Conversation

Notices

1. Embed this notice
   Rich Felker (dalias@hachyderm.io)'s status on Thursday, 30-Jan-2025 21:25:47 JST Rich Felker

   One really stupid thing about OpenAI & fellow scum saying they couldn't have this "advanced" "AI" without letting them infringe copyright with impunity:

   The approach they're using, with massive GPU expenditure for grossly inefficient training, is unnecessary and stupid. The only advancement is brazenness. We could have made better chatbots 30 years ago if someone spent the money on labor digitizing everything written and if they didn't have to respect copyright.

   • Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Thursday, 30-Jan-2025 23:01:58 JST Rich Felker

     in reply to

     The massive compute is making up for unwillingness to program any explicit logic, not any necessity. You could do equally impressive parlor tricks with moderate degree Markov chains and smart explicit NLP logic.

   • Embed this notice
     Chris Vest (chrisvest@mastodon.social)'s status on Friday, 31-Jan-2025 11:01:07 JST Chris Vest

     in reply to

     @dalias such a system could also be hooked up to a knowledge graph with data provenance, and boom now you can ask for sources and actually get them

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Friday, 31-Jan-2025 11:01:07 JST Rich Felker

     in reply to

     • Chris Vest

     @chrisvest But that would have taken respecting and paying expert human labor rather than billionaire wet dreams that doing so is unnecessary and obsolete.