Conversation

Notices

1. Embed this notice
   Sam Strong (samstrong@hachyderm.io)'s status on Monday, 03-Jun-2024 00:01:56 JST Sam Strong

   • Thomas 🔭🕹️

   @thomasfuchs Should the intent of the person posting the image matter? Post text + image give context? Should alt text be neutral?

   (Not that I think any of this “AI” stuff is going to end well, but I’m always curious about thoughts and opinions on alt text)

   • Embed this notice
     Sam Strong (samstrong@hachyderm.io)'s status on Monday, 03-Jun-2024 00:23:24 JST Sam Strong

     • Thomas 🔭🕹️

     @thomasfuchs Yeah the iOS photo identification stuff varies incredibly in accuracy. Most folks don’t even know the feature is there either.

     I doubt any of this is going to accurate enough to consistently match the tone of any given post. It reminds me of those gestures they added to the web cam software built into Mac OS that got in the news when folks accidentally set off celebratory reactions to bad news.

   • Embed this notice
     RevK :verified_r: (revk@toot.me.uk)'s status on Monday, 03-Jun-2024 02:05:57 JST RevK :verified_r:

     • Thomas 🔭🕹️

     @thomasfuchs one irony is if an image contains text the AI can usually read it. But if the body text of a post has used Unicode to make bold or serif, or some such, a lot of screen readers seem unable to read “what it looks like to a sighted person”. Some simple AI there may be sane!

   • Embed this notice
     cayleyh (cayleyh@hachyderm.io)'s status on Monday, 03-Jun-2024 03:41:23 JST cayleyh

     • Thomas 🔭🕹️

     @thomasfuchs this is something I would argue should be provided by the OS for all images / media. Feels like a great use of ML to improve accessibility for all similar cases.