• aallsbury@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    1 year ago

    My theory… they cracked AGI and have been running it in their backend for a bit. That’s what is training GPT5. It also explains weird data thats been coming out that appears to show the base gpt4 model remembering context across different threads, as well as some odd statements Altman has made about the AI learning from conversations. The board found out, and realized he was lying to the board, the gov, and the public. Fired. JUST A THEORY

    • nxqv@alien.topB
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      weird data thats been coming out that appears to show the base gpt4 model remembering context across different threads,

      Can I read more about this anywhere?