Highlights from Google I/O 2018: AI, TensorFlow.js, Android, PWA, and WebAssembly

Preface

Google I/O is an annual developer conference where Google showcases its latest open technologies, covering AI, IoT, web, Android, cloud, etc. The 11th edition follows a growing focus on AI and the recent endorsement of Kotlin.

Artificial Intelligence for Everyone

The core theme of Google I/O 2018 was making AI accessible to everyone. Google demonstrated the power of its AI with a live Google Assistant demo that booked a restaurant, showcasing advanced TTS and conversational capabilities. The event also highlighted TensorFlow‑based projects such as a high‑school student’s app that identifies plant species and health, and encouraged developers to experiment with TensorFlow models via TensorHub.

Google also introduced hardware kits like Vision Kit and Voice Kit, though the Voice Kit demo showed some rough edges, indicating the technology is still maturing.

Exhibits such as Donkey Car and Teachable Machine illustrated the practical impact of AI, while the massive TPU 3.0 hardware sparked curiosity about its size and performance.

TensorFlow.js

TensorFlow.js was showcased as a way for front‑end developers to run AI directly in the browser. The Teachable Machine demo let users train a multi‑class classifier with a webcam and control an on‑screen character, demonstrating impressive results though large models can cause lag.

Android

Android JetPack – a collection of libraries that consolidate common Android development practices.

Android App Bundle – a publishing format that delivers optimized splits of the app based on the device’s characteristics, supporting dynamic delivery.

PWA

Progressive Web Apps were highlighted with case studies from Starbucks and Instagram, which reported a 20 % increase in mobile users. Google also presented Workbox, a library that simplifies PWA creation, and discussed the current state of service workers in China.

Web Assembly

Google demonstrated how WebAssembly powers complex web applications such as a browser‑based AutoCAD viewer, loading a 6 GB desktop application in under 30 seconds. Other examples included a PDF viewer built with WebAssembly that runs smoothly in the browser.

Afterword

The author notes that the conference mainly presented overviews and encourages readers to explore the full videos. The article ends with a recruitment call for Android developers, algorithm engineers, and front‑end engineers at Liulishuo.