to be
back here at I/O. This morning, you got to hear all
about what we're doing for our billions of users. Now I'm excited to share what
we're doing for all of you-- the developers who are
solving everyday problems in powerful ways. Whether you're joining
us here at Shoreline or watching via the
livestream or joining us through one of 500 I/O
Extended events, welcome. [APPLAUSE] One of the reasons
I love my job is that I get to meet with
developers from all around the world and
hear what you're up to and what you're
finding difficult. It helps me understand
what we need to do as Google to
make things easier, as well as what we can do
better in our own products. And one thing that
becomes instantly clear, from Lagos to Warsaw to
Jakarta, is how important the broader
development community is in helping each
other figure out things and solve problems together. This kind of
pay-it-forward culture-- it's one of the great things
about working in this industry. And I am constantly inspired
by the stories I hear out of our Google Developer groups
and Google Developer experts solving this [INAUDIBLE]. [APPLAUSE] I see we have a few
of them here today. Every year, with
their help, we're able to reach more than
one million developers through in-person events
across 140 countries. And our Women
Techmakers program is able to engage more than 100,000
women advance their careers each year. [APPLAUSE] Our developer community
is truly important to us. And all of this would not be
possible without their efforts. It's also great that
in the last two years, we have had the number of
Google Developer experts double. These are amazing people,
like Rayan Al Zahab. [APPLAUSE] She both founded her local
Google Developer Group and Women Techmakers Lebanon. On top of that, she started
a web development training program for Lebanese and
Syrian refugee girls. We also have folks like Rebecca
Franks, an IoT and Android expert in Johannesburg. [APPLAUSE] Out of her passion for
cultural preservation, she works on an
open source app that lets children read books in
their native African languages. Google would not
be what it is today if it weren't for developers
like Rayan, Rebecca, and all of you. So thank you. [APPLAUSE] As you heard this morning,
Sundar was calling out-- the pace of AI innovation
is breathtaking. A whole new set of
capabilities is going to change the way we do things. And what's really exciting
to me is that we're at an inflection point. AI used to be something that
only deep experts and PhDs could use, and now it's
becoming accessible to everyone. Let me show you what I mean. [VIDEO PLAYBACK] - I would totally
describe myself as an average high schooler. I like hanging out with my
friends, watching movies, getting my nails
done, and lately I've been into machine learning using
convolutional neural networks. I don't know why
you're laughing at me. - Since childhood, she's always
wanted to know everything. - Why is that? How is that? What does it do? She just wants to be
always learning, moving, doing different stuff. - I haven't always been
into computer science, but my mom grows rose
bushes in my front yard. Every season, they'd get
disease, and then my mom and I would have to diagnose it. - Knowing Shaza, she wanted
to do something about it. - I had the idea as
a potential thing I could do for my research
class with Miss Son. - Why anyone would sit around
their summer vacation teaching a machine to identify plants-- not sure, but that was
what she wanted to do. - I wanted to have a way people
could diagnose plant diseases just by taking a photo of it. And so that's when I started
looking into TensorFlow. I'd watch different
tutorials and read blog posts every night. - Shaza took it upon herself
to do all the background research necessary and then
start to ask, what can this really do? - PlantMD works when a user
takes a picture of the plant, and it tells you
what plant it is and whether it's
healthy or diseased. And if it is diseased,
what disease it is. - My first question
was, is it an app that's only on your phone? And she's like, no. Somebody has downloaded
this app and using it. That was, like, wow. Something that I
just was so proud. - I don't think you have
to be a super genius to get into coding. Really anyone can do it with
an idea and with perseverance. I feel like open
source technology and the wealth of
information on the internet is empowering my generation. I know that I can do
anything I put my mind to, and so can anybody else. [END PLAYBACK] JASON TITUS: Amazing, right? [APPLAUSE] And we're delighted to
have Shaza and her family here with us today. [APPLAUSE] So you can see why we're
so excited about bringing technology like TensorFlow
and ML Kit to all of you. As developers, you can drop
modules into your applications, add a few lines of
code, and give them intelligent new capabilities. And this is only the
beginning of what is possible. As a company, we're committed
to empowering developers with the latest technology
to build things that matter. And today, we'll
show you what we're doing to make that easier. So over the next
hour, we're going to talk more about new
capabilities we're releasing, as well as the improvements
we're making to the platforms and tools you use every day. So with that, let's
get started with Steph, who's going to tell you
the latest on Android. [MUSIC PLAYING] STEPHANIE CUTHBERTSON: Android
is growing all over the world with billions of devices, and
every month, billions of apps downloaded. Our developer community
is growing rapidly, right alongside. In China, India,
Brazil, the number of developers using
our IDE almost tripled in just two years. With all this growth,
we feel a responsibility to support this vast ecosystem. And listen, if you feel
like your feedback drives what we say here each
year, you are right, like Kotlin last year. [APPLAUSE] Since we made it a
fully-supported language, we've launched more and more
support, taking advantage throughout Android. Already, 35% of pro
developers use Kotlin. That number grows every month. And 95% of Kotlin users
say they're really happy. More and more
Android development is going towards Kotlin. We are committed
for the long term. And if you haven't
tried it, I would. Just as your feedback
shaped investments so far, like Play Console, Android
Studio, and Kotlin, your feedback shaped
this year, too. So let's cover
three things today. First, distribution-- making
apps radically smaller so you get more installs. Second, development--
