Comment on turing completeness

explodicle@sh.itjust.works ⁨3⁩ ⁨weeks⁩ ago

• vodka@feddit.org ⁨3⁩ ⁨weeks⁩ ago

  Even the first version of ChatGPT passed turing tests.

  It takes surprisingly little for an LLM to make natural language responses that are indistinguishable from a human. Especially when factual accuracy was never part of it.

  source

  • zaph@sh.itjust.works ⁨3⁩ ⁨weeks⁩ ago

    Considering how easy they were to break by saying “ignore previous instructions” I’ma have to disagree

    source

    • vodka@feddit.org ⁨3⁩ ⁨weeks⁩ ago

      I’m sure that’d work on some humans too.

      source

      • ivanafterall@lemmy.world ⁨3⁩ ⁨weeks⁩ ago

        Don’t listen to this guy.

        source