• nonentity@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    14
    arrow-down
    5
    ·
    14 hours ago

    Thanks for proving my point, your service is invaluable.

    LLMs are explosively imprecise, statistically luke-warm, grey goo extrusion sphincters. They can’t produce anything novel, at best they can reveal potentially plausible information from the data they were trained on, which then requires proven domain expertise to validate.

    Intelligence requires comprehension, a trait LLMs are intrinsically incapable of performing.

      • nonentity@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        6
        arrow-down
        3
        ·
        9 hours ago

        Understanding, context[0].

        Computers are particularly adept at iteration, and LLMs are an application that leverages this effectively. To wit, I’ll concede LLMs are ‘AI’, where that initialisation represents Accelerated Iteration, or in management as Amplified Incompetence, but as an example of Intelligence, that remains for those who are proud of their lack of the naturally occurring variety.

        [0]: to frame a larger memory pool as a ‘context window’ is a misnomer, as context requires understanding and comprehension.