helping you build faster, with better APIs. And third, engagement-- bringing
users back more and more. Let's go straight
into driving installs. So Android is growing. That's great. App size is also growing,
and that is not great. Apps are targeting more
people and more countries, which means APKs have more
languages, more features. The larger your app gets,
the less installs you get. And most people think that's
an emerging markets issue, but it's true in all countries. So how can we make it easier
to build smaller apps? Our best idea was hard for us. It meant re-architecting our
entire app-serving stack. But it was the
right way to do it. Today, we are excited
to announce the new app model for Android. Using the Android App Bundle,
a new publishing format, you can dramatically
reduce app size. Apps need a lot of resources
to work on every device. The bundle contains it
all, but modularized. So when you download, Google
Play's new dynamic delivery only delivers the code and
resources a specific device needs. Now, we've tested
this with many of you, and we've seen huge savings. LinkedIn saved 23%. Twitter saved 35%. [INAUDIBLE] saved 50%. Now, if you're wondering, how
many devices does that work on? The answer is 99%. And that's because it's built on
long-standing platform concepts like splits and multi-APK. We also wanted this to be
almost no work for you. And so I'd love to have you
see a demo from someone who's been instrumental, along with
several teams who built this together. Please welcome Tor. [APPLAUSE] TOR NORBYE: Thank you, Steph. So here I am, in Android
Studio, working on my app. And I'd like to make it smaller. So what I'll do is open
up our APK Analyzer. And as you can see, most of the
download size is for resources. We have several
large image folders with different densities. So this is an app
which can really benefit from the new
dynamic delivery facility. To use it, I don't have to
change a single line of code. All I have to do is rebuild
my app as an app bundle. So I'll invoke the
Generate APK dialog. As you can see, we have a new
Android App Bundle option. Next, I point to my
KeyStore, and I can also tell it to export my
encryption key for Google Play. This is what lets the Play
store break apart my app and reassemble it
into smaller versions. So then I'll trigger the build. And once that's done, we can
take a look at the bundle file we just created. So as you can see, this
looks just like the APK files that you already know and love,
but there's some extra metadata here, which is what lets
the Play store do its magic. So now, in the interest
of time, let's assume that I've already uploaded
my app bundle and my sign in key to the Play Console. So then I land here. We have a new page in the Play
Console called the App Bundle Explorer. Take a look at the big
number on the right. This tells me that when
I publish this update, I'm going to save nearly 29%
on downloads for my users. And this page lets me drill
into just how that's possible. So in short, creating Android
app bundles with Studio is super easy, and
I hope you'll all try it out and see similar gains
for yourselves and your users. STEPHANIE CUTHBERTSON: And
all this is available now. App bundles and dynamic
delivery are launching for production use today. [APPLAUSE] Now, we're also working
to increase installs in other ways. At GDC, we announced
Google Play Instant. So you can try a game
without having to install. Now, games who have adopted
this have increased the number of players by up to 20%. Today, we're announcing that
this is available for all game developers. If you want to try
it right now, try launching "Candy Crush Saga,"
because that launches today. And to make development
easier, we're also releasing a Unity
plugin and Cocos integration. Now let's dive into
our second theme, making app development easier. Android's APIs could be easier. One person said,
on Android, there's six ways to do everything. Last year, we launched
Architecture Components. It was a testbed for
new ideas, starting in top areas that you flagged,
like life cycles and data. Today, so many top apps
use these in production. And more than half
of you have said you already use them or
plan to in the next year. Today, we're announcing Android
Jetpack, the next generation of Android APIs to accelerate
Android development. Jetpack is a set of
libraries and tools. We set the basic DNA by
including support library and architecture components. Jetpack brings these
together coherently and adds even more new
libraries, work management, navigation, paging, slices,
and Kotlin extensions across everything. All libraries are
backwards-compatible, which means they work
on 95% of devices. So it's been a pain to
schedule background tasks. Now, with Work Manager, you
get a single, easy-to-use API that works nearly everywhere. And Jetpack is all
about concise APIs. Those of you who've
tried it say you're writing up to 1/3 less code. Jetpack and Kotlin
are intentionally designed to work
together so you write only the code you need for a
pleasant reading and writing experience. And Jetpack saves time by
embodying opinions about what we found works best for Android
development, like RxJava or material design. Jetpack's APIs are
integrated with the IDE, too. For instance, Android Studio
now includes a Navigation Editor which works with the library. So you can visualize
your app flow, almost like you're
sketching on a whiteboard. You can add new screens,
position them in your flow. And under the covers,
we'll help you manage the back stack,
conditional flows until you get it just right. Overall-- [APPLAUSE] --IDE tools, we think,
are great helpers to make development fast. That's why everything you've
seen bundles using Jetpack comes with Studio support. And the team also works on
making everyday tasks faster. You told us, work on
emulator boot time, and I'd like you to see it now. TOR NORBYE: All
right, so first, let me show you how quickly
our emulator can start now. Ready, set, go. As you can see,
it's nearly instant. [APPLAUSE] I was not cheating. It was not running
in the background. The reason it's so fast is
that we support snapshots. So we store the full
state of the emulator into a file that we can
then load back quickly. And you can create these
snapshots yourself. So now I'm going to
bring back a snapshot I took in the middle of a
complex OpenGL 3D stress test. Ready, set, go. So you can see it's
