IO 2017 Show & Tell at GDG Seattle

At Google Developer Group (GDG) Seattle I/O Show & Tell, I shared what I have learned from Google I/O 2017: AI/machine learning, TensorFlow and Android.

Link to the meetup event - https://www.meetup.com/seattle-gdg/events/239377155/

Transcript

1. What I learned from I/O ‘17 GDG Seattle I/O Show & Tell Margaret Maynard-Reid, 5/30/2017 @margaretmz MargaretMaynardReid @margaretmz
2. @margaretmz | #IOExtended | @GDGSeattle About me • Android developer • AI/ML student • Organizer & WTM Lead of GDG Seattle • Taught Android @UW 2
3. @margaretmz | #IOExtended | @GDGSeattle Topics for today • I/O keynote • AI/ML • Android 3
4. @margaretmz | #IOExtended | @GDGSeattle Google I/O 2017 • 3 day conference for Google developers, May 17 - 19 2017 • Took place at Amphitheater next Googleplex in Mountain View • Incredible learning opportunity! ◦ 150 tech talk sessions ◦ 100 office hours ◦ 85 codelabs ◦ 19 sandboxes 4
5. Google Keynote 5
6. @margaretmz | #IOExtended | @GDGSeattle Keynote opening: “Mobile first to AI first” 6
7. @margaretmz | #IOExtended | @GDGSeattle Google Lens • A set of initiatives that uses computer vision • Visual search and then take actions • Integrate with Google Home & Google Photos Link to Keynote on Google Lens 7
8. @margaretmz | #IOExtended | @GDGSeattle Google Photos Added 3 new features, all use computer vision: 1. Suggested sharing 2. Shared libraries 3. Photo books Link to Keynote on Google Photos 8
9. @margaretmz | #IOExtended | @GDGSeattle AutoML - auto machine learning • Use neural networks to design other neural networks. • Read more about it in this blog on Google Research here. 9
10. @margaretmz | #IOExtended | @GDGSeattle TensorFlow Cloud TPU • TPU was announced at I/O last year, 15x faster than a GPU Limitation: only used for inference / running the ML model • Cloud TPU: ◦ much faster than a TPU, ◦ chained together to form a pod ◦ Can be used for ML training as well A short video clip here for you to see the AI/ML booth at I/O 10
11. @margaretmz | #IOExtended | @GDGSeattle Actions by Google • Google Home ◦ Voice calls ◦ Bluetooth support ◦ New partners added: Spotify & HBO now etc. ◦ Visual responses (connecting to phones, tablets & TVs) • Support payment transactions • Support smart devices 11
12. @margaretmz | #IOExtended | @GDGSeattle Android now on 2 billion devices 12
13. @margaretmz | #IOExtended | @GDGSeattle Keynote wrapup: “TensorFlow + Everyone” 13
14. TensorFlow 14
15. @margaretmz | #IOExtended | @GDGSeattle TensorFlow (currently 1.2) The most popular ML open source library • 17,5000+ commits since Nov 2015 • 475+ non-Google contributors (v 1.0) • Significant external commits Source: I/O ‘17 talk TensorFlow Frontier 15
16. @margaretmz | #IOExtended | @GDGSeattle TensorFlow architecture Source: I/O ‘17 talk TensorFlow Frontier 16
17. @margaretmz | #IOExtended | @GDGSeattle Android meets TensorFlow Add TensorFlow Android Inference library and Java API to build.gradle dependencies { compile 'org.tensorflow:tensorflow-android:1.2.0-preview' } Link to talk: Android meets TensorFlow 17
18. @margaretmz | #IOExtended | @GDGSeattle TensorFlow Lite • Small & fast • Part of TensorFlow • Optimized for mobile • Available later this year 18
19. What is new in Android? 19
20. @margaretmz | #IOExtended | @GDGSeattle What is new in Android? Biggest Android announcements at I/O: • Kotlin • Architecture component • Android Studio 20
21. @margaretmz | #IOExtended | @GDGSeattle Kotlin - now official language for Android! • Open source under Apache 2 • Drop in replacement bi-directionally • Bundled in Android Studio (3.0) • Concise and expressive • Null safety • Many other nice features... The official guide to Kotlin for Android developer is here. (short video) 21
22. @margaretmz | #IOExtended | @GDGSeattle Architecture Components Three talks at I/O on architecture components: 1. Intro 2. Solving the lifecycle problems 3. Persistence and offline Official guide to architecture components is here 22
23. @margaretmz | #IOExtended | @GDGSeattle Android Studio 3.0 Preview Install the Canary version alongside the stable version There are 3 major features: 1. App perf profiling tools (CPU, Memory & Network) 2. Support for Kotlin 3. Improve build speed for apps with large project size 23
24. What I learned from I/O 2017 summarized in my sketchnotes What I learned from I/O 2017 summarized in my blog post 24