OpenAI wants to stop ChatGPT from validating users’ political views

Comments

• UnderpantsWeevil@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  OpenAI is full of shit, as usual. They want engagement and they need revenue. They’ll do whatever it takes to get those things. Political ideology and sociological ramifications are an afterthought.

  source

  • yakko@feddit.uk ⁨3⁩ ⁨weeks⁩ ago

    Private equity having a normal one

    source
  • msage@programming.dev ⁨3⁩ ⁨weeks⁩ ago

    Political ideology and sociological ramifications are an afterthought

    for sale

    source
• tekato@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  The reason why I can’t stand LLMs is because they congratulate me before replying to anything I say. This could be a good thing.

  source

  • londos@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

    That’s a great point – and an important distinction.

    source
• ikidd@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  Somewhere, sometime, ChatGPT is telling the stupidest person in the world “Yes, you are absolutely correct!”

  source
• noxypaws@pawb.social ⁨3⁩ ⁨weeks⁩ ago

  Pretty telling that the headline is that they want to do something, rather than they did something.

  source
• latenightnoir@lemmy.blahaj.zone ⁨3⁩ ⁨weeks⁩ ago

  “”“”““Wants””“”“”

  source
• tal@lemmy.today ⁨3⁩ ⁨weeks⁩ ago

  You don’t have to do so, but given that people seem to like echo chambers, if you don’t, I bet that a competitor will.

  source

  • otp@sh.itjust.works ⁨3⁩ ⁨weeks⁩ ago

    Literally the motivation behind Grok.

    “ChatGPT doesn’t lie in the way that puts us in a positive light. So I’ll make my own AI to do just that!”

    source
• paraphrand@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  How does one even do that in a conversation centered on such things? These are conversations with the bot. So it’s analogous.

  Does it just refuse to interact? Does it just info dump and list all sides of the issue exhaustively?

  I’m sure I’m working under a weird limited view, but it just does not seem possible to do this without things being awkward, or, without it being designed around its own agenda.

  source

  • avatar@lemmy.zip ⁨3⁩ ⁨weeks⁩ ago

    The easiest example is probably similar to an objective out of the loop post response.

    source
• PK2@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

  Don’t they watch South Park: if they want to know how to do that, “Dude, just ask ChatGPT”.

  source
• monogram@feddit.nl ⁨3⁩ ⁨weeks⁩ ago

  Expectation: AI stops congratulating you

  Reality: AI calls you a gay fag for even asking commie questions.

  source