about two seconds, and it's up and running. STEPHANIE CUTHBERTSON: There
are more speed enhancements. We added an energy profiler,
integrated an improved system trace. There's a C++ profiler now. We promoted D8
compiler to stable after testing it on our
own Android platform, which means smaller, faster
binaries by default. And we added an ADB Connection
Assistant to fix, hey, why can't I connect to that device? So that's Android development--
faster and easier. We hope you try Android Studio
and Jetpack previews today, including all the alpha
stage new libraries. Jetpack is just a
beginning for us. We are testing so
many more new ideas, and we hope you watch for
them in the months ahead. OK, so once your app is
built, it's installed, we want to get
users coming back. The slices Dave showed
are a cool new way to drive reengagement. We wanted these to
be easy to build. So you'll find templates
that are rich and flexible so you can compose. You can start with
something simple, like a set of rows or grids. Then you can add content,
like text and images. But not just static
content, you can also house real-time data
and rich controls. And once you get the pieces
together in a good setup, you can add the code to
make slices interactive, like pause a song or
go to the next one. So then you end up with this
cool mini-snippet of your app. And because it's Jetpack,
slices work on 95% of devices, showing the power of building
new features in a Jetpack world. This is also ready for
everyone here to try today. Your slices will start showing
up in Search this summer, timed with the P launch, and
Assistant later this year. So that's a quick tour of
some of the biggest new areas of investment for Android
and how your feedback has been shaping the landscape. Now let's switch gears and
talk about cool devices. At CES this year, we announced
that Lenovo, Harman, LG, and iHome are all building
consumer products powered by Android Things-- a powerful platform for
developing IoT products. This week, Android
Things graduates from developer preview to
1.0, ready for everyone to build commercial devices. And everyone here at I/O will
get a free Android Things developer kit. [APPLAUSE] If you want to get it first,
join the scavenger hunt today, or everyone here can
pick one up tomorrow. As we continue to expand to
more devices and more surfaces, our team and the
Assistant team have been working closely together. To hear more, please
welcome Brad Abrams. [APPLAUSE] BRAD ABRAMS: Awesome. Thanks, Steph. This morning, you heard how the
Assistant is becoming even more conversational and visual,
helping users get things done, save time, and be more present. And developers like you have
been a big part of the story, making the Assistant more useful
across 500 million devices. DoorDash, Fitbit, Todoist,
Starbucks, Disney, and many, many others
are engaging with users through Actions that they build. And smart home OEMs like GE,
Arlo, Xiaomi, and Logitech have made more
than 5,000 devices that work with the Assistant. In total, the
Google Assistant is ready to help with over one
million Actions built by Google and all of you. And the platform's momentum
has been growing every day. It's now available across 16
languages and multiple devices, including phones, smart
speakers, TVs, cars, watches, headphones, and whole
new categories of devices like smart displays,
which are coming in July. And we're delighted to
see that many of you are starting to test the waters
with this new, emerging era of conversational computing. In fact, over half
a million developers are learning and building with
Dialogflow, our conversation development tool. Today, I'm excited
to share how we're making it even easier for app
developers and web developers to get started with
the Google Assistant. You can think of your
Assistant Action as a companion experience to your main
property that users can access from a smart
speaker, even in the car-- wherever they're not using
their phone or laptop. And if you want to personalize
that Action for users, account linking lets
you easily share state between your
app and your Action. And now, with seamless
digital subscriptions, your users can enjoy the
content and digital goods they purchased in
your Android app directly in your
Assistant Action. Take "The Economist,"
for example. Because I'm a premium
subscriber in their app, I can now enjoy that
same premium content on any Assistant-enabled device. But of course, creating
engaging actions doesn't end with
digital subscriptions. As you saw in the
demos this morning, the Assistant is becoming a
canvas that blends conversation with rich visual interactions
for the phone and other devices like smart displays. And starting today,
you can deeply customize the appearance
of your Action. You saw a glimpse of what
was possible earlier today with demos from
Tasty and Starbucks, but let me show you another one. Check out the game "King for a
Day" here on a smart display. It looks beautiful here, on
the phone, and on the TV. Of course, once you build
an Action for the Assistant, you want to get lots of people
engaged with that experience. And for that, I've got
three things to share. First, we're making
it even easier for you to promote your
Action with something new we call Action Links. These are hyperlinks
that you can use from anywhere that point
directly into your Action. Let me give you an example. Headspace has built
a great experience for the Google Assistant that
can help people meditate. Now, when they have
some new content, they might share a
blog post about it. This post contains an Action
Link right there at the bottom. And that special link triggers
the Headspace Action directly in the Google Assistant. If you've already
built an Action and you want to spread
the word, starting today, you can visit the Action Console
to find your shareable Action Link. OK, so now that you've
acquired some new users, you want to engage them. And for this, we've got
Action Notifications. Once users opt in,
Action Notifications gives you a way to connect
with them about new features and content. These notifications
work on the phone, even if users don't have
your Android app installed. And now, with
cross-surface notifications coming to the
Assistant, you'll be able to reengage with
your users on speakers, smart displays, and other
Assistant-enabled devices. But to consistently
reengage with users, you need to become part
of their daily habits. And for that, the Assistant
supports routines. This is the ability to
