The Firefox Test Pilot program has been the home of many interesting projects over the past couple of years, including its recent debut on mobile devices in the form of Lockbox and Notes. Today, the Firefox team is announcing a new experiment called Advance, an AI-powered service that tries to predict which pages you'll want to visit next.
To develop Advance, Mozilla partnered with Laserlike, a machine learning startup which has also built its own content search, discovery and personalization platform. With this power, Firefox will be able to suggest web pages that might be interesting to you based on your interests, which it will continuously learn as you interact with the experiment and the web. This will work for things such as recipes, blog posts, and news stories, content that lends itself to discoverability.
Mozilla is already known for its concerns regarding user data privacy and safety, and even as it's using AI and machine learning, it continues to be careful with how it treats the data collected by the service. The experiment is opt-in - you can start trying it from this page - and users can pause the data collection at any time, in addition to being able to view and delete all of it as well.
Earlier this year, the team introduced sponsored content powered by Pocket, and at that time, too, it promised that its approach was wary of privacy concerns, with data collection being done locally rather than being sent to a server. The organization is also investigating a machine learning technique that personalizes content locally by only sharing generated models with a server, rather than the data used to produce said modes.