Skip to main content

Google's New AI Approach Will Put iPhone to Shame

Google has set its sights on using artificial intelligence to improve your smartphone experience. And it's starting with a closer look at your smartphone's autocorrect features.

Credit: Sam Rutherford/Tom's Guide

(Image credit: Sam Rutherford/Tom's Guide)

In a recent blog post, Google describes a method called "Federated Learning" which uses artificial intelligence to boost the company's software services. Specifically, the method describes how Google collects information about your actions on your device, stores it locally, and then sends an update every now and then to its servers where it can compare those actions against other users. The end result, Google says, will be better autocorrect suggestions on your phone.

Using artificial intelligence to improve autocorrect suggestions is nothing new. However, Google and other companies have long sought to collect information about how you use a device, store it in the cloud, and then use that collective data to retrain algorithms to improve their services. 

More: iPhone 8's AI Boost Will Have Surprising Benefit

But Google contends Federated Learning is a better approach. According to Google, the traditional method of using AI to fine-tune services takes a long time to collect all your data, send it to the cloud, compare it to the behavior of others, make algorithm changes, and then send those algorithm changes to your device. With Federated Learning, all of your actions are stored on your device and artificial intelligence features built into your software are learning on the fly to deliver you better results.

Google says it will only occasionally request your anonymous data so it can generally improve its broader algorithm after comparing your data with all the other anonymous data it collects.

Google said that it's currently testing the feature in its Gboard keyboard app on Android devices. The app analyzes how users respond to suggestions and will attempt to improve them as time goes on. The information Gboard collects on your device is sent back to Google for interpreting and updating its app.

But there's more to Federated Learning than just smarter and faster access to autocorrect suggestions. Google says that Federated Learning delivers improved privacy by requiring far less communication between your device and Google's servers. And all of the updates are sent over encrypted channels to limit a hacker's ability to access your data.

Google hasn't said whether it will expand Federated Learning beyond Gboard in the future.