Conversation

Notices

1. Nire Bryce (nirebryce@hachyderm.io)'s status on Wednesday, 29-Mar-2023 11:40:09 JST Nire Bryce

   remember that pushing for stronger copyright protections re: "AI" only works if you push for it to apply retroactively, otherwise OpenAI just trained their models on all of this and all of their competitors can't.

   shit sucks.

   • Adrian Cochrane repeated this.
   • Nire Bryce (nirebryce@hachyderm.io)'s status on Wednesday, 29-Mar-2023 11:49:12 JST Nire Bryce

     in reply to

     what is it, $250,000 per infringement that they hit filesharers with? I figure it's gotta be at least 100k people they've violated the copyright of, minimum, right?

     25 billion seems like reasonable damages, but I'm sure they'd find a way to push it higher by lying to the judge

     Adrian Cochrane repeated this.
   • Nire Bryce (nirebryce@hachyderm.io)'s status on Wednesday, 29-Mar-2023 12:01:54 JST Nire Bryce

     in reply to

     this won't happen though because they aren't the 'small guys' so be careful pushing copyright strengthening without any sort of punishment. I've heard rumors openAI is lobbying for it, for exactly these reasons, and I'm willing to tentatively believe that at this point with how it's suddenly hit the broader media.

   • Adrian Cochrane (alcinnz@floss.social)'s status on Wednesday, 29-Mar-2023 12:29:44 JST Adrian Cochrane

     in reply to

     • Josias

     @jos @NireBryce Baldur Bjarnason suggests that regulation:

     1. Clarifies that those running the service are responsible as a publisher for the text their AIs spit out. To the same degree as they are for employees.
     2. Requires disclosure of the use of LLM technologies.
     3. Training sets should be opt-in & (I think more importantly) public.

     https://www.baldurbjarnason.com/2023/regulating-ai/

   • Josias (jos@freeradical.zone)'s status on Wednesday, 29-Mar-2023 12:29:48 JST Josias

     in reply to

     @NireBryce Maybe going in the opposite direction could help: abolish copyright entirely or decrease restrictions, and require OpenAI to release all the code and data for their models.