I’m looking for a privacy-respecting open-source android keyboard, and so far I’ve found:

Does anyone have any experience with these (or other alternative keyboards)? Which one would you recommend?

  • CAPSLOCKFTW@feddit.de
    link
    fedilink
    arrow-up
    19
    arrow-down
    1
    ·
    1 year ago

    I use FlorisBoard. It’s not perfect, but it has a fully featured clipboard and undo-redo functionality. Missing the swipe tho.

      • nottheengineer@feddit.de
        link
        fedilink
        arrow-up
        1
        ·
        1 year ago

        But it’s quite bad, given their care for privacy there’s no way they can compete with google who has all the data.

        I know a bit of machine learning and I’m pretty confident I could hack something together that’s better than what florisboard has right now, but I’d have no idea where to even start integrating it.

        • zerodawn@leaf.dance
          link
          fedilink
          arrow-up
          4
          ·
          1 year ago

          Agreed there isn’t any beating google, but everyone operates on different threat models. A bit of inconvenience for a bit of privacy may be worth it for some and not for others. From what little i used of it when i was trying it that yes it had much room for improvement but it was still useable. Once they get predictive text enabled i think the a swipe feature will really shine. But then again that’s just my humble opinion

          • nottheengineer@feddit.de
            link
            fedilink
            arrow-up
            2
            ·
            1 year ago

            Google’s swipe typing is pretty damn good, but it isn’t magic. It gets a lot better after using it for a while so I’m sure they use reinforcement learning in addition to the dictionary.

            They’ve had it for quite a while and it’s good, so I doubt they are putting much work into improving it. And since we have phones with TPUs and multimodal LLMs now, I think it’s possible to beat google.

            The LLM would only need 3-10 words of context and the swipe data as input to generate a single word, so it can be very small. I don’t know much about the power of cellphone TPUs, but I think training an LLM with about 10M parameters on the fly should be possible. If that’s the case, we could beat google while doing everything locally on the phone, so no privacy compromises.

            Now that I think about it, it sounds doable. But then again I never did anything like this so I’m probably underestimating it by an order of magnitude or two.

            • zerodawn@leaf.dance
              link
              fedilink
              arrow-up
              2
              ·
              1 year ago

              Beating google while being local would be the dream, wouldn’t it? What you say tracks but i’m like you, i’ve not done anything remotely like it so it’s very possible we’re underestimate indeed.