Conversation

Notices

1. Embed this notice
   翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Wednesday, 20-Sep-2023 21:18:15 JST 翠星石

   in reply to

   • Aaron Wolf
   @wolftune Except how such software on someone else's server only tends to be trained on information that was already under a free license or at least was publicly available - the proprietary works that one may wish to synthesize usually aren't inputted.
   
   I doubt that any court will conclude that a language model that regurgitates input text or images combined in interesting ways (there's a reason why microsoft didn't train co-pilot on any of its proprietary software, or the private repos of businesses after all) makes copyright no longer apply .
   
   
   Proprietary copyright licenses do hamper creativity, but free, copyleft licenses don't, so while copyright is still a thing, it's important to keep the copyright status crystal clear and use copyleft licenses, so businesses can't easily ruin everything.
   • Embed this notice
     Aaron Wolf (wolftune@social.coop)'s status on Wednesday, 20-Sep-2023 21:18:16 JST Aaron Wolf

     The #LLM, "AI" stuff is doing the same process as people do when we adapt and improve things in a free/libre/open (FLO) fashion.

     Provocative take: the *key* feature in #AI progress today is its *ignoring* of copyright limitations (because copyright hampers creativity, and ignoring it allows new valuable syntheses of ideas).

     (dangers are another matter, but point is human-driven #FLO creativity could also be freed from copyright, we just need *different* laws for the problems this brings up)