Public AI Inference Utility
publicai.co
A nonprofit, open-source service to make public and sovereign AI models more accessible.

Found from Bruce Shneier’s blog. This model is free, ad-free, privacy respecting, and likely to stay that way. If you or folks you know are heavily using GPT, and likely to be hurt when it starts introducing ads (and otherwise enshittifying) soon, do make sure they know there are alternatives like this.

This particular chat model uses a system prompt chosen by the swiss government, with the intention of providing LLM access as a public utility (like a library). I believe models are intentionally trained on ethical datasets (see the details of Aptertus here), with an effort towards sustainable energy use.

  • Dekkia@this.doesnotcut.itEnglish
    199·
    9 days ago

    This is the definition of ethically sourced data from the Apertus website:

    […] the training corpus builds only on data which is publicly available.

    So they still train on Websites, Blogs and Social Media. Ethical my ass.

    • theherk@lemmy.worldEnglish
      132·
      9 days ago

      That is at least an improvement over including in its corpus the entire worldwide collection of copyrighted materials.

      • Dekkia@this.doesnotcut.itEnglish
        41·
        9 days ago

        But the other stuff is copywritten as well most of the time.

        Just because it’s free to look at doesn’t mean it’s free to download, modify or feed into an AI.

        • theherk@lemmy.worldEnglish
          2·
          9 days ago

          Yeah, for sure. I’m not saying it is good at all, just that scraping some proportion of copyrighted material is an improvement over scraping all the copyrighted material.

    • Artisian@lemmy.worldOPEnglish
      62·
      9 days ago

      Let’s include the whole paragraph at least.

      Apertus was developed with due consideration to Swiss data protection laws, Swiss copyright laws, and the transparency obligations under the EU AI Act. Particular attention has been paid to data integrity and ethical standards: the training corpus builds only on data which is publicly available. It is filtered to respect machine-readable opt-out requests from websites, even retroactively, and to remove personal data, and other undesired content before training begins.

        • Artisian@lemmy.worldOPEnglish
          2·
          8 days ago

          As I read it, data must be available according to swiss copyright law, not personal, available using the open web. Further, they retroactively respect opt-out requests.

    • kalkulat@lemmy.worldEnglish
      1·
      8 days ago

      Sounds like a ripping good way to keep corporate data (and government secrets) from the public radar.

      That way we won’t find out whose hands public taxdollars (or public-owned structures rented to corporations) wind up in.