Google to Use ML to Reduce Chrome’s Memory Usage

Machine Learning and Artificial Intelligence or ML and AI are two things that you hear Google talking about all the time. If you’ve been following news from Google I/O 2018, you’d have heard these dozens of times already. I am however yet to hear machine learning in connection with Google Chrome. That was till I spotted this code change request for Chromium:

Tab Ranker: Predict reactivation of tabs

Adds an ML model to score tabs based on whether they will be
reactivated. Currently this is only used in tests; a future CL will add
an experiment to use these scores as part of the tab discarder’s tab
selection algorithm.

The model includes hard-coded weights generated by tf.native. This is
documented in tab_ranker/native_inference.md.

Bug: 836898

This description makes it clear without any doubt. Google wants to improve “tab discarder” better with ML, Machine Learning. Now let’s take a look at what Tab Discarding is. Here is Google’s own definition:

Tab discarding allows Chrome to automatically discard tabs that aren’t of great interest to you when it’s detected that system memory is running pretty low. What do we mean by discarding? Well, a discarded tab doesn’t go anywhere. We kill it but it’s still visible on the Chrome tab strip. If you navigate back to a tab that’s been discarded, it’ll reload when clicked. Form content, scroll position and so on are saved and restored the same way they would be during forward/backward tab navigation.

In simple words, it is Chrome pausing one of your tabs when you are not actively using it. Google wants to use Machine Learning to improve Chrome’s ability to predict which tab can be paused.

This is the first time I am seeing Chrome getting help from machine learning. Don’t be surprised if you hear more about ML and AI in Chrome and Chrome OS! Stay subscribed.

Leave a Reply

Your email address will not be published.