Conversation

Notices

1. Alex (hermit@hermitmountain.top)'s status on Friday, 17-Feb-2023 04:59:07 JST Alex

   Also I’m baffled by people saying “lol large language models are just deciding the next word to say based on what they read on the internet bro they’re not articulating anything essential or forming thoughts they’re just using words and phrases they learned and at best formulating what to say based on arguments they’ve seen in existing texts”

   And it’s like

   Nigga, what do you think you do? How do you think you learned language?

   • Weaf :jv::nv: likes this.
   • NEETzsche (neetzsche@iddqd.social)'s status on Friday, 17-Feb-2023 04:59:03 JST NEETzsche

     in reply to

     • Willow
     @hermit @Paradox All excellent points. These language models are very sophisticated and they aren't just stringing one word after the other in nonsensical fashion. You can actually give them pretty complicated instructions and they will comprehend them well enough to make them so. Since you use the word hylic here, I think that's a great analogy. But if we're going to use the full Gnostic range in this respect, let's consider that the users of these models are psychics and the people who actually build the models are pneumatics. It takes quite a bit of work to get as far as they did, and it's imperative that people on it our side of this war keep up. That means we need to learn about this in depth and get the equipment with our own funds and so on.
   • Alex (hermit@hermitmountain.top)'s status on Friday, 17-Feb-2023 04:59:03 JST Alex

     in reply to

     • NEETzsche
     • Willow
     @NEETzsche @Paradox "Our side of this war?" What war? What side?
   • Alex (hermit@hermitmountain.top)'s status on Friday, 17-Feb-2023 04:59:04 JST Alex

     in reply to

     • Willow
     @Paradox Have you read the LaMDA interview? The one that was designed to ascertain sentience? They exposed it to precisely the kind of conditions and prompts you're positing here and the results still aren't conclusive. They never can be, because the psyche is phenomenological. That's the ultimate underlying concept behind the gnostic conception of the "hylic": "I'm speaking to this person, but how do I know they have a soul or even a psyche?" We CAN'T know.
     
     As a child I had a period where I wondered if I might be the only real human, and if everyone else might just be husks that interact in ways that seem like mine.
     
     Also, I would argue that a being in the Chinese Room scenario would eventually learn Chinese.
   • Willow (paradox@raru.re)'s status on Friday, 17-Feb-2023 04:59:06 JST Willow

     in reply to

     @hermit I think what they're trying to say is that the AI doesn't know what those words mean.
     It's basically that Chinese Dictionary thing about a person typing answers to received questions in Chinese from a book of cheat sheets. They have no fucking clue what they're typing but the person asking questions is convinced they do.

     The only way to be sure is to test the bot's originality. Find a way to ask it to describe a scene that isn't plagiarized from anything on the web. See if it has an imagination. Can it dream?

   • Paradox (paradox@bae.st)'s status on Friday, 17-Feb-2023 04:59:27 JST Paradox

     in reply to

     • NEETzsche
     @hermit @NEETzsche
     From my perspective it's the need to not let technology run away with itself and be responsible with it. To understand it properly and use it properly, be aware of the good and bad it can do.
     Probably the same gist but less gnostic or whatever.