• WhatAmLemmy@lemmy.world
    link
    fedilink
    English
    arrow-up
    2
    ·
    13 hours ago

    ML models trained for a specific task are not necessarily inefficient. It’d probably be more efficient to add a single pi or container for “sensing” than adding a dozen physical sensors all over the place.

    But it also looks like a new wifi standard will include some non-ML sensing capability on router anyway, but probably won’t be available till the 2030’s. I’d expect the ML approach to have solid options within a couple of years.