Developing a More Helpful Browser with ML Tech
At Google, they leverage techs such as Machine Learning (ML) to develop more beneficial products, i.e., from filtering email spam to providing relevant search results. Google also leverages ML to create web images more accessible to users who are blind or have low vision and generate real-time captions for online videos in service of a user in noisy surroundings and those who have a hard time hearing.
This Chrome work continues, so they wanted to share a few future ML enhancements that provide users a safer, accessible, and more customized browsing experience. Moreover, these updates are powered by on-device ML models, which means our data stays private.
Less Vexatious Prompts and More of Mind Peace
Safe Browsing in Chrome helps protect billions of devices each day by signaling warnings when users try to navigate to dangerous websites or download corrupted files. In addition, in March 2022, Google rolled out a novel ML model that determines 2.5 times more potentially hazardous sites and phishing attacks than the prior models, resulting in a safe and secure web.
For further improvement of the browsing experience, Google is also evolving how users interact with web notifications. On the one hand, page notifications aid in delivering updates from sites you care about; on the other, notification permission prompts can become annoying. To assist users in browsing with less interruption, Chrome predicts when permission prompts are unlikely to be granted related to how people previously interacted with similar prompts and silences the undesired ones.
A Browser Just for Us
Maybe you like to read news articles in the morning with your phone in one hand and the other holding a cereal spoon, so you share many Chrome links. Or maybe voice search is more your thing, as you ask a few questions during your journey to work. Either way, Google wants to ensure Chrome is meeting us where we are, so in the coming years, we'll be leveraging ML to adjust the toolbar in real-time, displaying the most suitable action in that scenario.