Apparently, they did go rogue.

  • a_non_monotonic_function@lemmy.world
    link
    fedilink
    arrow-up
    1
    ·
    4 days ago

    There are new models. I guarantee you they’re trying to train them.

    What they’ve discovered is they’ve hit a wall in terms of the diminishing returns of just funneling more data in and doubling the size of the actual models.