@Gargron Yup. They're not intelligence. They're not hallucinating. They're not making new connections.
They're data-center scale autocorrect.
Top-level
@Gargron Yup. They're not intelligence. They're not hallucinating. They're not making new connections. They're data-center scale autocorrect. 13 comments
@drsbaitso @Gargron Data Center auto correct is such a good fit, and is exactly my experience. Chatbots are so unreliable you have to check each and every answer, and that makes using them entirely pointless. @drsbaitso @Gargron Not at all, language models build an internal world model and understand what they're talking about. @drsbaitso If you want to talk with adults you're going to need actual arguments instead of "lalala I can't hear you" @sicko You're probably right that the racism was a bad choice. But don't think I missed you using the n word to conclusively lose the argument. |
@drsbaitso @Gargron Awhile back, I asked GPT-4 to come up with a meme in response to this viewpoint