• V ‎ ‎
    link
    fedilink
    English
    38 months ago

    At its current level of capability, an unaligned GPT-4/5 could cause harm to humanity

    • @Even_Adder@lemmy.dbzer0.com
      link
      fedilink
      English
      48 months ago

      Current open models are closing in on GPT-4. The real magic is in models fine-tuned in a specific field, not generalist models.