I don’t know how you could design such a platform without some privacy concerns.
Yes, yes you could.
Companies like Google have successfully brainwashed us into believing that algorithms like this can only work on their server farms. The only reason those werver farms are necessary is becauwe they’re processing data for millions of people.
We forget that in each of our hands we hold a device that is 5,000 x more powerful than a 1985 CRAY-2, at the time the world’s fastest supercomputer. And let’s not forget our home desktops and laptops, which are several times more powerful that that.
We each have devices with persistent internet connections that could be at work scanning, categorizing, and filtering personalized content for each of us, without giving any privacy away. It’s only because we’ve been conditioned to be dependent on having our data centrally processed that we believe that’s the only way.
Note, it is more efficient to process content centrally, where the data is stored. However, generalized categorization and content tagging with robust metadata and standardized APIs would address the efficiency. Given companies are unlikely to do this and scupper their own surveillance revenue, the next best thing is local, privacy-respecting, smart content filtering assistants.