At Google, we employ technologies like machine learning (ML) to create more helpful products, such as screening out email spam, updating maps, and providing more relevant search results. Chrome is no exception: we use ML to make web pictures more accessible to those who are blind or have limited vision, and we also produce real-time subtitles for online films to help people who are hard of hearing or live in loud situations.
Because this work in Chrome is ongoing, we wanted to highlight some current and upcoming ML enhancements that provide a safer, more accessible and customised browsing experience. Notably, these updates are enabled by on-device ML models, ensuring that your data remains private and never leaves your smartphone.
More peace of mind, less vexing prompts
Every day, Safe Browsing in Chrome protects billions of devices by displaying warnings when users attempt to visit risky websites or download dangerous files (see the big red example below). We launched a new machine learning model in March of this year that identifies 2.5 times more potentially harmful sites and phishing attempts than the prior model, resulting in a safer and more secure online.
We’re also changing how consumers interact with online alerts to improve the surfing experience. Page notifications, on the one hand, assist send information from sites you care about; on the other hand, notification permission requests can be annoying. Google Chrome predicts when permission prompts are unlikely to be given based on how the user previously interacts with similar permission questions and silences these unwanted messages to enable people to browse the web with little interruption. We’ll be releasing an ML model that produces these predictions totally on-device in the future version of Chrome.
This is what you will see if a phishing attempt is detected in Chrome’s next release (Left), and Chrome will present permission requests silently when the user is unlikely to approve them (Right).
Finding out what matters, always in your language
We created Journeys earlier this year to let consumers retrace their journeys online. As an example: You may spend weeks researching attractions, comparing flights, and buying gear before visiting a national park. Chrome links together the pages you’ve visited regarding a certain topic with ML and Journeys, making it simple to pick up where you left off (vs. SCR o o o l l ling through your browser history).
When it comes to hiking boots and camping guides, we’re also leveraging machine learning to make those web pages available in your preferred language. We’ve released an upgraded language identification model in particular to determine the language of the website and if it needs to be translated to meet your preferences. As a result, tens of millions more successful translations are being produced every day.
A browser designed specifically for you
Perhaps you enjoy reading news stories in the morning – phone in one hand, cereal spoon in the other – so you share a lot of Google Chrome links. Perhaps voice search is more your style, as you slip in a few inquiries throughout your commute to work. In any case, we want Chrome to meet you where you are, so in the near future, we’ll be using ML to change the toolbar in real-time, highlighting the action that’s most beneficial at the time (e.g., share the link, voice search, etc.). Of course, you will be able to manually tweak it as well.
Our objective is to create a Google browser that is truly and consistently useful, and we’re enthusiastic about the potential that machine learning offers. At the end of the day, your experience is truly important, so please tweet us at @googlechrome with your comments.