A machine learning librarian at Hugging Face just released a dataset composed of one million Bluesky posts, complete with when they were posted and who posted them, intended for machine learning research.

Daniel van Strien posted about the dataset on Bluesky on Tuesday:

“This dataset contains 1 million public posts collected from Bluesky Social’s firehose API, intended for machine learning research and experimentation with social media data,” the dataset description says. “Each post contains text content, metadata, and information about media attachments and reply relationships.”

The data isn’t anonymous. In the dataset, each post is listed alongside the users’ decentralized identifier, or DID; van Strien also made a search tool for finding users based on their DID and published it on Hugging Face. A quick skim through the first few hundred of the million posts shows people doing normal types of Bluesky posting—arguing about politics, talking about concerts, saying stuff like “The cat is gay” and “When’s the last time yall had Boston baked beans?”—but the dataset has also swept up a lot of adult content, too.

  • foremanguy@lemmy.ml
    link
    fedilink
    English
    arrow-up
    23
    ·
    4 hours ago

    If you post something publicly, that thing will be used to train AI. Nevertheless the privacy speaks of the company.

    • Brumefey@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      2
      ·
      3 hours ago

      I don’t know why social media are used for training. It’s like the worst quality of data ever and it results to answers like « go kill youself » when prompted about something sad…

  • FaceDeer@fedia.io
    link
    fedilink
    arrow-up
    60
    ·
    7 hours ago

    I don’t know why anyone would be surprised about this. Bluesky is a distributed system using an open protocol. The whole point of it is that there’s no central control.

    Same goes for the Fediverse, of course. Everybody should be prepared for the “surprise” that all our posts and comments here are also being used for AI training purposes.

    • webghost0101@sopuli.xyz
      link
      fedilink
      English
      arrow-up
      12
      ·
      edit-2
      4 hours ago

      Lemmy AI is going to be one bean of a star trek meming communist. That if not running on linux ends any response with “install Arch btw” 💪 💪

    • mosiacmango@lemm.ee
      link
      fedilink
      English
      arrow-up
      11
      ·
      edit-2
      5 hours ago

      It’s not distributed, nor really designed at all like the fediverse. It is deeply centralized, and its architecture requires it to be centralized, or at least to have only huge players with a “gods eye view” for it to work.

      Atproto was initially designed as a straight drop in replacement for twitter, so its design makes sense, but its not at all like the Fediverse.

      One of the authorities of ActivityPub, the fediverse protocol, just did a very kind but still very blunt breakdown of Bluesky’s design choices. she is a big fan of the people involved and some of its positives, but it is not fediverse like, not at all. In her words, it doesn’t scale down, only up. You cant have a small bluesky server. To work, you need all data sent to everyone, on every instance. The data demands for just the current influx is TBs/month of data, and climbing (according to the link below, they use 16TB of nvme storage right now after the recent surge, which would be thousands /month on any cloud service. This will climb dramatically).

      All data being public is a design choice by Bluesky. It is also a different design choice by the fediverse that comes to the same outcome, but that does have an answer if we want it. I know gotosocial did something interesting to make fully private votes by using a empty shell profile that votes, but tying that in a tricky way to your account. So there are fediverse answers to privacy, but there may not be bluesky answers.

      EDIT: One of the blueksy/atproto devs replied to the above link today. The gist reinforces the point that the service is intended to be run by large orgs, including corporations, but also big non profits like the internet archive or Wikipedia. His take is that user experience is key, and for that you need big money and easy features. They are hoping that since the pieces of atproto can be hosted separately by separate giant orgs, that market forces will make it viable to be decentralized.

      • CleoTheWizard@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        3 hours ago

        From what I understand of the protocol, the federation just isn’t the same but provides some of the same benefits. Im not an expert, correct if wrong.

        Essentially when I looked into it, the main benefits are stuff I actually prefer as opposed to the current implementation on fediverse in some regards.

        The main idea being that users own their data on their own server (or collective server) and can choose to remove or take that data elsewhere to different apps or potentially even accounts. This is a lacking feature in the fediverse and it’s a common contention. If I get blocked on Lemmy or Mastodon, my data goes away. Especially since most people are not likely to host an instance themselves (since it’s an awful user experience) whereas BlueSky data can easily be stored by a third party that is trusted.

        But yes you’re right, this still promotes large platforms. However again it gives users more control over what they host on which platforms and keeps their data in one place. That’s a huge advantage imo.

        I don’t so much mind this future. It’s not quite the free speech platform that the fediverse is but it’s closer. Moderation can be much more lax and focus on TOS breaking or illegal things. And hey if at some point BlueSky is too woke or whatever the hell people say, they can literally pick up their server with their content and build an app elsewhere. The implementation is different but the end point is largely the same which is cool.

    • cm0002@lemmy.world
      link
      fedilink
      English
      arrow-up
      22
      ·
      7 hours ago

      I’ve said it before and I’ll say it again, if you post something publicly, expect it to be used publicly.

  • 👍Maximum Derek👍@discuss.tchncs.de
    link
    fedilink
    English
    arrow-up
    98
    ·
    edit-2
    8 hours ago

    That’s why I always pepper all my social media posts with misinformation.

    BTW, did you know most convenience stores offer free ATMs to anyone who can haul them away? You don’t even need to ask.

    • paraphrand@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      4 hours ago

      I keep noticing situations on social media and wikis were the only way I can frame it now is that it’s just data every for AI models.

    • _wizard@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      5 hours ago

      Wish I could. My job has me knee deep in it everyday trying to keep up with all this.

      • CrowAirbrush@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        4 hours ago

        Modern day life over here requires the internet, our government has taken away options that use paper and replaced it with websites. Same goes for banks.

    • Stovetop@lemmy.world
      link
      fedilink
      English
      arrow-up
      34
      arrow-down
      1
      ·
      edit-2
      8 hours ago

      There are definitely bots mining fediverse content as well. When the Reddit exodus was ongoing, there were entire Lemmy instances with no users but bots. Not posting or reposting, just…watching and waiting, I guess.

      Not that it’s of any consolation, just better to assume that nowhere is safe from being mined for AI training.

    • GhostlyPixel@lemmy.world
      link
      fedilink
      English
      arrow-up
      5
      ·
      6 hours ago

      Legality hasn’t stopped AI training in the past, I’d say they beg forgiveness instead of ask for permission, but they don’t even do that lol

    • FaceDeer@fedia.io
      link
      fedilink
      arrow-up
      4
      ·
      7 hours ago

      From the Bluesky TOS:

      Bluesky Social is available as a desktop application at bsky.app and bsky.social (each a “Site”) and a mobile application (“Bluesky App” or “the App”).

      These terms only apply to social networking that happens on Bluesky Social services, including the Sites and Bluesky App. If you’re using another social networking application on the AT Protocol that isn’t Bluesky Social (we call this a “Developer Application”), the developers of the other service will provide the terms and conditions that govern your experience.

      So looks like the Bluesky TOS simply doesn’t apply. Create a developer application and give it whatever training-friendly TOS you want.

      • Bilb!@lem.monster
        link
        fedilink
        English
        arrow-up
        1
        ·
        6 hours ago

        It looks like they’re considering adding some equivalent of a robot.txt to express consent or non-consent for posts on ATProto, but of course as they say:

        “Bluesky won’t be able to enforce this consent outside of our systems. It will be up to outside developers to respect these settings.”