Conversation

Notices

1. Embed this notice
   Nelson (skyfaller@jawns.club)'s status on Wednesday, 01-Jul-2026 22:55:18 JST Nelson

   • Evan Prodromou

   @evan Yes, but some bad ideas may be cognitohazards and you don't need to deeply understand them to get why they're bad and worth avoiding.

   I reluctantly cite Maciej (who himself seems to have fallen down some fascist rabbit holes) on superintelligence: https://idlewords.com/talks/superintelligence.htm

   I like his discussion of "The Outside Argument", where you don't need to study superintelligence, simply look at "What kind of person does sincerely believing this stuff turn you into?" Superintelligence nuts suck.

   • Embed this notice
     Evan Prodromou (evan@cosocial.ca)'s status on Wednesday, 01-Jul-2026 22:55:17 JST Evan Prodromou

     in reply to

     @skyfaller this is a good talk and I like the framework but his argument was more convincing in 2016 than 2026.

   • Embed this notice
     Evan Prodromou (evan@cosocial.ca)'s status on Wednesday, 01-Jul-2026 23:06:50 JST Evan Prodromou

     in reply to

     @skyfaller We also see AI being used for software development as one of its primary use cases, including by the people developing the AI systems.

   • Embed this notice
     Nelson (skyfaller@jawns.club)'s status on Wednesday, 01-Jul-2026 23:06:51 JST Nelson

     in reply to

     • Evan Prodromou

     @evan why? If anything we've seen even more shitty behavior from superintelligence bros, starting with Elon committing mass murder through DOGE

   • Embed this notice
     Evan Prodromou (evan@cosocial.ca)'s status on Wednesday, 01-Jul-2026 23:41:04 JST Evan Prodromou

     in reply to

     • Ethan Black

     @golemwire @skyfaller I think the idea of the "woke mind virus" is kind of similar, from the other perspective.

     https://www.wsj.com/tech/elon-musk-woke-mind-virus-41576aa6

   • Embed this notice
     Ethan Black (golemwire@social.golemwire.com)'s status on Wednesday, 01-Jul-2026 23:41:05 JST Ethan Black

     in reply to

     • Evan Prodromou
     @skyfaller@jawns.club
     "Cognitohazards"... if we're looking at things that way, wouldn't the "other side" be viewing a lot of "your side"'s views that way? In that case, how would one be able to know the ideas they choose to not think about aren't the right ones? Because the "other side" could be doing the same thing.
     @evan@cosocial.ca
     
   • Embed this notice
     Evan Prodromou (evan@cosocial.ca)'s status on Wednesday, 01-Jul-2026 23:50:29 JST Evan Prodromou

     in reply to

     @skyfaller

     Are LLMs becoming more intelligent? Yes.

     "LLMs certainly aren't intelligent" - We don't understand the nature of intelligence, its origins or its definition. It is probably a spectrum -- a human is more intelligent than a wombat which is more intelligent than some gneiss. "Not intelligent" might not be a meaningful term.

     "Self improving" - The use of LLMs provides training data for the next generation of LLMs.

   • Embed this notice
     Nelson (skyfaller@jawns.club)'s status on Wednesday, 01-Jul-2026 23:50:31 JST Nelson

     in reply to

     • Evan Prodromou

     @evan Are you saying that spicy autocomplete is on the road to superintelligence? LLMs certainly aren't intelligent or recursively self-improving, instead feeding them to themselves leads to model collapse, which is the opposite. LLMs are basically irrelevant to this talk, except as a flag for who has had their brain eaten.

   • Embed this notice
     Evan Prodromou (evan@cosocial.ca)'s status on Wednesday, 01-Jul-2026 23:51:28 JST Evan Prodromou

     in reply to

     @skyfaller Question for you: do you think artificial intelligence is impossible, or just impossible with LLMs?

   • Embed this notice
     Evan Prodromou (evan@cosocial.ca)'s status on Thursday, 02-Jul-2026 00:59:07 JST Evan Prodromou

     in reply to

     @skyfaller good answer!

   • Embed this notice
     Nelson (skyfaller@jawns.club)'s status on Thursday, 02-Jul-2026 00:59:09 JST Nelson

     in reply to

     • Evan Prodromou

     @evan I think intelligence is a fake idea, promoted by eugenics boosters: https://aworkinglibrary.com/writing/toolmen

     If you are like me, and you were told you were very intelligent from a young age, it may be hard to let that part of your identity go, but I think we've seen enough damage from the concept of intelligence to discard it and reconsider.

     So to answer your question, I think intelligence is imaginary, so artificial intelligence is just extra bullshit on top, causing further harm.