• Seankala@alien.topB
    link
    fedilink
    English
    arrow-up
    1
    ·
    11 months ago

    Nothing about this is novel though; the fact that language models are able to uncover sensitive training information has been a thing for a while now.