Conversation

Notices

1. Embed this notice
   Casey (kc@chaos.social)'s status on Friday, 16-Feb-2024 16:56:24 JST Casey

   in reply to

   Like language models, they will be used to discriminate against marginalized people, including people with disabilities. They will reproduce stereotypes in never-before seen ways.

   Unlike language models, they have rather sparse uses for #accessibility. Maybe in the future, they will be able to generate accurately signing sign language avatars, but that’s also something nobody wants or has ever asked for except for some weird nondisabled techbros.

   • Embed this notice
     Casey (kc@chaos.social)'s status on Friday, 16-Feb-2024 16:56:25 JST Casey

     I am extremely terrified about OpenAIs new video generation capabilities with their new Sora model.

     With my particular vision impairment, I'm unable to really notice the inaccuracies in just about any single one of their examples, except where they highlight its weaknesses specifically. Even when others point them out to me.

     That said, I did see this coming. It's really about time we have some public conversation of whether we really want this. #inclusion #a11y

     GreenSkyOverMe (Monika) repeated this.