• Buttermilk@lemmy.ml
    link
    fedilink
    English
    arrow-up
    3
    ·
    1 year ago

    Gradient decent does strike me as the sort of problem that would work amazingly on a quantum computer. I think the issue is byte limitations. This paper is certainly over my head, but aren’t we struggling to get a 265 bit number calculated through a quantum computer, and wouldn’t a 48 gigabyte language model be basically impossible with current methods?