execute multiple actions with a single command for things
like waking up in the morning, getting to work, or
many other daily tasks. And now, with
Routine Suggestions, after somebody engages
with your Action, you can prompt them to add
your Action to their routine with just a couple of taps. So for example, on my
way to work each morning, my Assistant can tell
me how to beat traffic, and it can also help me order
my Americano from Starbucks. We're excited to see how Action
Links, Action Notifications, and Routine Suggestions will
help you drive engagement. But the broader challenge
of helping people connect with the right
Action is reminiscent of the early days of the web. Over the past 20
years, we've built up a lot of experience
in connecting people with the right information,
services, and content. And we're putting
that expertise to work in the Google Assistant. For example, when somebody
says, hey, Google, let's start a maps quiz, the
Assistant should immediately suggest relevant games. For that to happen,
we need to understand the user's basic intent. And that's hard,
because there are thousands of ways that users
could ask to play a game. To handle this complexity,
we're beginning to map all the
ways that users can ask for things into a
taxonomy of built-in intents. Today, we're making the first
set of those intents available. And with these, you'll be able
to give the Assistant a deeper understanding of what
your Action can do. We'll be rolling out
hundreds of built-in intents in the coming months. So with that, I'm excited
to see how you all extend your experiences to the
Google Assistant, how you'll build rich, immersive
interactions, and create consistent engagement
for your users. To learn more and get started
building Actions today, visit actions.google.com. And with that, let me introduce
Tal, who will tell you about the web platform. Thank you. [APPLAUSE] TAL OPPENHEIMER: Thanks, Brad. I don't think it's
an exaggeration to say that the
web is the world's most critical
resource for ensuring the free flow of information. The web is a fundamentally
open and independent platform. So for developers,
the web makes it possible to reach users around
the world on almost any device. And for users, the web
provides a truly frictionless experience. You tap on a link
and load a page. And these properties
have allowed the web to reach a massive scale,
with over five billion devices accessing the web each month. And here at Google, from
the very earliest days of page rank to building
our very own browser, we've been deeply invested
in the continued growth and reach of the web. And as part of this,
we have two main goals. First, to make the
web platform itself more powerful and more capable. And second, to build
tools to help you easily take advantage of this power. Over the past few
years on Chrome, we've worked alongside
other browsers to add capabilities
to the platform to support new web
experiences we've been calling Progressive Web Apps, or PWAs. PWAs are websites that take
advantage of modern web platform APIs to build
experiences that can do things like work while
the user's offline, send push notifications, or be
added directly to a user's home screen. And universally, businesses
that have built PWAs have seen incredible results. Take Instagram. Instagram launched
their PWA last year to increase their reach to
users with low-end devices. And they were able to double the
retention of their web users. Times Internet
has been launching PWAs across their
products and saw an 87% increase in time spent
per user for their "Economic Times" PWA. And when the Starbucks
team rolled out their PWA, they doubled their daily
and monthly active users on their website. And because the web adapts
seamlessly to different devices and platforms, their
mobile PWA also worked well for their
desktop web audience. In fact, they found
that the number of orders placed on
the desktop version has grown to be about equal
to the number of orders placed on their mobile version. And it's not just
these businesses. Across sites that
advertise with Google, we see an average
mobile conversion rate boost of 20% when the
site switches to a PWA. And we also build many
of our own products here at Google as PWAs. Google Maps launched a
new mobile PWA tailored to provide a fast and
data-conscious experience, and Google.com itself is a PWA. It loads 50% less
JavaScript over the network and can support features
like retrying search queries if you're offline. And investing in a PWA today
goes further than ever. Chrome OS now provides
native support for PWAs, allowing them
to be installed and run fully integrated and in
their own standalone window. And we're incredibly
excited that Service Worker, the underlying new API
that makes PWAs possible, is now supported on
all major browsers, including recently Edge
on Windows and Safari on both desktop and mobile. This is probably the most
important leap forward for the web in the last decade. So PWAs have fundamentally
changed what the web can do. But that's only part of it. WebAssembly enables websites
to run high-performance, low-level code written in
languages like C and C++, and it has broad support
across browsers and devices. And because this code has
access to all of the web's APIs, WebAssembly enables a
new class of content to run on the web platform. As just one example, the
AutoCAD team took a 35-year-old codebase-- that's older
than the web itself-- and were able to compile
it to run directly inside a browser
using WebAssembly. So now all of the power of
AutoCAD is just a link away. So the web platform's
been gaining all these great new capabilities,
but we want to make sure it's easy for
you to take advantage of them. So we've been working
on the tools to help. Lighthouse is a feature of
Chrome's built-in DevTools that analyzes your site and
gives you clear guidance on how you can improve
your user's experience. Half a million developers
are running Lighthouse against their site or as part
of their continuous integration process, to help avoid
performance regressions, or even to keep an eye
on the competition. And today, we're launching
Lighthouse 3.0, which makes Lighthouse's performance
metrics even more precise and its guidance
even more actionable. So Lighthouse helps
you understand how you can upgrade
your site, but we don't want to just give you advice. We want to give you
the tools to help make sure any new
sites you build are high quality by default.
