A big theme for Google recently is the on-device machine learning applications. Last week Google officially launched the Android Pie. Inside the new mobile system, Google integrates many AI applications and open up the infrastructure for external developers.

For example, it learns the usage of apps and prioritize system resources for apps that are mostly used in each device, so as to save some battery. The new Android also has the smart text selection tool to recognize the selected text and suggest relevant actions. For example, it can detect the address information in the message and ask if the user want to open the map to search for the location. Later on, Google plans to add more support to recognize flight numbers, time, and IBAN, all through machine learning techniques. It also opens the technology to developers through its updated Smart Linkify API.

An example use of the smart text selection tool in Google's new Android Pie

In order to have all these machine learning models applicable in mobile devices, Google optimized the model complexity and latency of the model to speed up the inference on the mobile devices; it also upgraded its Neural Networks API to accelerate the on-device machine learning.

All of these efforts show Google’s commitment on bringing AI to mobile devices. it aims to improve the Android system and the developer ecosystem with the new AI technology, so that it can consolidate its dominance in the market.