@njoseph I absolutely understand why they would want to try this. Silos invade privacy trying to gleam some of your browser history in order to recommend things to you and profit from biasing that.
It's a real problem that browsers are in a position to solve. But it actually has to be better, which is challenging.
I certainly want to eventually implement it myself in my own browser, and I think I know how to do it. But I really need to talk to cryptographer first, I can't afford mistakes.
@alcinnz I wouldn't complain if they showed their suggestions based on the current page only without trying to personalize it. Just like how #DuckDuckGo shows their ads based on the search query only.
If the browser directly sends requests to the third party AI startup on every page visit, it's equivalent to giving them your entire browsing history, one entry at a time.
#Mozilla should serve requests directly, with a cache of recommendations that get updated by the third party frequently.
@njoseph To me that sounds like it'd be simpler clientside than what they described. And it matches the UI design they're going for.
I don't know about serverside though.
This is a general instance supporting toots in English and తెలుగు.
Hero image credit: Sean O'Brien (CC BY-NC-SA 3.0)