• sourdough@lemmy.world
    link
    fedilink
    English
    arrow-up
    4
    ·
    18 hours ago

    They would need to be trained for open ended creative tasks, which is just hard in the current reinforcement learning paradigm.