Bluesky's open API means anyone can scrape your data for AI training | TechCrunch
techcrunch.com
Bluesky might not be training AI systems on user content, but there's little stopping third-parties from doing so.
  • @[email protected]English
    26•3 days ago

    Can’t really have an open federated protocol if you want to be able to prevent scraping 🤷

    • BETYU
      2•3 days ago

      the protocol might be open but joining your own server is not open so not really like the fediverse.

      • @[email protected]English
        1•2 days ago

        Weelll, depends on what you mean with “joining your own server.” The model is definitely different from ActivityPub’s, but lots of people on the network already have their own PDSs or “personal data servers”. They do still go through Bluesky’s relay, but nothing is really stopping people from running relays as such, it’s just fairly costly (as in some hundreds of dollars per month) currently as they need to hold the full state and history of the network (but apparently that’s being worked on.)

        But it’s definitely a federated protocol even though it’s different from AP

        • BETYU
          1•2 days ago

          but you kind of said yourself that in practice its not really.

              • @[email protected]English
                1•1 day ago

                So just because someone else isn’t running a relay now while they figure out how to make them less costly to run, it’s pointless for the protocol to be partially federated already and support further federation?

  • LvxferreEnglish
    5•3 days ago

    This isn’t exactly surprising, is it? Odds are that Lemmy is also being scrapped the shit out of.

    And it would be fine if it was a bunch of nobodies training their homebrew small language models, for the sake of whatever.

    Except that it isn’t - it’s a bunch of big arse companies, with a “NEED MOAR DATAS!” approach, and more than enough money to bake the already too warm planet, since they struggle with the fact that those “things” called “humans” care about consent. “This thing didn’t opt out, so training on its data is fair game!”. Just to shove the tech back into the thing’s throat, in the hopes that it makes the tech eventually profitable.

    …I guess that my point is that this should be handled legally, not through closing down the protocol. The issue is not people scraping it, but who does it, and why.

    • @buddascrayonEnglish
      2•2 days ago

      Anyone who doesn’t believe that literally everything and anything they post on the internet is being scraped for LLM’s is an idiot.