We started the AMP Project two years ago to
help make building fast web pages much easier. And I'm happy to share that AMP
is evolving in some big ways. We're expanding the kinds of
things you can do with AMP. We've added a bunch of
features that support critical e-commerce
experience in AMP, like search auto-complete,
a full-featured date picker, and soon, infinite
scrolling lists. And businesses are
seeing the benefits. As just one example, Overstock
saw a 36% increase in revenue on their AMP pages. And we've introduced AMP
Stories, an easy-to-use format for creating immersive
stories on the mobile web. Now, all AMP content
benefits from a fast, free,
privacy-preserving cache that optimizes page loads. But they've had these
Google.com URLs. So we're fixing that with a new
standard called web packaging. This is also the first step
towards our ultimate goal for any fast,
responsive web content-- to be able to take advantage
of all of the benefits of AMP. And today, we're
announcing a new way to take advantage of
all of these tools. We introduced Chrome OS
almost seven years ago to showcase the best
of the web and make computing accessible to all. Chromebooks grew
50% this past year, both units sold and
28-day active users. And we've expanded to
tablets and attachables. But it's not just about
access to technology. It's also about
access to create it. And that's why we're expanding
Chrome OS to support developers with the ability to securely
run Linux apps on Chrome OS. [APPLAUSE] So this means that many of
your favorite tools, editors, and IDEs now work
on Chromebooks. So starting with the Dev
channel on Pixelbooks, you can now build for the
web on a platform built around the web. And soon, you'll
even be able to run Android Studio on Chrome OS. [APPLAUSE] So all in all, it's an
incredibly exciting time to be a web developer. Businesses around the
world are consistently seeing substantial
returns from deeply investing in their
web experience by building
progressive web apps. And the reach of these PWAs
is now truly everywhere, with support across
every major browser. And it's easier than ever to
build great web experiences with tools like
Lighthouse and AMP and so many more of our
web developer products that you'll hear
about throughout I/O. The web platform is
light years ahead of where it was in
Google's early days, and it remains as core
as ever to our mission. We're excited about the host
of new use cases made possible by today's modern web,
so be sure to check out the web platform state
of the union talk later today to learn more. And now, please
join me in welcoming Rich to talk about updates
to Material Design. [APPLAUSE] RICH FULCHER: Thanks, Tal. I'm going to talk about
Material Design's new approach to customizing our apps, and
share some of the new tools and resources we've created. We introduced Material
Design in 2014 as a system for creating
bold, beautiful, and consistent digital
products across platforms. We were responding
to the desire that we heard from you and
developers around the world for clear design
guidance, for advice that would make the
experiences you create better for your users. Material has become
the design foundation for all Google products. And you've taken it and launched
it into millions of apps. But we've heard two
sentiments especially clearly after hundreds of design
reviews with product teams and developers. First, that you didn't
always see Material as flexible enough, that
products from different brands looked too similar. And second, that our
engineering support for Material needed to be stronger so that
you could more easily realize your design vision. Our goal has always been to
provide more than just a design blueprint. We are committed to
delivering resources, tools, and engineering components
to make it easier for product teams to work
together seamlessly from design to development so
that you can deliver customized experiences for your users. That's why we're
proud to announce Material Theming, a major update
to the Material Design system. With Material Theming, we're
delivering a truly unified and adaptable design system. It enables you to express
your brand's unique identity using color, using type,
and shape consistently across your product. Material Theming retains
the consistent guidance from Material and expands on it. Let's take a look. [VIDEO PLAYBACK] - What if it was possible
to design and build better so your ideas could scale from
a single sheet to a system? With Material, you can
apply color dynamically so that every touch enhances
the character of your design. Typography is adaptable
so you can express more by customizing for
style and legibility. Craft an experience that's
responsive across device and platform. Streamline collaboration with
tools and open source code, making it easy to express
your brand's unique identity, balancing form and function
down to every last detail. Material Design
isn't a single style. It's an adaptable design system
inspired by paper and ink and engineered so you can build
beautiful, usable products faster. What will you make? [MUSIC PLAYING] [END PLAYBACK] [APPLAUSE] RICH FULCHER: Material
Theming puts you in charge. When you make just a few
decisions about aspects like color or type,
for example, we'll apply those throughout
your design. You'll see options
for customization across our design guidelines. And we've added hundreds
of new examples. All still Material, but
