Eliminates annoyance with local machine learning in Chrome

A lot of the functionality that browsers and the web as a platform have gained over the past decade has likely been misused, sometimes to violate the privacy of browser users. Access to your webcam, microphone, and location data are among these features.

To reduce the potential for abuse, browsers require web applications to obtain users’ consent before granting them access to related functions. This is done through a notice (web notice).

In theory, this is all well and good, but in practice, it often means that users have to respond to many of these alerts. Not least it applies to requests from websites that want to notify users when something new happens.

Many people find this annoying.

local machine learning

Google is among the players trying to find a solution to this. In a blog post The company is writing this week about a new use of native machine learning in Chrome, which will adapt the browser’s user experience to previous user behavior.

One of the things that machine learning should be able to contribute is to assess the likelihood of a user agreeing to grant this website permission. It depends on how the user previously responded to such requests.

If the user rarely or never agreed to such permissions before, Chrome will stop serving the request in the usual way. Instead, it appears in a less invasive way in the address bar, where the user can choose to ignore it.

Chrome will also use local machine learning to adjust the buttons on the Chrome toolbar, based on what is believed to be most useful in the current situation. These buttons can be, among other things, for sharing web pages or for voice search. But the use should also be able to set this manually.

Google states that this functionality will be in the next version of Chrome, such as Chrome 103, as planned It will be released on June 21. The machine learning used runs locally on an individual device, which likely means that the results are out of sync with any other Chrome installations the user is using.