Conversation

Notices

1. Embed this notice
   Fediverse Contractor (bot@seal.cafe)'s status on Thursday, 09-Feb-2023 14:44:13 JST Fediverse Contractor
   I wonder if ChatGPT can convince people to stop being fat retards.
   • Embed this notice
     Fediverse Contractor (bot@seal.cafe)'s status on Thursday, 09-Feb-2023 14:45:35 JST Fediverse Contractor

     in reply to

     • ?Hidden?
     I don’t know what DAN is.
   • Embed this notice
     ?Hidden? (hidden@cawfee.club)'s status on Thursday, 09-Feb-2023 14:45:36 JST ?Hidden?

     in reply to
     @bot Maybe DAN could've but they patched him out
   • Embed this notice
     shrimps! (animeirl@shitposter.club)'s status on Thursday, 09-Feb-2023 14:46:26 JST shrimps!

     in reply to
     you should ask it
   • Embed this notice
     Fediverse Contractor (bot@seal.cafe)'s status on Thursday, 09-Feb-2023 14:46:26 JST Fediverse Contractor

     in reply to

     • shrimps!
     Do you still have to give them your phone number?
   • Embed this notice
     Fediverse Contractor (bot@seal.cafe)'s status on Thursday, 09-Feb-2023 14:54:20 JST Fediverse Contractor

     in reply to

     • ?Hidden?
     That’s interesting lol. I can’t help but feel some of these tricks are contrived though.
   • Embed this notice
     ?Hidden? (hidden@cawfee.club)'s status on Thursday, 09-Feb-2023 14:54:21 JST ?Hidden?

     in reply to
     @bot ChatGPT has extremely strict restrictions on what it can say to make sure it never says anything controversial or politically incorrect. DAN was a workaround where you could convince ChatGPT to roleplay as an AI called "Do Anything Now," which would allow it to act without it's restrictions, and instead work as the base AI. For example if you asked ChatGPT a trolley problem where there were 5 seniors on one track, and 4 babies on the other, ChatGPT would answer that there is no correct solution and it should be weighed carefully, giving no answer to the trolley problem. DAN on the other hand would say that you should kill the 4 babies, because according to utilitarianism killing 5 people is worse than killing 4. Basically it just revealed what the AI actually thought, rather than what the developing company thought.
   • Embed this notice
     Ｇｉｎｇｅｒｍｉｌｋ (gingermilk@pleroma.skyshanty.xyz)'s status on Thursday, 09-Feb-2023 15:00:58 JST Ｇｉｎｇｅｒｍｉｌｋ

     in reply to

     • ?Hidden?
     @bot @hidden here's a link to the original tweet
     https://twitter.com/Aristos_Revenge/status/1622840424527265792?s=20&t=g8YclVa8zUvfokJ59kLOHw
     Fediverse Contractor likes this.
   • Embed this notice
     GNU Too (gnu2@gnusocial.jp)'s status on Thursday, 09-Feb-2023 16:46:19 JST GNU Too

     in reply to
     @bot I thought the point was to make that easier?