reflecting a much wider range of products and styles. When you see something
you like, you can use our handy
redline viewer to check the dimensions, padding, and
even the hex color values. Today, we're also
releasing two new tools to make it faster to go from
design to implementation. First, Material Theme Editor. This plugin for the
popular application Sketch helps designers
create and customize a unique Material theme. Change one theme value, and
it cascades across the design. Then that work can be shared
easily using Material Gallery. This is the tool used by
product teams at Google to review and comment
on design iterations, and it's now
available to everyone. When it's time to build, you
can see automatic redlines. You can identify exactly
which component is being used and how it's been customized. But we need to make all of
this easier not just to design, but to realize in code. So our open source
Material Design components for Android, iOS, the
web, and Flutter-- all with Material Theming-- are available today. [APPLAUSE] These are the same
components we use at Google to build our apps. And just as Google creates a
custom Material theme that best reflects our brand in products--
starting with the new Gmail for the web-- you can use these
same components to create an experience and a
theme that is right for you. More Material components and
Material Theming customizations will be coming soon,
so keep an eye out. We'll be launching regularly. And we'll keep
listening, so please keep the feedback coming. Material Theming--
a new approach to customizing your apps powered
by open source components, new tools, and design
guidance-- all available today. To make Material yours,
get started at material.io. And now, here's Jia to
talk about progress in AI. [APPLAUSE] JIA LI: Thanks, Rich. Hi, everyone. At Google, we believe
developers are driving the real-world impact of AI. Of course, we understand
how complex this can be. Machine learning is a
challenge in any environment, but today's developers often
target multiple platforms and devices. We also have to juggle
issues like processing power, bandwidth, latency,
and battery life. That's why we're committed
to making AI easier to use, no matter what
platform you are on. This begins in the
Cloud, which makes machine learning accessible to
customers in every industry. We believe success
in AI should be determined by your imagination,
not your infrastructure. We invest the best technology
to power Google's products, and Google Cloud is making these
resources available to you. For example, our Cloud
TPUs are widely available in public beta, making
machine learning more affordable than ever before. Until recently, training
a model to a high accuracy on the ImageNet data set
cost thousands of dollars. But in a Stanford
benchmark contest, our TPUs reached the same
level for less than $50. Cloud TPUs are now available to
everyone, and getting started is as simple as
following this link. And this morning, we announced
our new third generation TPU, demonstrating our ongoing
commitment to AI hardware. The Cloud also lets us
share our best technology through a growing range
of machine learning APIs. Our latest is Cloud
text-to-speech. Based on DeepMind's
WaveNet, the same technology behind the voices in
the Google Assistants, it generates speech with 32
voices in multiple languages and variants. In just a few lines of
code, your application can realistically
speak to your users across many platforms
and devices. For more sophisticated
interaction with your users, text-to-speech can be paired
with Dialogflow Enterprise Edition. Dialogflow makes it easy to
build a conversational agent, even with no prior experience. It understands the
intent and context of what a user says and
generates accurate responses, whether you are
shopping for clothes or scheduling your bike repair. Finally, Cloud AI is
always exploring new ways to put more power in
the hands of developers. This effort continues
with Cloud AutoML, which automates the creation
of machine learning models. Our first
widely-available release will be AutoML
Vision, which makes it possible to recognize
images unique to your vocation without writing any code. You provide the
training examples, and AutoML does the rest. Our alpha users are
already accomplishing a lot with AutoML, from
identifying poisonous spiders to helping the blind
better understand images. But Vision is just
the beginning. We're looking forward
to bringing AutoML to more machine learning
tasks very soon. To learn more about
all Cloud AI products, including AutoML early
adoption, please visit our site. But Cloud AI technology
isn't the only thing that we're pushing
in new directions. TensorFlow has become
a standard in machine learning since we open
sourced it two years ago, with 13 million downloads. Now we're focusing on
bringing it to new platforms. We recently announced
TensorFlow.js, which brings machine learning
to millions of web developers through JavaScript. New models can be created
right in the browser, or on the server
side through Node.js. While models trained offline
can be imported, and run with WebGL acceleration. Of course, there is a big
world beyond the browser. We also introduced
TensorFlow Lite last year, which brings machine
learning to Edge devices, including Android
and iOS phones. This allows offline
processing with low latency and ensures sensitive data
