• jarfil@beehaw.org
    link
    fedilink
    arrow-up
    1
    ·
    edit-2
    1 year ago

    No, that’s why I said “Ideally”, meaning it as a goal.

    I don’t think we have the means to do it yet, or at least I don’t know of any platform working like that, but I have some ideas of how some of it could be done. Back in the days of Digg, with some people, we spitballed some ideas for social networks, among them a movie ranking one (that turned out to be a flop because different people would categorize films differently), and a kind of PageRank for social networks, that back then was computationally impractical. But with modern LLMs running trillions of parameters, and further hardware advances, even O(n²) with n=millions becomes feasible in real time, and in practice it wouldn’t need to do nearly that much work. With the right tuning, and dynamic message visibility, I think something like that could create the exact echo chambers that would attract X people, allow in des-X people, while keeping everyone else out and unbothered.

    Of course there is a dark side, in that a platform could use the same strategy to mold the opinion of any group… and I wouldn’t be surprised to learn that Meta had been doing exactly that.