Login
You're viewing the functional.cafe public feed.
  • Nov 12, 2026, 3:49 PM

    @AstralAudio @zarfeblong Feel free to shut me down on this point, as I'm speaking outside of my specialty here, but it seems to me that it could end up being better for accessibility if fewer investment dollars were going to big, extractive, general purpose AI projects, making it possible to argue for their redirection to purpose-built accessibility projects that don't get muddied up with other imperatives.

    💬 1🔄 0⭐ 0

Replies

  • Nov 12, 2026, 4:32 PM

    @lrhodes Three points it would benefit you to research to draw your own conclusions. I'm no expert so just as in the dark:

    1. JAWS for Windows (specifically, how it uses smart AI features).

    2. How blind users are vibe coding projects for accessibility in general as well as mods for gaming. (Not touching the gaming ports which I disagree with on fundamental levels.)

    3. Big models vs open source. (Gosh do I advocate for the latter but realities ... bleh.) @zarfeblong

    💬 1🔄 0⭐ 0
  • Nov 12, 2026, 4:39 PM

    @AstralAudio @zarfeblong I'll definitely look into the others, but I don't think there's really any such thing as an "open source" LLM. There are open weight models, but the ethical improvements of those models tend to be overstated, in part because open weight models are good at obscuring impacts that are more visible in cloud-based models.

    💬 1🔄 0⭐ 0
  • Nov 12, 2026, 5:51 PM

    @lrhodes @AstralAudio @zarfeblong …and there's nothing "open" about so-called "open weight" models, they're just downloadable binaries. You don't know how they were trained, you often don't have the training data either. It's industry openwashing.

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0