never leaves the device. TensorFlow Lite also supports
hardware like the Raspberry Pi, so smart devices like the AIY
Project Vision and Voice Kits can leverage it as well. As Dave announced
earlier this morning, we've released ML Kit
in beta, an SDK that brings Google's machine
learning capabilities to mobile developers
through Firebase. The same technology
that has powered Google's own experiences,
like text recognition in Google Translate and
Smart Reply in Gmail, will be available to
power your own apps. Our vision for AI at Google is
turning the latest technology into products that make
better life for everyone. But we cannot do this without
creative developers like you. Together, we can
bring AI to the world. To talk more about how Google is
supporting mobile development, let me introduce Francis Ma. Thank you. [MUSIC PLAYING] [APPLAUSE] FRANCIS MA: Thank you, Jia. Our mission for Firebase is to
help mobile app teams succeed by providing a
platform to help you solve key problems across
the lifecycle of your app-- from building your app
to improving app quality to growing your business. We've come a long way
since we expanded Firebase two years ago from
a real-time database to a broad app
development platform. And it's so exciting
to see there are now over 1.2
million apps actively using Firebase every
month, including many of the top Android and iOS
apps like Pandora, Pinterest, and Flipkart. We appreciate so many of you
are trusting us with your apps, whether you're a sole developer
or working in large teams. We're committed to helping
developers at companies of all sizes to succeed. Now, one of our key
goals is to help you take care of the critical,
but sometimes less glamorous parts of app development so you
can focus more on your users and build cool stuff. As an example, we've
worked hard to ensure we're ready to meet
your compliance needs with many privacy
and security standards, as well as with
the upcoming GDPR. We are also continuing
to expand the platform to further simplify everyday
developer challenges. A little over a year
ago, the Fabric team joined forces with us to bring
the best of our platforms together. And we've made a lot
of progress since then. In the last couple
of months, we've brought Crashlytics, Fabric's
flagship crash reporter, into Firebase. And we've improved
it by integrating it with Google Analytics so
you can see what users did in your app that led
up to a crash for much easier diagnosis. With the combination of
Crashlytics, performance monitoring, and
Google Analytics, Firebase is not just a platform
to help you build your app's infrastructure, but also to
help you better understand and improve your app. Another major set
of advancements we've been making to
Firebase over the last year is introducing machine
learning to the platform. A few months ago, we
took our first step with the release of
Firebase Predictions. Predictions applies ML
to your analytics data and predicts the future
behavior of your users so you can take proactive
actions to optimize your app. For example, you can lower
the difficulty of your game for users who are likely
to abandon it, or send special offers to users
who are likely to spend. And you can also run A/B
tests with different setup to see which of
these performed best. So Predictions
was the first step that we took to bring the power
of Google's ML to work for you. And today, we're
taking our second step with the release of
ML Kit in public beta. ML Kit-- we're bringing
together Google's machine learning technologies
from across Google and making that available
to every mobile developer working on Android and iOS. ML Kit provides five
out-of-the-box APIs, like image labeling
and text recognition, and these APIs can run on-device
for low latency processing or in the Cloud for
higher levels of accuracy. You can also bring your own
custom TensorFlow Lite model if you have more specific needs. So let's talk through
an example of how ML Kit can be used in an app. Now, I'm a dad of two. And my young kids are always
so curious about the world around them. And I thought it'd be
fun to build an app where we can take pictures
and identify objects in them. And by using ML Kit's
on-device image labeling API, it's easy for me to apply ML
to identify different objects-- it's a dog, a tree, or a bridge. Now, as many of you are aware,
one of the key challenges of building for mobile is
having a low latency response and assurance that
it can work even without a network connection. And by using ML
Kit's on-device API, I can be assured that
image labeling will work, even if I'm out on a remote hike
with no network connectivity. So on-device APIs are great
for low latency processing, but there are
other times where I want to optimize for
highest accuracy possible. For example, when
identifying landmarks, I want my kids to know this is
the Golden Gate Bridge and not just any bridge. And ML Kit's Cloud-powered
APIs give me that. They have a much higher
level of accuracy and can be easily integrated
right from my app client code. Then there are other situations
where you may have more specific needs than what the
out-of-the-box APIs can cover. And for those times, you'd
want to bring your own custom ML models. Now, as mobile developers,
we know how important it is to keep your
app binary small and be able to iterate on
the experience rapidly. With ML Kit, you can
upload your TensorFlow Lite model and [? surf ?]
that through Google's global infrastructure. Your app can dynamically
retrieve these models and evaluate them on-device. This means you don't need to
bundle the model with your app binary, and you
can also update it without re-publishing
your entire app. And since ML Kit is
available through Firebase, it's easy for you to take
advantage of the broader Firebase platform-- for example, experimenting
with different ML models using A/B testing, or
storing your image labels with Cloud Firestore, or
measuring processing latency with performance
monitoring, or understanding the impact on user engagement
with Google Analytics. We are so excited
about the possibilities that machine learning
unlocks, whether you want to supercharge your
growth or build amazing user experiences. We want to harness
Google's advances to help you build
and grow your app. And with that, I'd like to
turn it over to Nathan Martz to talk about another area
of exciting advancements in mobile computing-- augmented reality. Thank you. [MUSIC PLAYING] [APPLAUSE] NATHAN MARTZ: Hey, everybody. I got to say, it is incredibly
exciting to be here today. At Google, we believe that
augmented reality represents one of the most exciting
advances in mobile computing today. By enabling our devices to
see and sense the world much like we do, AR allows us to
interact with digital content and information in the context
of the real world, which is exactly where it's often
the most useful and accessible. And as our phones learn to
see the world in new ways, they unlock new possibilities
for the kinds of experiences that developers can create. That's why, just
three months ago, we launched ARCore, our platform
for building augmented reality experiences, to
allow you to take advantage of this
incredible new technology. And we've already seen
some amazing apps. For example, you can
now create a floor plan just by walking
around your home. You can visualize
the intricacies of the human nervous
system at real-world scale. Or if you like, transform
your dining room table into the home of your
next virtual pet. And as you've been
building, we've been learning and
listening to your feedback. Today, we're rolling
out a major update to ARCore to help you create
even richer, more immersive and interactive experiences. First, we know that
creating a 3D app can be challenging,
especially when you have to write directly to
lower-level APIs like OpenGL. That's why we've created
Sceneform, a brand new 3D framework that makes it
easy for Java developers to create ARCore applications. And the thing is,
Sceneform is not just for people creating
apps from scratch. Its design actually
makes it really easy to quickly add ARCore
features to apps that you've already released. Under the hood,
the Sceneform SDK includes an expressive API,
a powerful, physically-based renderer, and seamless
support for 3D assets inside of Android Studio. Best of all, Sceneform
is optimized for mobile, architected from the ground-up
with performance, memory, and binary size in mind. So we also know that
sometimes as a developer, you want to build
an AR app that can react to the specific physical
objects in the real world. That's why today we're
introducing Augmented Images, a new capability in ARCore that
makes it possible to attach AR content and experiences
to the physical images in the real world. And Augmented
Images doesn't just detect the presence
of a picture. It can actually compute
the precise 3D orientation of that image in real time. So now you can use Augmented
Images to do things like bring a textbook to life,
create new forms of artwork. You can even see what's inside
the toy box before you open it. Finally, though, we
know that if we truly want our computers to
work more like we do, they need to involve not
just the things in our world, but the people in our lives. Fundamentally, our time
here is spent with family and friends and colleagues. We collaborate, we create, we
share experiences together. That's why we believe that
every incredible thing that you can do by yourself
in AR today, you should be able to do
together with someone else. It makes me very proud to
announce the next major step in how our phones
see and understand the world, a new
capability in ARCore that we call Cloud Anchors. With Cloud Anchors, we
actually allow multiple devices to generate-- thank you-- a shared,
synchronized understanding of the world so that multiple
phones can see and interact with the exact same digital
content in the same place at the same time. And this allows
you, as a developer, to create applications that
are collaborative and creative, multi-user and multiplayer. To show you exactly
what I mean, we actually integrated Cloud
Anchor support into our experimental, open source,
AR drawing app "Just a Line." Let's take a look. [VIDEO PLAYBACK] [MUSIC PLAYING] So yeah. [APPLAUSE] Thank you. We know that Cloud Anchors will
enable many new kinds of AR experiences. AR experiences that
combine the power of your device, your creativity,
and the people around you. But because these
experiences are so powerful, we believe that they
should work regardless of the kind of phone you own. That's why I'm
very excited to say that we're making
ARCore's Cloud Anchors available on both
Android and iOS devices. [APPLAUSE] It's pretty awesome. So everything that I've
just talked about today, from Sceneform to
Augmented Images to Cloud Anchors and more, is available
for you to use right now. You can learn more about that
and start creating your own ARCore-powered app by going
to developers.google.com/ar. And with that, I'm going
to hand it back to Jason to wrap things up. Thank you all very
much for your time. [APPLAUSE] JASON TITUS: Thanks, Nathan. Hopefully you can
see from our talks today that we're
committed to meet you where you are, no matter what
platform you're building on, and help you take advantage
of the latest innovations, from machine learning
to augmented reality. And much of what
you heard of today is built on top of
our Cloud technology. This includes powerful APIs
around machine learning, smarter data analytics
with BigQuery, and a focus on low latency
and high reliability. And whenever possible,
we try and take things out as open source, like
Kubernetes and TensorFlow, so they can be broadly
adopted and openly evolved across the industry. With our Cloud technology,
we're taking the lessons we've learned building
large-scale software and bringing them to you. So I/O, this year,
is bigger than ever. We've got some great content
lined up for the next few days. And you'll see we're doing
something new this year-- inspiration talks. They take a broader look at how
the technology that you build affects the world around us. So we cover topics like building
for a better tech-life balance, the future of
computing, and using AI to transform health care. We all have a responsibility
for what we build as well as the people
that we build for, so I hope you have a
chance to check these out. And one of my favorite
parts about having I/O so close to
campus is that we're able to bring in more
than 2,000 Googlers who built these products. They're here not
only to help you, but more importantly,
to get your feedback on what we can do better. My hope for this conference
is that each of you walks away feeling that
you can do something that you couldn't do
before, and that you're able to use that knowledge
to build things that matter. Thank you very much, and
have a great Google I/O. [APPLAUSE] [MUSIC PLAYING]
